Bitcoin Forum

I have been working with 2 other members of the bitcoin community to develop a custom FPGA mining board. Our goal was a quiet, efficient, highly profitable platform for bitcoin mining… with a much lower pricetag than previous FPGA boards.

http://lh4.ggpht.com/-X9trDy0Z9IE/Tk0rIrKkSKI/AAAAAAAAAYM/lVss4ZKsCcM/IMG_1077_thumb.jpg?imgmax=800

We have made and tested a prototype with pool mining. With the most recent program it gets 100 MHash/s and uses 6.8W of power. In the future we expect incremental improvements to the mining speed due to better software. You can see from the Mining Hardware Comparison (https://en.bitcoin.it/wiki/Mining_hardware_comparison) that our custom board is far better than the commercial FPGA alternatives in terms of $/Mhash.

The primary members of this FPGA team were:
fpgaminer, USA, Engineer
li_gangyi, Singapore, Electronics Hobbyist
newMeat1, USA, Engineer

We also want to thank makomk for his contributions to optimizing the code and NF6X for his advice on communication interfaces. It is a testament to the global nature of the bitcoin community that we were able to bring together such great talent from across the world.

Check out Li’s blog on the development of the board for more pictures! (http://fillwithcoolblogname.blogspot.com/2011/08/heres-where-its-at-now.html)

We have 4 boards for sale at this moment. The cost is $440 equivalent based on the MT Gox weighted average, to be paid in BTC. Shipping costs will be added. We do international shipping—please message me if interested and we’ll work out the details. Up-front payment is required to cover our hardware costs. The boards will ship within 7 days after receipt of payment.

In the next few weeks, we will be putting together a bulk order, so don’t worry much if you miss out on this initial batch.

What do you need to run these boards on your computer?
•   An open USB port
•   An open 4-pin Molex power connector. Most computers have an extra one or two of these inside (my PC has 4!)
http://images.bit-tech.net/content_images/2008/04/thermaltake_toughpower_1500w_w0171_psu/molex-8.jpg
•   Download and install Xilinx's free ISE software, here (http://www.xilinx.com/support/download/index.htm)
•   A Xilinx Platform Cable (http://cgi.ebay.com/Xilinx-FPGA-CPLD-USB-download-Cable-JTAG-programmer-new-/250831239457?pt=LH_DefaultDomain_0&hash=item3a66b4f921#ht_2447wt_993)- we recommend eBay to get it cheap
•   Operating system: Windows, Red Hat Enterprise Linux WS and SUSE Linux Enterprise are supported
o   Other versions of Linux may work, but no guarantees

We'll be posting a walk-through soon about how to set up the board. If you got the bitcoin client running, you shouldn’t have any problems setting up the FPGA board  :)

Get yours now- the Bitcoin Dominator X5000!
https://i.imgur.com/r0ekF.png

Give me a price. Let me know about bulk orders too in terms of price, qty and shipping to LAX.

It might take me a while to respond to a message because I can only reply to 10 msgs per hour. Sorry about that. We're only selling 1 to each customer at the moment, to help spread the word. If you want to buy all 4, go for it! We will certainly discount shipping and maybe we can work on another discount for you

It will take us at least a week to work out the details of the bulk order to come.

question though, if it only uses 6.8W of power, why not give it a male molex connector + female usb connector so that you can stack these cards?

you should look into optimizing these so that you can build custom boxes to store rows of these without needing tons of molex connectors and usb ports

This design is fairly modular already. Yes, you need multiple USB ports and Molex connectors, but those are cheap. We went for a simple solution first, to make sure it would work. Communicating over USB isn't trivial

Down the road... we have some changes in store, along those lines.

While stackable would indeed be nice, you can use splitters for the time being. Each rail on a regular computer power supply is 12v @ 5a, or about 60 watts. Even the cheapest power supplies have two rails with regular molex connectors for a total of 120 watts. If you load both rails to 80% (96 watts) you could put 7 boards on each rail, thus pulling 1.4 GH/s for less power draw than a single 5830. Of course right now that'd cost you $3,080 but who knows, given enough time price could very feasibly come down :)

Edit: oh yeah, and a couple USB hubs. I'm assuming the amount of bandwidth these things take is trivial, right?

Yeah. I don't know exactly (not my part of the project), but Xilinx says you can run 127 from one computer.

Jtag itself is fairly scalable, you can chain multiple devices into the loops I.E use the same Xilinx Platform cable to manage multiple boards with a modified harness.

I've yet to test this out properly though, so it'll take some time to implement (software side).

The cool thing about this minimalist design is that, with this as the base, we can add functionality later without having to ditch the entire board and dropping support. This avoids us having to re-spin the entire board, going through all the testing processes and then screwing the previous buyers over.

And yes, the bandwidth required is trivial.

Shakaru- looks like shipping to the US will be $10 for one.

Awesome! I'm almost buying it. Unfortunately you will not allow using an escrow service for my protection, because you need the money up front, right?

What do you think, Li? He's a senior member. Should we let our first buyer use an escrow service? I think we have enough dough to cover the first 1 or 2 out-of-pocket

Please message me, molecular, so we can work out payment addresses and escrow stuff

Go ahead, we've got no reputation at this point of time, and we need to build the confidence. Alternatively we can use offer to use paypal or something. Whichever the buyer prefers.

I would really prefer bitcoin, it's just way cooler. Unfortunately I cannot find an escrow service. Clearcoin is closed. Maybe we could find a trusted community member? Paypal is an option however, if we cannot find someone.

Also: I would prefer some form of secured, tracked shipping, would hate to see this precious thing lost in shipping. I'm in germany. Can you give me a quote on shipping?

EDIT: it seems all xilinx platform cables I can find on ebay.de (german site) are shipped from hongkong... maybe you could even get one for me and include it to save a little on shipping? I'd have to shell out $57 incl. shipping to get one myself.

I have to admit I'm kind of jealous of you guys.  :-[

But isn't the Board overengineered? Those voltage regulators look like they could drive at least a dozen spartan-6s. Out of curiosity how much LUTs do the 32bit rotations consume?

Overengineered- absolutely! Like I said, our priority was making sure it worked. I've had problems before with noisy or inadequate PSU's. These voltage regulators were chosen because they have especially low radiated EMI. On the version that we sell, we will downgrade them to a cheaper version, but with the same package.

I'll mention, too, that we will test everything before we ship it.

That's a question for fpgaminer, who is at work right now...


Molecular- we talked about including a Platform cable with the FPGA board. I think we should do this for you. Let me see what Li works out as far as shipping and stuff is concerned.

Molecular: I can register the mail and then we'd be able to track it, I can't seem to find how much it'll cost to register everything, but I do know the base charges. Shouldn't cost more then USD$12 to get this packed and shipped to you registered.

ElectricMucus: wow what a nick, yeah it was over-engineered, I had no idea how much current the 1.2V rail will consume in the future, didn't want to re-roll or re-work everything if there is a limitation, going by the measurements I've carried out so far, that size (10A capable) is about right for VccINT, so I wouldn't say it'd be enough to power a dozen Spartan-6s, maybe another 1 at the current power consumption levels. The VccIO and VccAux rails require fair less current then I thought it would, a lower spec part is avail in the same footprint, I'll put it into the remaining production boards.

Here's where we stand... if you buy a Platform cable on eBay, it will take 2 weeks to ship to you. We could buy it for you, but then we just have to ship it twice and there's no cost/time savings.

We are working on a design now to eliminate the Platform cable. If that ever comes to fruition, it will lower the board cost quite a bit. Also, with future bulk orders, price will go down more.

So, we need to cover our costs now. We want to get some boards out in the community for evaluation. At the same time, we don't want to screw our first buyers by charging a ridiculous amount, then undercutting that price by $50 or so in a month's time.

Here's the best solution I can come up with-- $440+shipping for the board. Buy your own Platform cable on eBay (it is a knockoff, but we use them- they work). If you ever decide to buy from us again, you get 10% off. This offer stands for our first 4 buyers.

Essentially, you trade a high cost right now for a much better bargain in the future. What do you think?

Here's a link to the absolute cheapest Platform cable- http://cgi.ebay.com/Xilinx-FPGA-CPLD-USB-download-Cable-JTAG-programmer-new-/250831239457?pt=LH_DefaultDomain_0&hash=item3a66b4f921#ht_2447wt_993 (http://cgi.ebay.com/Xilinx-FPGA-CPLD-USB-download-Cable-JTAG-programmer-new-/250831239457?pt=LH_DefaultDomain_0&hash=item3a66b4f921#ht_2447wt_993)

OMG.....

When I followed your link to eBay, then I do a search on it using key words:"Xilinx FPGA USB Cable".
I found there more than half of them are designed  and produced by me, at least they use my product pictures for advertisement... I sell more then 2000 pcs of this staff in China, on TAOBAO.com, and some for middle man. In the past year....

Well can you hook us up with an insider price?    ;D   Maybe you could help us design an on-board alternative?

Man I love the global connections of bitcoin

Remember, GPU mining may or may not be profitable 2 months from now. An FPGA miner should be profitable for years to come

The on board jtag cable's SCH  is public. Xilinx open source of it in SP605 Development Board

follow this link:

http://www.xilinx.com/support/documentation/boards_and_kits/xtp067_sp605_schematics.pdf

on page 32.

But what I want to say is: would you like to implement a more universe interface for the FPGA? For example, an UART port. This protocol is very easy to achieve in FPGA logic, and easy to use. It's only needs a MAX232 chip for voltage level translator.

Have A Nice Day

ngzhang-- thank you for the advice! I have a feeling that you might be instrumental as we develop the next round of these boards. We always have a lower $/Mhash target in mind. My brain is fried for now, but we will certainly be talking to you later

It does make things easier with a UART port, maybe with a USB to TTL converter instead of a MAX232, however we'd now need to add an SPI flash to load the bitstream initially. I was thinking more of a custom Jtag to USB solution using FTDI chips for eg.

Add one SPI flash may become a good alter on your next board, I think. An WINBOND W25Q64CV chip is only about 1$. So the mining board will not need to be re-config every time on the power up.

Best wishes

EDIT:

Why not start another thread in " Development & Technical Discussion".
I apologized about disturbing your selling thread.

im really quite interested in buying one of these, sent you a pm.

Hey Li, can you please answer arij's questions? I'm still up against my message limit.

Congratulations on getting this up and running very quickly! It's good to see an FPGA mining out in the wild.

On the other hand, I'm surprised that the two of you branched off of the open source modular miner project (https://bitcointalk.org/index.php?topic=22426.0) to make this board on your own. Why not just work together to get the best possible design and make it open source? You're already discussing adding more features and it that we've already worked hard to design for.  Clearly, you're going to continue moving in the same direction (i.e., back towards our design), so why split up our efforts like this?

Obviously, the design is basically a pared down version of the open source one. Is your design open source? I'm not an expert in open source licensing but aren't you required to release your design under the GPL because it is based off of other work covered by the GPL? It seems like you're getting into a sticky area there.

It's not based off yours in the least. We developed it completely independently. I guess the voltage regulators are the same, but... you would expect that since Li did the voltage regs for both projects. And he also did most of the layout for your design, up to a month ago, didn't he?

Yes, I would be very happy to work with you and O_shovah. I already tried once before! But, don't get me anywhere near an MCU   :)   I like to keep things simple

Ah, of course. Forgive me, for some reason I thought you were based in Asia and could get to the cable quicker. I'll look around if I can get it faster. How fast will shipping of the board be?


This is a fair offer. I've actually been thinking, maybe I should wait for better versions, but that is now out of the question.

Let me sleep over my decision, though, I've made the experience that it's better to sleep a night before making a sizeable purchase and let's figure out the payment and shipping details in pm tomorrow, ok?

$12 to ship a small package from USA to germany. That sounds too good to be true. Which service is that?

I'm in Singapore and will be assembling the 1st few boards. Singpost (our national postal service) will do it, the entire board itself weighs about 200 grammes inclusive of packaging (bare board is about 60g). If you want to cut down on shipping even further, I can ship without the heatsink+fan combination and you can add them later (assuming you can find something suitable locally).

They do offer the option of registering the item, so it can be tracked, I've shipped with them a fair amount of times, with some fairly high value items, and it hasn't been a problem thus far.

How about 1 Ghash/s and 68W of power?
and <$500

Sub $500 will just not be possible with current FPGA prices/performance for 1Ghash/s

Jordoss: It runs on a Spartan-6 FPGA, basically most of the calculation process is implemented in hardware, thus making it fast and power efficient.

Hello you two,

Well first of all i have to cangratulate you for your FPGA board finished and operational.
This event will certainly pave the way for the new FPGA generation.

Nontheless i have to admit the striking similarity to the modular fpga project gives me some stomache ake.
Altough i personally don't blame you as it is an open project, but i dont know how the other contributers think about that.

I also have to admit that the modular fpga project has slowed down seriously with some of the core personell missing.
I may not complete it alone as it takes me a lot of time to get trained for the MSP 430 coding.
I will keep up the development but currently at a slow speed.

Therefore i personally would appreciate any future coporation with you, in case i din't miss my chance the one time i turned you request down.    

 

Your design is way different. The only similarities are: the same voltage regulators, and the same FPGA's. As far as your project being open source- it really isn't. I have no access to your project, so I'm not even sure about the details on it.

How about this- I'll extend an olive branch. Drop the MSP 430. It's silly and it's overkill. Help us put a simple USB interface on our board, and help us get it manufactured en masse. I have no problem letting you and fizzisist in on this project.

As i said ,i don't acuse you of having stolen our ideas or something like that. I just picked up some pm's i got.


My final target is to provide the community with a working fpga solution no matter who it originates from.
So as i said, i will gladly contribute to your board if it is still appreciated     

Cool. I count you in, O_Shovah. If Li is OK with it, I'll start cc-ing you on emails and stuff like that. Let's make GPUs obsolete   :)

Hi,

If you use many board, can they be daisy chained or each board need a USB cable? If you're using JTAG, Xilinx devices can easily be daisy chained (1.TDO-->2.TDI-->2.TDO-->3.TDI-->etc.-->1.TDI)

Currently each board needs its own Platform programming cable. We thought about using jumpers or something to daisy chain them, but figured there would be noise issues. We need to find a better solution. The next batch of boards won't use Platform programmers, I'm almost sure.

I'll try daisy chaining them on the next board I produce. I'm sure within the Xilinx ISE it'll work fine, but outside of that (the mining script etc.) I don't think it's set up just yet to use 2 FPGAs, pretty sure some code changes in the mining script at least should take care of that.

This is really really cool and I'm glad to see a working board for sale so quickly! Not needing PCIe is like a dream and mining via usb on my old AMD duron machine would rock! Best of luck!

I have little experience with FPGAs, but I know that you can substitute the programmer for a microcontroller in order to load code on them, I have a ProxMark3 (RFID hacking tool, and only device I own where I can access a FPGA) that I've been dabbling with FPGA code on, it uses its ARM processor to load the FPGA. Might I suggest adding a basic one, like a PIC18 in order to avoid having to buy the expensive cable and make it easier to chain them? The PIC18F2550 has USB and can be chained easily enough using I2C. The 67J60 has a built in ethernet controller and that combined with a FPGA miner could be completely standalone.


Also, is there any chance that the FPGA miner code will run on a Spartan XC2S30? If so, I can probably create a version that will run on a ProxMark (~$300)

I'm sure people will want to buy more than one board...so some daisy-chaining mecanism would be good for your business :)

wait what else can fpga board do??
i dont have much expirence with them.

diid u get your chips from http://avnetexpress.avnet.com?

Consider me in. I think the perfect FPGA miner is something in between this design and the sophisticated one we've been developing in the "modular" thread. Let's work together and get things moving, at a reasonable cost and in reasonable time.

Errrm... none, because the rotations are just rearranging bits in a fixed way. They do use up routing resources though. (I'm not involved with the design and sale of this board but I did contribute a few performance-related improvements to the FPGA code.)

Are you certain? Aren't the LUTs involved in routing too? I always thought that they were either configured as logic or routing node.

I have to admit my understandings of the workings of fpgas isn't quite as good.

I would so buy one or two but $440 is a bit too much to invest in 100 Mhash. It would take a seriously long time to earn back that original investment. I just picked up a 5870 for $180 and that nets me 393 Mhash/s , so 3x the amount this board does at a more reasonable cost.

According to my calculations it would take  23.66 months to pay for this board - assuming btc prices stay at ~10.98.

That's OK. People like artForz get it. What happens when only half as many bitcoins are created with every block? Or farther down the road, when miners are supported mostly by transaction fees?

What I'm getting at is-- the farther along we go, the more important computational efficiency becomes. At some point GPU's aren't gonna cut it.

*Sorry to be curt in my original reply

I've read the irc log and just discovered that there is a wiki entry. But are those claims reasonable?
Besides the rumors, I don't think that a asic would be that desirable even by someone with alot of budget. The feature size is much too large for these low production units and really cornering the market wouldn't be possible considering that the latest fpgas, even with all its disadvantages have the smallest feature size of any product now.
If this trend continues it might even surpass the efficiency of a low production asic or any asic for that matter if they become the dominant technology.

But the strongest reason imo is that you can't update asics, so I think he can suck it  :P

LUTs are sometimes involved in certain specialised kinds of routing, but most of the routing is done through a dedicated routing fabric in which the LUTs are embedded. As I understand it this is one of the things that distinguishes FPGAs from other forms of programmable logic like CPLDs.

Yeah, FPGAs very power-efficient but the upfront cost is nasty for every solution I've seen so far. On the plus side, as newMeat1 says you can keep going well after GPU miners have to give up...

Edit:
That reminds me, an interesting factoid: unless he's been deliberately misleading people, at the time he had the ASICs made he didn't actually know about the optimisation of computing H+K+W[0] in the previous round, which improves performance for free by reducing the critical path (at least on FPGAs).

On optimization of mining algo (be it on GPUs, CPUs or FPGAs) is there a post that describes all available optimizations?  I'd like to learn about it.  Can someone point me to it please?  or is the computing H+K+W[0] in the previous round pretty much the only optimization so far?

I'm mostly just the firmware designer, but I will try to chime in on a few things here for everyone's benefit :)

First and foremost, I do not think it has been made clear that these boards are firmware upgradeable. As improvements are made to the Open Source FPGA Bitcoin Mining project, new firmware will be generated and made available. Just like we see improvements to GPU software and you can drive more out of your dusty 5850s ;)

Personally, I am very excited about this little board as a development platform :) Beats the heck out of my $1000 dev kit  >:(


Connecting Multiple Boards
As newMeat1 has hinted at, this was being fleshed out and worked on even before these first boards were sent into production. I'll be playing around with FTDI chips and/or micros, and see if we can get a nice, simple solution that will be backwards compatible with existing boards, streamline future designs, and scale well for multiple boards.

I have an ~8 year old laptop that could probably drive these things :P All current GPU mining rigs can also drive it, in addition to driving their GPUs.

The Spartan-2 series is likely too old to push many MH, if any. That particular chip only has ~1K CLBs, for example.

Think of an FPGA as a massive breadboard, with LUTs glued onto it in columns and rows. The breadboards are your "routing fabric" and allow you to choose how you connect the LUTs, and of course you can load the LUTs with whatever configuration you want. That's your most basic FPGA architecture. Just note that there is a "routing cost" proportional to the distance you try to route. For example, connecting one LUT to another half way across your massive breadboard would lead to a massive signal delay.

This thread (https://bitcointalk.org/index.php?topic=33817.0) gives a good run down of all the optimizations being implemented on GPUs.

This thread (https://bitcointalk.org/index.php?topic=33817.0) gives a good run down of all the optimizations being implemented on GPUs.
[/quote]

thanks! i did see this before and had lost track of it. appreciate the link.  so basically similar algo optimizations were implemented on your fpga code?

Not really. Those are GPU optimizations, and are focused on reducing the total number of calculations needed to compute the final hash. That benefits the GPU's performance, but none of them would really have a useful impact on an FPGA's MH/s performance. So I haven't implemented most of them, because they aren't too helpful, would clutter the code, and my time is better spent elsewhere for now.

I was just wondering ... why does a 6.8W FPGA need a fan?

A cheap 40mm fan is like 0.07Amps so about 0.8Watts - an increase is power of about 12% - 12% is a lot.

My 6950's I set to not run the fan when they are under 60degress.
At 350Mh/s they stay over 60degress, but a drop to something like 250-300Mh/s and they will stay under the 60degrees and not even need the fans on.
... and we are talking over 100Watts (350Mh/s is about 180Watts)

Just wondering about the decision to bother to put a fan on it all ...

I presume you have done the obvious optimisations that clearly mean less calculations per double hash:
pre-calculating the first 3 rounds and not bothering to calculate the last 3.5 rounds since you don't need them at all ...

6.8W isn't a lot compared to a GPU but it is a TON compared to these FPGAs that normally run at 3 watts or below. FPGAs are really being pushed in these applications.

That's a smart question...

You would be surprised how much heat 6W can generate. I talked about this a long time ago with fpgaminer. He said something like, "It's not so surprising that you need to put a fan on an FPGA. The surprising thing is that a 6990 doesn't burn your house down."    :P   Maybe the difference is, a 6990 already gets some airflow, being inside the computer case.

But yeah, we might be able to get away with no fan. I think the chip doesn't start incinerating 'til over 100C. fpgaminer was working on an error-detecting code. If the chip gets too hot and starts making mistakes, it will detect the mistakes and stop.

You bitcoin crowd sure are a tough sell. I think I will try to sell these on eBay

P.S. The board has been running non-stop for about 18 hours now    8)

It is because the FG484 chip is about 28mmx28mm, plastic packaging. Only a 28X28mm heatsink can add to the chip. So a small fan is needed.

Here's something else I just realized. Li wrote this in another thread earlier today:


So if the chip gets hot, it starts using more power. Looks like the ~0.5W of the fan is justified...

It's just helps to justify it more, there's no easy way to cool these things well with such a small footprint, I haven't seen a heatsink that's alot higher then it's length/width (at least not where I live), I can't cool these passively, so adding a small fan ontop was the easiest way without having to scrounge for specialty heatsinks (the current heatsinks I use are about USD$1 each, affordable, the fan adds a few more bucks.)

Without a fan you'd be running into finger scorching temps. I have no idea what the actual on die temps are like (the chip is encased in a plastic epoxy package), but I'd definitely not be running 24/7 at those temps.

I know at this point the board may seem to be pricey, but the costs for the board and components for low volume orders are high, let me put it into context. The spartan-6 based dev board that FPGAminer uses during dev costs USD$1000, this is less then 1/2 that cost, yes it doesn't have all the other features, but that's what we want for a mining board.

We'd like to at this point sell off a couple of the boards I have on hand and recoup some of the original investment to continue spinning the next batch of boards.

Support for the original board (this batch) will continue for as long as feasible, upgrades may come in little modules that you can plug into the Jtag port (for chaining etc.), we aren't planning to screw the early adopters over.

Very cool. Looks promising! Unfortunately the investment price is too high to make much sense.

nice stuff !! i support but price put me down!  ;D

Well, look up how much an LX150 dev board costs! I'll tell you- it's gonna be at least $600. And a board like that probably won't stand up to mining for very long.

Now maybe does the price look a little better?

That price is definitely good for an FPGA miner, I think though the most savings will come from mass production due to the power, heat and space saving costs. If you can manage to get 5 on a board and sell it for ~1k and make it stackable it would not be a bad investment. I'm sure people would put down some serious $$ then.

5 FPGAs alone (without board any parts yet), will suck up the 1k budget already. We aren't trying to sell the boards with a sky high margin, the costs of the parts alone are just quite high at the moment.

Well the wiki says "-3N 484-pin chip is ~$150" which is cheap for a spartan 6 but 5 of them would only be 750$ the rest of the board can be 250$ and you could charge 200$ per board to make it worth your while and It would still be a good investment.

That's where the wiki is wrong and needs to be edited. I couldn't find -3N parts in the 484 package in stock with any of the distributors (Digikey and Avnet), else I would've gone with them. Even if you go with a super cheap power supply, at small quantities, it's quite impossible to make a good PCB and keep component costs below $200 (a cheap PSU maybe $30 per FPGA? x5 that's $150 right there and then you'd need to include the PCB). Not to factor in the cost of soldering the BGAs. Maybe at 1.3k it might be doable.

How about world peace?

It would probably take even longer if prices rise again, because that would attract a lot of gpu and increase difficulty. It might be faster if price falls to $1, because that would drive out power-inefficient mining and puts loads of pressure on difficulty. So this is actually insurance against price drop ;)

I'm happy to report to the thread that I just ordered my FPGA miner.

Both guys (Li and newMeat1) have been very helpful so far in providing me with documents I asked for (an invoice).

I'm now waiting for them to send me an order confirmation with payment instructions. I will pay using paypal, unfortunately, because that's safer for me.

So I'd say: probably ONLY 3 LEFT!

I'll keep you guys updated on how it goes and I hope Li and newMeat1 can build some confidence this way.

UPDATE0 Aug 19th 10:00 - I ordered Xilinx USB platform cable through ebay.de for €41.50 = $59 incl. shipping from HongKong.

UPDATE1 Aug 19th 10:10 - I received an invoice from Li via paypal and payed $440 (+$12 for shipping and handling) = $452. I noticed my shipping address in paypal was a very old address of mine and PMed Li with the correct address. I hope it'll work out.

UPDATE2 Aug 19th 10:20 - Li confirmed receiving the payment and also confirmed he would send to the correct shipping address.

It should have the second of those two optimisations (not calculating the last 3.5 rounds) but I don't think it has the first one, mostly because it's harder to do and requires changes to the supporting software.

That's surprising actually. PC motherboards and graphics cards have similar power supply requirements, and they're definitely available for below $200...

The effect volume has on component price blows my mind too. It's really quite unfair, but that's supply and demand for ya :/

Congrats! Who wants to chisel that into the Bitcoin history book?  :D

I'm sorry but they manufacture motherboards and other stuffs in millions quantity. If we can build our mining boards in this quantity, we even can build them with ASIC, which 10 times cheaper than fpga, and 10 times faster than fpga, and the power consuming is 10 times lower than fpga, all these features are provided on a single board.

molecular, thank you very much! You won't be disappointed. Also, as one of our first customers, you locked in a 10% discount for life

Edit: We are down to 2 boards left now. And we are trying to work out a selling deal with Cablesaurus!!!

Happy to report as well we'll be working together to maintain a seamless vein of supply to the community as this begins to find utilization.

Another important question: what is the expected lifespan of such an FPGA board running at 100% 24 hours a day 7 days a week? 6.8 watts sounds awesome but without knowing an average lifespan to amortize the hefty initial cost over I can't say with any certainty whether this is better than GPU mining, and if so by how large a factor.

Have you looked at the microsemi smartfusion devices? They offer a huge amount of gates at a much lower price & you get a extra arm processor.

am I missing something  ???

We aren't using the LX45, that's a puny part with 43+k logic cells. We've on the board a LX150 part that has 147+k logic cells, also the -2 plays a part in how fast the chip can run. -3 is what we have.

http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=122-1725-ND

I have no reason to believe it won't last as long as any Graphics card, if not better. The current cooling solution keeps the chip under 50deg, GPUs constantly run higher then that. The board itself is populated with quality components, most capacitors are ceramic and the remaining are polymer capacitors (that don't leak electrolyte).

The question remains of course, what if the fan fails? I'm sourcing for good ball bearing reliable fans. That probably is the thing that'd kill the FPGA, the rest of the board should still work though, those LMZ modules are very hardy and can take a direct short to the output. If you send the board back I can fix it at the cost of a new FPGA.

I realized that and edited the post, sry ;)

But what about those?
http://www2.mouser.com/ProductDetail/Microsemi/A2F500M3G-FGG256/?qs=9C5AG8cDq%252bW6haUt%252bCawKA%3d%3d

There's no real way to compare logic elements vs system gates, the only way would be to port the code over and see it'll fit.
From the rough looks of it, not gonna happen, there's the extra microcontroller and extra features sucking up silicon real estate.

are you asking just to ask or have you compared the technical capabilities of the chips?

idk, I suggested the chip because I assumed gates are gates, no matter on which fpga they are. So I did some comparison.
This wasn't meant as a critique or that you choose the wrong chip, just me trying to help  :-[

I've heard that the markups of fpgas are quite high, so I assumed it is possible that a cheaper chip is available which could be used with similar or superior capability. Xilinx is the market leader so I think it's save to assume that competitors try to offer cheaper products to get a higher market share.     

Yeah, I know, thanks a lot for the discount. That really made the decision a lot easier.

Your help is appreciated, and it's always good to learn about new options.

A lot of FPGA products will list their gate count in promotional material, but gates are completely meaningless on an FPGA so it is just a marketing "scam."  :-\ Took Xilinx a few generations to stop that nonsense and start listing their products in terms of 4-LUTs.

I didn't read it as critique, I was wondering if you had found something...  i don't know much about fpga's so that's why I ask.  I wouldn't know how to compare them... at least not yet. :)

Still have the 2?

I'll buy both if you can accept PayPal as I don't have enough BTC.  Please pm me details.

3 of them have sold, for sure. There's one that payment hasn't gone through on yet, but I consider it spoken for. Pipesnake- if the buyer doesn't come through on this last one, I'll contact you right away.

About 4 of us got together and had a really nerdy meeting this afternoon, planning the next version of the board. We think we have a solution to the USB interface problem. It's scalable and it costs less than $10!

fpgaminer is going to test whether our idea works in the coming week. Then we will have to order more boards, which takes 2 weeks. So, in 3-3.5 weeks, hopefully we will be announcing a newer board.

Ok I'm in for sure on the last board or the next batch.  Awesome work!!

https://bitcointalk.org/Themes/custom1/images/post/thumbup.gif on gettin' em sold! I would have liked to pick one up but the price was a bit high (understandable though) and I'd really like to make my own.

Nice to see some progress on this, props for being first to market.

1337 views  ;D

Are you guys using the DSP48A1s (@100MH/s?) They will not necessarily
increase Fmax, but, they should reduce LUT count and perhaps let you
fit more stages/cores into the part.

-rph

Hi,

What would be the minimum spec for the computer controlling this?

Would this http://mini-itx.com/store/?c=34#jnf94-270 (http://mini-itx.com/store/?c=34#jnf94-270) do?

Thanks!

Well, you will need an internet connection, for sure, so that you can communicate with the mining pool. It looks like this board has an ethernet connector and USB ports, so maybe it would work. If you could load Red Hat Enterprise Linux WS or SUSE Linux Enterprise onto this little Atom board from a USB drive, then it wouldn't be too hard. Our mining scripts and the Platform cable would work.

I bet fpgaminer can give you a lot better answer when he's online. He's our main programmer.

A more cost-effective option, and easier, would be to buy an old laptop off eBay or a garage sale or something.

Anyways, let us know if you decide to use the Atom board. I want to see how it turns out, and I like the idea of a highly efficient computer.

EDIT: I'm forgetting about the Molex power supply. Somehow you need to give our FPGA board 12V.

oooooooooh look at this project http://www.kickstarter.com/projects/794668827/aliencortex-av

Sounds good doesn't it? But the FPGA on it is pretty weak. It probably has only about 20% of the logic elements of our board, so you would be looking at maybe 10 Mhash/s from that thing. (The relationship between # of LE's and hash rate isn't linear... the more LE's you have, the more efficiently you can mine. Up to a point)

Also just noticed it only runs at 24 MHz... so really it would only be good for 2-3 MHash/s

Congrats, looks like great work getting the first FPGA miner up and running!  Better keep moving, there are likely many others right on your heels. :)

My hours are booked solid until November, with no time to fiddle with a new board as much as I'd like to, but I'll likely jump in on the next round of boards.

I hope my suggestion wont be taken as trolling...

how about setting up a company to garner resource for research and selling the end product to the btc community like what you guys are doing here..

except that those that actually buy into the company will either get profit/discount on the product...?

IE...

total dev cost is $5000
total manufacturing cost is $5000

(these are just assumptions.. i'm sure someone can make a better accurate depiction of how to do this...)

so you would sell 10k worth of shares for the first round to buffer for this...
the dev who put in time and money would be covered by the shares representing their hardware cost and dev time...  may be 1/3 of the total shares?
the rest is open for public buy-in...

when boards are sold... the profit is divided upon the shares hold... and distributed among share holders..? if shareholders wants to buy at discount they get a % related to the % of shares they hold...?

just a suggestion... nothing more ...

If shares for said theroetical FPGA comany could be purchased on GLBSE i would certainly buy some myself.

+1

+1 as well... just trying to get these R&D and mining hotter with less heat... (pun intended)

dovewing, Pipesnake, and deslok... I appreciate the votes of confidence. We definitely thought about going with GLBSE about a month ago. In fact, we even had some meetings with IBB about investing in the project.

Trouble is, we already have 5 talented people working on these boards, so we split our meager profits 5 ways. With 5 different people, managing the project is already tough. I think dealing with investors would make it even worse.

We're already pushing to the next level as rapidly as we can, being held back only by shipping delays and board manufacture delays.

understood - from investor's point of view.. i think it's more to do with profit (ie i want the money) and less to do with how u guys design it... the product/how it's developed/researched is not really something investors would think too much about, unless they want to be doing R&D...
the reason i bring up GLBSE is because that would immediately give some buffer for R&D/Manufacturing and give you guys some money back from investing in dev kits or material used, etcetc...

but either way I do hope you guys can make headways as you can tell a lot of interest is garnered from this thread if not due to the "price of entry" for the current incarnation of the product...

on a side note though... if the product can be daisy chained with usb cable.. you can probably use regular heat sink to cool it due to the larger surface area..

               usb cable
[ board 1 ]========[board 2]
     ||                         ||
[board 3 ]========[board 4]

just suggestions/ideas...

My personal favorite solution for passive cooling is mineral oil. Here's a link (http://www.pugetsystems.com/submerged.php) to a fish tank-mineral oil PC. If I had one of these boards a.t.m., it's what I would do. But, mineral oil is expensive and some people probably don't want oily, drippy boards.

I guess I can say this- if you want a custom cooling solution for your board, and it's easy to accomodate, just tell us! Like, maybe you don't want a fan on yours because you will be buying a different heatsink. We can accomodate that

I've been thinking more about your suggestion. I think we could make it work, but not exactly as you imagined.

A big problem for us:
On one hand, we don't know how many boards we can sell. On the other hand, bulk discounts could drastically reduce the price of these boards. To give you an idea of how much... if we only make 5, each PCB costs about $60. If we make 70, each PCB costs less than $10 (I think $5 is the exact #, if i remember right).

So maybe we could take pre-orders. It would be to the benefit of everyone. We know exactly how many to make, you get a bulk discount. The only problem I see is, we don't know exactly how much the boards will cost yet. Any suggestions? I'm thinking, right now, that perhaps we could require a $50 payment to secure your order slot. The remainder, whatever it turns out to be, would be paid prior to board manufacture.

I will have to talk this over with the rest of the dev team. By the way, fizzisist has started helping us out. He has a lot of knowledge relating to communication interfaces, circuit design, and probably more skills that I don't know about yet. So, if you see him on this thread, know that he's a core member of the  team.

Cool - i think the pre-order deposit idea would work (ie 10% of the price for more than 5 units and a set amount for small quantity)

I assume pcb is the entire board? (can we see a few pics of the board?)

The PCB is just the green silicon thing, like in this picture:

http://lh4.ggpht.com/-DseG3N374yk/TksrJ0wFaFI/AAAAAAAAAWY/5CvlvnxWFco/IMG_1063_thumb.jpg?imgmax=800

The newest batch of boards will likely look very similar to our old ones-- see the first post in this thread for pics.

let's say next batch... if people are willing to order in bulk... what will be the pricing be like? (if i get way ahead of the schedule let me know... just asking questions as they pop up in my head)

How big is "bulk"? 10? 100? It makes a huge difference.

I'll just give you a rough idea. We'll flesh this out better later on.

Total community order size    -      Cost per board
5 boards - $400
10 boards - $395
25 - $390
.
.
.
70 boards - $335

Li, please correct me if I'm way off

I wrote down some very rough profitability calculations for this product:

https://bitcoin.org.uk/forums/topic/88-mining-for-profit-or-for-fun/page__view__findpost__p__418

It should be clear to anyone that at this point in time, such an investment can only be considered to be "for fun" or "a step in the door". Not trying to drive people away, though, this is the future of mining (alongside ASIC mining), just not the immediate future.

that's why i suggested the company idea... to create a R&D pool and to aim for a good Mh/s per $ ratio... instead of pushing out 100Mh/s boards...
may be aim for improving to 300-400 (without optimization) per board with this current price? let me know how feasible it is guys?

The problem is fpgas do not scale as well if you use multiple chips (depends actually, could be linear) there are real monsters though, virtex-7 and so on which could deliver that performance but cost about this price just for a chip (or more)

The next thing is the FPGAs wouldn't get any cheaper (that means a fraction of the price) unless in _huge_ numbers. And even then the costs wouldn't go down as drastically since afik if you buy a whole lot you get them untested...

It does scale linearly with each FPGA addition, until of course you run out of network bandwidth or CPU power to feed the FPGAs quick enough.

To answer the previous question on whether the ITX motherboards are fast/good enough. On the current mining script (which isn't very well optimsed) my Q6600 quad core processor reports a utilisation of 3-4% max in windows when it's running. No idea how much that scales to when it's running on an atom, but it should be more then good enough. I need to test this out though, do have an atom board lying around.

Why couldn't we put 2 or more FPGA chips on one PCB tho, ala (and forgive the reference) Voodoo 5 5500 or 6000?

So ordering 1 FPGA board is $440, and ordering 4 would be $1760

but if we can get 4 on 1 PCB, and 20 boards together, the each chip is $360, and a board sells for $1440.

(Which is using the cost of 1 FPGA miner sold in the thread, and newMeat1's prices for bulk orders above.)

You're reducing your ROI from 22.5 months to 19.2... still well after the block reward halves, but in the right direction.

Now, I don't know how hard this would be to program for... or even if it's possible... but it seems that getting more chips on the PCB would be beneficial to all.

It doesn't quite scale like that, since you're now using a larger PCB (slightly increased cost per board), but yes we are looking into that to make sre that's feasible, the original plan was to keep everything simple and easy to debug, thus only 1 FPGA per board, once you start going upwards there's gonna be alot more involved like software/communication issues that need to be worked out, no one I know has a dual FPGA board to test this out. Not to worry though, development is still on-going in the background and we'll bring them out when they're ready.

300-400 Mhash/s for $440 is out of the question, I think. Each FPGA alone costs $200, and we push them as hard as they will go. You might see 150 Mhash/s out of each FPGA chip, dependent on programming improvements. If we put 2 on a board (which fizzisist is working on right now), maybe you would see 300 MHash/s for $500. I think that's the absolute best we can expect from a Spartan 6.

Somebody mentioned the Virtex-7 series of chips. Those are the replacement for the Spartan 6's. When those become available... maybe your figures would be possible. It's far too early to say.

I just wanted to add a note of support for this project, I am looking forward to your next implementation: I considered the previous model already, but I don't quite have the capitol for a $500 investment with a two year payoff at my power rates... although, it is tempting, and the next incarnation could break my resolve :p

The architecture of the next batch is settled. We are still working out some details, but we should have a working board in 3 weeks.

There will be space for 2 FPGA's on the next board. If you only want 1, then we will leave one spot on the board blank.

We will be doing pre-ordering through Cablesaurus (our exclusive distributor in the future). It's $50 to preorder your board (same cost whether you want one or 2 FPGAs). Just before manufacturing the boards, we will ask you to pay the remaining balance. With this preorder system, we will make the right # of boards as well as getting a volume discount. Somebody will post here again when the preorder system is up and running on cablesaurus.com (http://cablesaurus.com) (it's not up quite yet). Cablesaurus takes Paypal and bitcoins.

The more preorders we get, the cheaper the boards will be for everybody!

Here are our basic guarantees about the boards:
1. Delivery by Nov. 1st 2011 (hopefully sooner, I left some extra time in case we screw up)
2. Only a USB cord and 4-pin Molex will be required to run it (no Platform cable)

Specific to 1-FPGA boards:
- at least 90 Mhash/s avg hash rate
- less than 7.5W power draw
- costs less than $420 (likely quite a bit less, especially if we get decent volume discounts)

Specific to 2-FPGA boards:
-at least 180 Mhash/s avg hash rate
-less than 15W power draw
-costs less than $620

For the record- these are our first 4 customers who get 10% discounts on future purchases:
molecular, fpgaminer, thirdlight, and Pipesnake.

The obvious question: Will there also be a discount for the 3rd+ generation for generation 2 buyers?
Will the 1 piece board be upgradeable with a hotplate & some luck?
Will the preorder deposit be refundable if you don't have the money (I know that's cheap lol)

Good day. I am from Russia. English is not their native language. Used an interpreter Google.
Wishes to the new version.
Interested in this card, but at a lower price. And just the ability to work without the programmer, such as usb. Also included is not needed radiator and fan. This fee will reduce the price and ease delivery. You just need to provide for the installation of a standard computer with a radiator fan, for example by video or from a Pentium 3. As an additional power connector to provide instead that a Molex type PHU. To facilitate the assembly of clusters requires the mounting holes on the board. Also included should be idle intuitive software, something like a modernized GUIMiner.

ElectricMucus- those are hard questions. I'm gonna have to consult with the team before i answer them.

SolomonVR- the next batch (coming out in 3 weeks) will be much cheaper. See 2 posts above yours. Yes, there will be mounting holes on the next batch (it was planned for the first batch but things got confused...). Yes, we eliminated the Programmer for the next batch. About the rest of your comment, I'm not sure what to say. Maybe a member of my FPGA team can help me out.

SolomonVR, leaving the heat sink off the board is probably a good idea, especially for shipping. Maybe Cablesaurus can offer some alongside the boards as an "accessory," although some kind of heat sink is definitely required. When you say additional power connector, what type did you have in mind? The Molex type is ideal, because we can be pretty sure that someone won't connect a different voltage than required, or reverse the polarity. We can consider adding a second connector (or at least the footprint for you to solder it on yourself) if you have a good option.

As for the software, fpgaminer is in charge of the development. I'm sure he'd appreciate help if you're offering!

Nice looks like your down to 3$/mh I'm starting to get temped  ;D I think I'll probably wait another gen.

Well just because we expect a nice price drop from gen 1 to gen 2, don't expect any miracles after that! Gen 1 had some obvious problems, but Gen 2 should take the Spartan 6 just about as far as it will go.

I bet we probably won't do a Gen 3 until the Virtex-7 comes out, but who knows when that will be?

I wonder what the best way to mount a bunch of these would be. Surely just laying them somewhere cannot be most efficient.

What are the physical dimensions with the heat sink on?

Have you tried Fritzing Fab (http://fab.fritzing.org/fritzing-fab)? Just going from their examples it seems like they'd be cheaper. Not sure if you can have them leave the white paper off since you're working with SMD stuff (or if they leave a hole in it for the component?) but I thought I'd put it out there.

Fritzing Fab- that's a nice service, but it looks like they only make 2 layer boards   :/

Ah, didn't notice that. Also I honestly forgot you'd mentioned you're working with 4-layer boards, guess that would drive the cost of fabrication up a bit :P

ElectricMucus, I finally got some answers for you...

1. I'm sorry to say it, but no. The only reason we gave a discount to the first 4 buyers is because they knew that a cheaper version of the board would be coming out soon.

The Gen 2 boards will be pretty well optimized (in terms of $/Mhash) so we shouldn't need to give any extra incentives to buy them. Also, if we promised future discounts, it would be a nightmare for Cablesaurus to keep track of.

2. The FPGA chips are 484-pin Fine Ball Grid Array-type (FBGA). What that means is there are 484 little balls of solder underneath it. Soldering these things is difficult because all 484 balls of solder need to melt at once. Otherwise the chip will come down crooked or you will get a bad solder joint.

I've tried soldering 484-FBGA before myself with a toaster oven. I followed this guide:
http://www.instructables.com/id/Toaster-Oven-Reflow-Soldering-BGA/ (http://www.instructables.com/id/Toaster-Oven-Reflow-Soldering-BGA/)
It's tricky because you have to put the right amount of flux down, plus get an even temperature everywhere. For me, it looked good but I had some bad solder joints. So although it will be possible to upgrade from a 1-chip board to a 2-chip board, I sure wouldn't try it unless you have lots of experience.

3. I'm gonna have to say no again. Sorry!

Sirky-

For the next batch, we are looking at a 3.4in x 4.1in rectangle. I don't know exactly what the height will be~ probably about 1.5in with the heatsink. We're definitely gonna put some mounting holes and standoffs in, so you can stack them vertically.

Here's the kind of standoff I'm talking about:
http://media.digikey.com/pdf/Data%20Sheets/Daburn%20PDFs/10-40_1.pdf (http://media.digikey.com/pdf/Data%20Sheets/Daburn%20PDFs/10-40_1.pdf)

Actually, something like this would probably work best for assembling a stack of boards: http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=4808K-ND

Maybe Cablesaurus can sell some of these as accessories as well, so that folks can get everything they need in one order.

Cool project!


I don't know too much about the hardware side of this stuff, so here are a few newbie questions:

If you would control this from a laptop, where would you get the 4-pin Molex connector from? Do you need an additional AC power adapter in that situation?

Is there such a thing as a USB to 4-pin Molex adapter? Or does a USB port not provide enough power for this?

Is my understanding correct, that previously the platform cable was used to upload some type of firmware? Is this now done over USB? or what was the function of the platform cable and by what is it replaced now?

Thanks in advance. :-)

- I'll have PSU Dummy Plugs in stock which would allow you to power a PSU from basically anywhere without needing to plug it into a motherboard
- I'm not aware of a USB->Molex adapter but the PSU dummy plug would solve such an issue

I'll add some pictures and more details surrounding the plugs soon.

jav-

That's exactly right. The next batch will only need a USB cable. It plugs into the board and a chip handles the USB-FPGA interface. Exactly what type of chip handles the interface is a trade secret at this point  :)

I understand the preorder can be paid with bitcoins, right? Also, what would be the manufacturing date?



For me this is a deciding factor. Good job on this!



Does that include the $50 preorder? (ie is the total cost 420+50 or 370+50?) How much would that become if you get, say, 20 preorders? (I assume the same percent discounts apply to the 2-FPGA version and I'll calculate them myself).

Yes, Cablesaurus has always taken BTC. My best guess on manufacturing date (when the PCB order is submitted for manufacture), if things go smoothly, is Sep 15th.


Thanks. We realized pretty fast that it was crucial


370+50. 420 total would be the most that you would have to pay for a single FPGA board. Let us get back to you on the quantity-price breakdown. fizzisist and Li are working on some spreadsheets right now. My top-of-the-head guess is $385 for 20 orders.

EDIT: The preorder page on Cablesaurus still isn't up yet (it's just a test version right now). Expect a post from him soon

Being pushed ...

Does that affect the expected lifetime of these?
What's the expected lifetime at the normal 3W?
and the expected change to that at 6.8W?

Mining is 24/7 - so those lifetime expectations would be rather critical.

Just wondering.

If I want to give this a try, what's the most barebone computer I can use? Let's say I want to build a dedicated FPGA miner. I can get a cheap motherboard since I don't need PCIe slots. I can go with a small PSU and if I do LinuxCoin on a USB stick, I don't even need a hard drive. If I want to go all out, how many of these can I put on a single motherboard?

When will this be available for pre-order?

all you really need are usb ports so I imagine a lot.

Xilinx publishes some reports on reliability tests: http://www.xilinx.com/support/documentation/user_guides/ug116.pdf

I haven't read through that all yet, but the interesting test is probably the High Temperature Operating Life Test. It looks like they tested the Spartan 6 (XC6Sxxx) at Tj > 125F for 533,351 hours. They somehow scaled it to "equivalent hours at 125F" to 1,161,865 hours. That's 132 years. Not sure how to interpret the fail rate, but it's either 0 or 10%... Either way, I don't think failure is going to be a problem, but please use some type of heat sink. The better that you cool the FPGA, the less power it will consume, just like with GPUs or any semiconductor. We plan to ship the boards without heat sinks, so that you can put on whatever you like. We'll offer a suggested one, that should work well enough without a fan. Increasing airflow with a simple fan never hurts, though.


The USB bus allows for up to a maximum of 127 devices, so you should just need a bunch of hubs to get pretty much as many as you like. There might be some other issues with running multiple boards, though, because it's never been tested yet, but that should all be worked out in software. The smart thing to do would be to wait until someone confirms that they got multiple running before you place an order for 127 of them :)

kano- that was a tough question   :D   You made me do some serious googling. Here's a paper about how Altera tests their FPGAs. Altera is the other major FPGA maker besides Xilinx.
http://www.altera.com/literature/rr/rr.pdf?GSA_pos=1&WT.oss_r=1&WT.oss=Altera (http://www.altera.com/literature/rr/rr.pdf?GSA_pos=1&WT.oss_r=1&WT.oss=Altera)

They test by doing voltage cycles, temperature cycles, humidity cycles, configuration cycles... you get the idea. The good data starts on pg 12. It looks like all of their devices should last at least 10 million hours!

EDIT- fizzisist beat me to it by a couple minutes! Well, now you have 2 independent sources for your reading pleasure

Thanks - I guess I could have done it myself in hindsight, but I thought you guys would already know :)
Well certainly suggests they should last more than at least a few years even with adverse usage.

That's 1141 Years...

I doubt this figure is somewhere in the datasheet.

A common mtbf for ics is usally between 100000 to 500000 hours which is still enough.
mtbf means mean time between failures which is the time by which 50% of the devices have failed.

Although failure rate follows the bathtub curve which means some ics fail very early after which almost no failures occur over a long period of time after which it exponentially increases.  Modern ics and especially high revenue items like fpgas are extensively tested so that almost any initial failures are sorted out.

The important part though is that the mtbf is only viable for running the chip within specifications, if that is the case here only the developers can answer... hint, hint ;)

UPDATE3 Aug 23rd 06:42 - PM from Li, saying he shipped the board! Gave me tracking-# and ETA 6-8 days.

Another question: Could this be cooled by submerging it into a small aquarium filled with mineral oil? Would one still use a heatsink in such a situation or is that unnecessary, with a liquid cooling solution?

Are all pre-orders sold out?

Jav: It could be, but it's seriously overkill for this. I'm sure it'l be seriously messy though.

The initial 4 boards are sold out, I do have the 1st board that I assembled. I can sell that @ the same price if you're interested, let me know.

I'm interested in ordering one of the Gen2 boards. Is there any consideration for adding a mounting bracket to them so they can be mounted in a case like a PCI card (just not plugged into the motherboard)?

It is within specs. We definitely aren't overvoltaging or overclocking them

No- Cablesaurus hasn't started taking preorders yet. Look for a post on here when he starts taking them.

I like the mineral oil idea, too! I had a post about this earlier-- you wouldn't even need an aquarium, just a tin box or something that is about 5in x 4in x 2in. I'm sure you wouldn't need a heatsink, because there would be mineral oil both above and beneath the chip. Also because of the convection of the oil. From what I've heard, mineral oil conducts heat 10X better than air.

From a practical standpoint- if you put a big-ass heatsink on each chip, then passive air cooling will be just fine. We're still talking about it, but that's probably what we'll end up doing.

Our only mounting consideration is some holes for standoffs (aka spacers), which would allow you to stack boards vertically. My best idea, for putting one inside the case, would be to put some zip ties through these holes. I'm sure you could find some wires or something to zip tie it to.

Why would you want to do this?

I wouldn't want the small, whiny fan of this design anywhere near my apartment, but there are lots of better cooling solutions, and you can often go passive with a little thought. These typical designs where a fan is trying to push air through the chip, are clearly results of absolutely no thought to fluid dynamics. They are sometimes better off without the fan, especially if there is some airflow in the case already.

There are over 100-watt GPUs with fanless air cooling. I have one of these (a HD5770), also most of my CPUs have fanless heatsinks. My FPGA board is also doing fine (http://iki.fi/teknohog/hacks/fpga/#de2cooling) with a passive heatsink.

as support of the project I would like to pre-order one as soon you guys are ready

ps
I understand the cost it's very high but if you just considering how much $/Watt u saving in the long run just right here u making the payback especially if u get a quantity, where the compounding get quicker. I am considering only one right now just because I like to get familiar with the system where I understand the writing but I am not familiar with the practice.

I am trusting the project as well I am trusting the Bitcoin community where I have already spend a lot of my time and money not for greed but for a developing of a better system and community.

The next board is planned to have a passive heat sink only, and we will most likely sell some boards with no heat sink so that you can use one of your own choosing. I believe the previous heatsink/fan combo was used just because Li had those sitting around already, and it works.

I could better justify the price tag to myself, if I can build something interesting to put into my living room or somewhere. For that I want it to be silent and "hey, have a look at my money-making aquarium" seems like a nice conversation starter. :-) From reading the beginning of the thread, I got the impression that quite a bit of cooling would be necessary, so that's why I liked newMeat1's mineral oil idea.

But of course, if passive cooling is possible and safe, I'm all for it! So will this second generation of boards already be cooled passively? If that is the case I'm probably interested in a pre-order.

Hi folks,

I saw up top there was an issue with lack of escrow, but that it got sorted.

I'm willing to provide a manual escrow service, through the end of September, for this project's sales only. Zero fees.

Most barebone? SheevaPlug - $100 and done.

http://www.slashgear.com/marvell-sheevaplug-99-linux-pc-hidden-in-a-wall-wart-2435556/

Apparently there's a whole community regarding these things - complete with servers, etc.  http://www.plugcomputer.org/

I can imagine a power strip with these things hanging off of it with usb hubs hanging off of them with stacks of fpga cards.

Good luck running the requisite Xilinx software on ARM. Even if you can program the FPGA (using something open like UrJTAG), further communication for actual mining might turn out difficult. This is one reason why I implemented serial port communications in my version of the FPGA miner.

lolwut.

They plan on using a legitimate usb interface for the next gen chips as for the xilinix software, that it entirely independent once programmed a fpga can be anything, this is kind of the point of fpgas.

The next generation won't require Xilinx software. fpgaminer is working on the python interface, which I expect will run on Arch Linux ARM (http://archlinuxarm.org/). I have a Dockstar that I plan to experiment with, for this purpose. I'll let you know if I have any success (or if I don't). By the way, I set up a wiki at fpgamining.com (http://fpgamining.com) for the purpose of putting guides and stuff somewhere. Hopefully there will be some useful information there someday (nothing right now).

Pre orders are now enabled @ www.cablesaurus.com

It seems to me you're trying to solve a problem that doesn't really exist. Think "pentium 90" in terms of cooling here.

Just to be clear, are the 2 chip and 1 chip boards the same?

Yes. Whichever one you choose, it will help us get a volume discount on the PCB's. My team is going to buy 10 to start us off!   -- to sell later on eBay or for people who miss preorders

I am strongly considering preordering 10 as well.

I just ordered 2 of the dual ones. Looking forward to them!

I apologize if this has been answered before, I can only read back so much :)

What is the expected final price for the 1 and 2 fpga boards when they hit the market? (I assume this will be a max price and the price will be based in $ not BTC).

Also, paying the pre-order in BTC today, will that count as BTC when the final payment is due or as $50?

Paying in BTC today will count as $50. Max prices are $420 for single FPGA and $620 for dual FPGA

Unfortunately Digikey doesn't take BTC yet, so we will be converting to USD to pay for hardware. You have a couple weeks before the final paymet is due, so... if you feel like BTC value is going up, it would be smart to wait on your preorder. It won't hurt us

got 10 Dual FPGA X6500

do ur best guys can't wait see what happen next :)

So I'm putting in my pre-order, & being charged $30 for shipping.  ???

If it costs $30 to ship a promise, how much is it to ship a board?  ;D

Wow! Thanks. We will make it happen.

You guys take a risk I am with Ya ;D

I might considered getting one or two. When is the pre-order cutoff date? Any way to keep a running tally of what the final cost will get down to? Or is that impossible to know until production starts?

We'll make Sep. 12th be the preorder cutoff. We should have or be very close to having a first prototype by then.

Not impossible, but- it would take a lot of work to keep such a running tally. Maybe we can organize ourselves and give some periodic updates

I am concerned about communicating with the miner, because the original open source FPGA miner uses Altera's closed x86 software to talk over USB. Of course, there are ways to communicate via USB with open source software, but even then there can be driver issues. Serial port is safer in this sense, which is why I'm not the only one to choose it for FPGA mining.

Of course, the FPGA can be programmed with a different computer from the one that is using the miner. However, things are often much more convenient if it is the same machine.

Since I'm already starting to sound trollish, I may as well generalize that it's a bad idea to rely on closed software. For example, a friend of mine is worried about having an FPGA board that is only programmable from 32-bit Windows, as the support for XP is nearing its end. It's entirely possible to program and mine with Altera and Xilinx boards using open source on ARM, if you do your homework.

Synthesizing the FPGA code still needs closed software, but at least it is free as in beer for some chips. This LX150 needs an expensive license that must be renewed annually. I guess this board is worth its price if you only consider it as bitcoin-only hardware, and don't plan on developing anything else.

teknohog- you realize that the user "fpgaminer" has been helping us quite a bit? He developed the original open-source FPGA software. So don't worry about someday buying a new computer that doesn't work on your old FPGA board. We'll make sure you get current software and a bitstream. And we are switching to a USB==> serial communication interface. I assume that's pretty safe, since USB won't be going anywhere for a long time to come. The hardware that we use to implement this interface is extremely common.

He is right, though-- you do need a Xilinx license if you want to synthesize your own firmware for the FPGA. (Think of this as compiling your own custom code. All the vocab is different with FPGAs).

Most people who would want to do that probably have access to the software through school or work, I would bet.

EDIT- Here's the list of drivers for our interface chip.
http://i1116.photobucket.com/albums/k575/newMeat1/drivers.png

Thanks, this is nice to know :)

I want to support this,  but I am not in a position to place an order at this time.  PM me if there is anything I can do!

About to pre-order but have a few questions...

How does mining actually work with these? I see you need Windows for the driver, that's fine, but will it work with existing windows mining software?? Have you been able to mine with the prototypes?

Will the frimware become any more optimised do you think, or is the current quoted hash rate the most we are ever likely to get out of the devices? Can they be overclocked at all to give faster rates?

Is the FPGA reprogrammable? Do I need special cables or software to do so or can it all be done via USB?

What kind of warranty are you going to offer? This is experimental equipment so I would expect a few failures here and there...

fpgaminer wrote some scripts to do the mining. This is a picture of it in action, mining from Eligius-
http://lh5.ggpht.com/-uBIoMCfNHgk/Tk0q-Il_5iI/AAAAAAAAAYA/uWzDQE7JBqU/s1600/Test%25255B2%25255D.png

So, it's not going to work with your GUIminer or whatever mining software you use with GPUs. It has its own software. If you want a more detailed answer, feel free to PM fpgaminer. We still need to get a user guide written and hosted somewhere- I realize this.

Li has been mining non-stop for 5 days with our first prototype


Well, I hear that artForz has gotten up to 113 Mhash/s with this exact same chip. That probably gives you something close to an upper limit. So you will probably see some small increases in the future. That leads directly into your next question...


Yes it is. All you need is the USB cable.


Well, would it cause a mutiny if I said I don't think we can afford any type of warranty? Our profit margin just isn't high enough, and we know how people abuse mining hardware. But, keep in mind, we'll test everything before we ship it. I really don't expect "a few failures here and there".

Maybe we could look into doing something where the customer pays to ship it back to Li, and he tries to fix it. He already fixes Xboxes and PS3's so I think there's a good possibility of this... Check his blog out to see him repair a cellphone. It's pretty cool. (http://fillwithcoolblogname.blogspot.com/2011_07_01_archive.html)

newMeat1, that's the mining software for the old board, and the new board is expected to be very different. Also, I'm not sure where you got that driver list... drivers are available for Windows, Mac, and Linux. The software is still in development, so the actual requirements to run it are still to be determined. fpgaminer is in charge of developing this software, so any questions about it should be directed towards him.

If the lack of details available at this point make you uncomfortable, please don't pre-order yet! More specifics will be ironed out soon.

Nice work everyone

-rph

The LX150 is a 484-pin BGA part... Do you send the boards to a professional assembly house for the BGA reflow, or do you reflow it yourself?

How much can I buy this board for with all needed items to get up and running quickly?
How much to buy? in USD and GBP
Academic discounts (as I work in an academic institution)?
Where can I buy them from?
Power use?
MHash/s performance?

Thanx

I do the BGA + QFN soldering with a soldering station at work.

Unfortunately we're sold out on the initial batch of boards. We're working on a better board the next round and will have firm pricing later.
Power use and performance expected can be found at the 1st post.

Well, we have 42 preorders so far. That will help us get some nice bulk discounts. To give you guys an idea-- We have already dropped the cost of our heatsink by $2.51 due to this bulk order. With another few units sold, it will drop by another $0.87. So, on a dual-FPGA board, we are saving over $5 on the heatsinks alone. Thanks everybody for making this happen!

Nope sorry. We aren't like some industrial giant that charges outrageous rates for commercial customers, then gives stuff away to students practically for free.


cablesaurus.com (http://cablesaurus.com)
For a dual FPGA board, expect less than 15W power draw and >=180 Mhash/s in real pool mining conditions

Actually, the size of the order is still going to be quite small, despite the 42 pre-orders. Unfortunately, Cablesaurus can't use any of the deposits that came from Paypal, because of the risk of chargebacks. None of the Paypal money can be touched until the actual order is shipped with a tracking number.

It's great to see that there is so much interest, though! As of right now, the $420/620 price point is still about right, with only a small profit in there for us to compensate us for our time.

The biggest problem with quantity discounts is that we get absolutely no discount for ordering FPGAs in bulk. It's going to be the same price to buy 1 as it is to buy 50. All of the rest of the parts are minor compared to the FPGAs, so I don't see the prices dropping as significantly as newMeat has hoped. Of course, we'll do our best!

So since there are no bulk discounts for FPGA's, and they are the most expensive part, you don't have to buy them until you get the full order in for people (or maybe just before).

If you need front money just to buy the boards, and some of the cheaper stuff, you should let us know. I am willing to front some of the money for this, or pay for entire cards up front if it helps you obtain discounts (which you can hopefully pass on to us somewhat).

QFN, sure, of course.

But BGA?  A soldering station isn't really any help with BGA devices -- the pins are underneath the chip; you can't get at them with an iron or hot air gun.  The only approaches involve heating the whole board and chip using a convection oven or hot-plate of some kind.

I guess I'm just curious to know if you're using a PID+Toaster-Oven approach for the BGA part; I've heard very good things about this approach and have been meaning to try it myself.  It would be nice to know at least one person has gotten it to work with a 484-pin Spartan chip -- these parts don't get used very often in small-run products.  Have you gotten this to work?  If so, do you have any tips or hints?

Thanks!

I think that transparency is really good here and it builds credibility.  We are not you typical buyers either I don't think anyone here would scoff at you guys getting a fair margin for your efforts.  I personally am amazed at what you've created.  I can't wait to see 8 or 16 FPGA's on a single board, ;)   (is that possible?)   

That being said what is the price on the FPGA's?  per unit?  I assume the 150 Model is the best price/performance hence its choice.   

Also I love fabricating so perhaps you could over a heat-sink less version?

so you get about 160 mh/s with one of these for what was it $450?? i dont understand how these are better than buying a graphics card for 150 and getting 350 mhs

No I'm not using a toaster oven, at my current workplace I routinely fix Xbox360s and PS3 consoles with a IR soldering station. That's what I use.

Its a lot more complicated than that Arij!

Yes you are going to pay around $450 and get only 100-110 Mh/s it seems,  thats almost 4x-5x more cost than buying GPU's.  But these things are almost magic and this is why.

They connect via USB, So your rig can be anything that has a internet connection (and likely free...,  This means no more $300 Power supplies.  Hell you could run 10 of these cards on a old ATX 300 Power supply from a computer going to the landfill...)

They consume roughly 1/8th 1/10th the electricity.  Perhaps less?  So every month the power savings will slowly eat up the 'savings' you think you have buy buying a gpu.

Really in theory it seems you can put over a 100 of these on a single computer with enough connectivity.  

Its really amazing!

So the SheevaPlug might work if the mining software will run on ARM. But you still need to find a molex connector from somewhere. The usb does not have enough power.

I imagine in 1-2 years, you guys will come out with gen10 that is just a device (like the SheevaPlug) that has has 10x FPGA inside and a ethernet port. You just plug it in and it will mine at 1Gh/s. Now, that would be cool. :)

Hmm, I think I must be missing something. How is it possible that the estimation for 20 orders was $385 and now with 42 pre-orders the $420 is still right? Is it only due to the "order" ↔ "pre-order" difference, or are there more reasons? Mind you, I'm fine with that, just a bit surprised about the difference.

That's a good point, but then you have to deal with shipping on each of those FPGA orders. Li works in Singapore, so that's international shipping, too. Also, like I said, we can't really use any Paypal money to order parts until the finished board is shipped to the customer. We're working something out, though, so we'll get back to you soon.

The X6000 and X6500 boards are being designed and tweaked as we speak, as well as the logistics of fulfilling orders. Consider those ballpark figures, rough estimations based on the information we have at hand. Everyone is working diligently to get good price quotes, but it's a little bit difficult when the technology is still in flux :P It's like trying to figure out how much your new PC will cost, with an ever changing landscape of discounts and rebates, and you're still waiting for your last paycheck to clear before deciding whether you want blue or red LED fans :P

Please don't take this post too negatively. I do appreciate what you guys are putting in and I think you should be compensated appropriately. But in the future, please try to come up with better estimates on the price before you ask people to preorder. If you are going to say that >20 orders means $385, then you should stick with it. If not, please don't promise something that you may not be able to deliver. I'm guessing a lot of people ordered them expecting it to come out cheaper than $420 because they saw multiple people already pledging to order 10 of them. I ordered 2, so it doesn't make that much of a difference to me. But still... since everyone who preordered already have $50 in, if you don't give a discount, people will still pay you $420 each, and you guys can pocket the savings on the bulk order. I'm not saying you plan to do that, but there's really no incentives for you to pass along the savings to us since we are already committed to the $420 price. Know what I mean?

I'm still optimistic that we can get some good cost savings. Some of the other team members are more reserved. We are going to be going over spreadsheets in the next 24 hours to find out for sure. So look for a better answer soon.

We absolutely aren't going to pocket the money due to discount savings. We have a minimum profit in mind that we are going to split for each board. If bulk discounts save us more than that, it will be passed directly on to those who preordered.

I believe that. Keep up the good work. Looking forward to receiving my FPGA miners.

At 400 each and 100 mhash even with low electricity cost it would still take over 3 years to make back you investment at this diffculty.....I think I'll wait

For the dual FPGA one, at $620 and 200 mhash, it will take a year and a half. It's definitely not very cost effective right now. But I'm definitely supporting this project.

Let me know for real. I am happy to pay for a dual FPGA up front right now with bitcoins if it helps you raise the needed capital to fulfill all your preorders.

That's interesting. Someone who replied to one of my /. comments reckoned FPGAs were about half the retail price if bought in bulk from the manufacturer... of course he could be talking BS or the required volume could be higher than you're going to reach, and the lead times are probably horrid.

In my experience ordering microcontrollers, bulk discounts are usually pretty small under 100, and increase substantially in the 1K-10K range.

Actually, newMeat is planning to call them, but I'm skeptical he'll have much luck. I expect the quantity needs to be more like 1k, and that's definitely way out of our reach.

50 != bulk.
Hell, that's not even one tray.

^
^
^

This post from somebody who is rumored to hold 25% of the network's hashing power.  (https://en.bitcoin.it/wiki/ArtForz)  We are honored.

But, we are now closing in on 100 chips needed.

There have been 2 inquiries about orders in the triple-digit sizes, we'll see if either of them pans out. That's another reason why we have to look at our numbers very carefully.

That was a loooong time ago, nowadays it's < 1%.

lol probably still better than my 0.0001% but its nice to see these fpga orders pick up. Once/if the price can get down to somewhere around 2$/mh you'll probably see more mass orders. Because imo fpgas are more ideal for bitcoin mining especially for long term profit.

You guys are all smartypants, so let me know if I am correct.

Let's say sometime down the line we switch to SHA3 (or whatever new hashing algo).

FPGA's/GPU's/CPU's would be fine, while ASIC's would be hosed, correct? Or am I wrong?

All mining software would have to be redeveloped, and re-optimized. ASICs would then be useless, but I'm sure you could find a buyer in the server (crypto accelerator) or hacking community if you had a mass number of ASICs that calculate SHA256...

That is what I suspected. Thanks!

what about anticipate 25% or more (just ask for a start up budget) trough Dwolla or BTC . What I try to say if u need $100 today as now we give to u BTC 9.26 as today the value is BTC 1=$10.80 and you have to convert them and spend for what ever u need! This is why we do this right?! let's start to think how we can use this BTC and mediate other currency as we need! We have to spend BTC to make it worth it other ways it will stall all this system!!!

I don't need Escrow, I am doing this on trust we already take a risk to freez money trough paypal the deposit thats already an escrow! If you need cash, well I reading there is some people like me to help so let's get focus to make things done in some ways!

Thank you for reading!

BTW if u need to figure out to finalize the price it's okey long of the way in some what, you will not drop the price up to 50% and if u do so mean we can have more cards

You are wrong. For the majority of Bitcoin users who pay an average worldwide electricity rate of ~$0.10/kWh, buying a $450 FPGA board for only ~100Mh/s makes absolutely no sense from a financial viewpoint, given the current Bitcoin difficulty/exchange rate ratio. You wrote your post based on feelings, not math. I encourage you to run the numbers yourself.

The only rational decisions for buying these FPGA boards are either (1) your power is a lot more expensive (2x or 3x more, eg. $0.20-0.30/kWh), or (2) you bet on the difficulty/exchange rate to sharply increase in the very short term, or (3) you are significantly power-constrained (eg. want to run a mining farm out of a small apartment).

I would be happy to pay another deposit in BTC to help with the funding, if cablesaurus can work it out.

That's the UK, then. Ave price (before recent ~20% rises) was £0.13 / kWh, about $0.21.  :(

Here in northern California, I'm paying $.34 per kWh. So the math works out a bit better... still not cost effective though.

Well these are approximate price ranges I quoted. Personally, at $0.20/kWh I would be on the fence... and probably still keep my GPUs.

Doing both... keeping GPU & expanding into FPGA. Lowers ave cost to produce a coin, but increases ave capital cost / MH.

I'm expecting difficulty to increase, electric cost to increase, & coin value to ??

If value increases, gpu is viable longer, if value decreases, I can always switch off the heaters!

Ignoring capital, I won't mine for less than electric - that's crazy. I'll tolerate no return on capital because I enjoy the project.

Just me...  :D

A daily update for you guys...

fpgaminer tested out our USB communication interface on one of his dev boards last night. This is a huge step forward! He was able to read the ID codes of all the chips correctly. He said a bit was getting lost somewhere when he tried to communicate, so he will be trying to hunt this bug down in his code. Now that we're 90% sure this communication idea works, we'll be able to finalize some board design details. It looks like we are sticking to the original schedule so far.

Li has been overworked recently with soldering and stuff, but he showed us a preliminary cost breakdown today, for 1-10 units. He's still waiting on soldering and PCB quotes. Expect him to post a price list soon for 1-100 units. If you preordered a board or two, you can figure out the final amount owed by ( final shown price - $50 for your preorder )

Thank you  for the update :)

fpgaminer testing on the dev board...

http://bitcoin-mining.com/DSCF0044.JPG

How many amps does it pull through the Molex?

I am looking at external molex adapters, and am wondering how many boards I can run off one.

Let's see... 15W power draw for the 2-FPGA board. We will be using the 5V side of the Molex on this next batch (not the 12V side). So 15W/ 5V = 3 Amps

Damn, most of the external molex online seems to use the 12V.

Will continue looking!

A cheapie power supply like this one (http://www.newegg.com/Product/Product.aspx?Item=N82E16817182006) should work fine for about 11 dual-FPGA boards, although if you're spending that much money on FPGAs, you'll probably want to buy something a little more efficient and reliable. :)

There's one caveat, though. You'll need to short the /PS_ON pin to GND on the main connector. Cablesaurus has mentioned selling some dummy plugs for that purpose.

Also, the boards will still accept 12V as the input if you'd prefer to use that. You'll just need to use an adapter to move the 12V pin over to where 5V usually comes out. You could do that with one of these (http://www.amazon.com/Fan-Molex-Pin-MB/dp/B000BSJGL0), but splicing the 12V line over to the 5V line. Maybe we can arrange for Cablesaurus to make and sell custom adapters to do that, as well.

Supposedly there's some trouble with the line regulation on the 12V rail when you don't have a significant load on the 5V rail, though, so you'll want to be aware of that. That's one reason that we're using the 5V rail instead.

Hi everyone.
I've just get out from the newbie section, but I'm monitoring this topic and few others topics related to FPGA for a few weeks. I'm interested in buying one of your dual FPGA boards....
And few questions comes to my mind...
Is there will be an option to buy a self assembly kit? (Yes I know what I,m doing) for a few less $$ (BTC?)
Why not to use SATA power connector to save some $$. There is a 3.3V rail that can be used to power the VCCIO (FTDI chip also?) and after cheap LDO linear voltage regulator to supply VCCAUX on the FPGA side. One LMZ is out, replaced by less then 1$ LDO.
5V rail? LMZ specification says that it can be operated from 6V... There is almost the same efficency when operating from 12V rail, but less current will be drown from that rail... So there will be option to drive four LX150 from one SATA power connector...
And that's it for now... to much beer for tonight ;)
Some more questions (sugestions?) will come later....

PS. Sorry for my English, it's not my native language...

Possimably, we've yet to set the final price yet, once we work all of that out I'll see what kinda discount you get.

Some of the PC PSUs I know, don't have the 3.3V stuffed. The upcoming board will work with anything from 5-19V, so it's up to the user if they wanna make some kinda adapter cable or what not to put on their own PSU solution.


New boards most probably won't be using the LMZ modules, I'll update as I go along the design decisions.


It isn't too bad actually.

I'm so happy!

BTW: Have you chaps considered this hardware for future boards? >> http://www.tabula.com/technology/technology.php

Quote from Dexter

I've used LDO's before. Isn't it a problem that LDO efficiency is based on Vout/Vin? So, if we go from 3.3V to 2.5V, that's only 76% efficient.

Since Vaux uses less than a watt, however... maybe this idea is worth looking into. Thanks! I will have to see if any LDO's meet Xilinx's specs for the power supplies.

Well, I placed my pre-order on a X6000 unit. I hope that means the second series without the need for the system cable. I'm in Australia so hope the shipping arrangements will be OK. I plan on going to the Bitcoin confrence/expo thing (https://bitcointalk.org/index.php?topic=38867.0) in Melbourne 18 -20 of November, as I'm planing to put together a standalone solar mining rig. This will be a great unit to incorporate. I wonder what the most minimal type of processor/board I can look at running for the software. Maybe a simple embedded system board running a Linux kernel?

Oh! Great work guys. I'm very impressed.  ;)

Hmmm this is brand new to me. It seems to be an FPGA that reconfigures while it's running, to stay optimized. Very interesting. I'm sure it's way too expensive for us right now

I can answer some of Karmicad's questions...
That's right. I'm sorry this thread is so cluttered now, we should start a new one with the details clearly marked See link in my signature for FPGA details.

International shipping hasn't been a problem so far. I'm sure you'll be OK

I'll try to have somebody else get back to you about the minimum computer

EDIT- And thanks for your order!

Has anyone looked at these? I posted over in the other FPGA thread and didn't get any response. Pricing looks to be in the $200 range.

http://www.achronix.com/products/speedster.html

It's hard to say about those achronix chips. They don't give any prices or even much data about them. I guess they are probably really high-end chips. Like, "If you need to ask the price, it's too much" kind of thing.

Xilinx datasheets reports that current drown from VCCAUX is 600 mA max. So, from 3,3V drop to 2,5V means 0.8V. Multiply that with 0.6A and theres less that 0.5 W of loses. Every cheap LDO with output current of 1A can do this without any problem. Yes, it's something around of 75% efficiency, but consider price below 1$ for LDO and 20$ for LMZ. Difference in loses are something around 0.25W. I've read that prices of 1kWh are something around 0.2$/kWh. Thats 380000 hours when using of LMZ starts to pay off !!!

Well, I guess there are some other considerations--

The board design is 95% done so we're pretty well locked into using a 5-19V supply at this time (no 3.3V). If we used an LDO and somebody powered the board with 12V, it would only be 2.5/12= 20.8% efficient. And all that extra waste heat wouldn't be pleasant.

Besides, our new regulators are cheaper than $20.

PSU 24-Pin ATX Power-On Dummy Plug (https://cablesaurus.com/index.php?main_page=product_info&products_id=45)
https://cablesaurus.com/bmz_cache/c/ce31cccf755e3012ffdf6da9d94245f9.image.200x150.png
* Allows the powering on of a supplementary power supply
* 24-Pin ATX Connector

Here are the dummy plugs which will allow powering on of a standard PSU, the PSU can then provide power to FPGA cards and stand apart from the laptop.

Or for about 1/1000 the price of one of your PSU dummy plugs you could cross the green and any black wire with a paperclip.

Heres the other post I was referring to,

https://bitcointalk.org/index.php?topic=9047.msg472048#msg472048

Those two articles I linked are the only place I can find the price mentioned. I'll try emailing them using my university email and see if I can get a response... If these FPGAs are as fast as their advertising leads you to believe, $200-400 depending on which one might be a very good deal.

Thanks.

No Probs. If it does the job, the pleasure will be all mine (get it? - MINE) http://i53.tinypic.com/2njaetd.jpg

True, true...

That is a knowlage only you poses :)

About speedsters...
I can't find it now, but I saw somwhere a price for their development kit. It was way beyond 1000$...
Intel just starts manufacturing in 22nm 3D gate process, so speedsters22i won't be avaible until 1Q 2012...

QUOTE:
"The SPD60 is currently available to customers. Achronix says that volume pricing for the Speedster FPGA family will range from under $200 for their smallest part (the SPD 30) in high production volumes to $2500 for their largest part (the SPD180) in low volumes. Given that the SPD60 is in the middle of the size range, it’s fair to assume that, depending on volume, it will probably cost between $800 and $1500."

SPD60 have 47k LUTs, and it comes in FBGA 1285 package...

See, I was thinking of pairing your unit with something like this for WiFi:
 

I can't figure out what processor is on there (ARM perhaps?), but it must have one, because is says it's running the eCos - OS (http://ecos.sourceware.org/). Unfortunately as far as I can tell, eCos only allows for the running one process. On the WiFly, that would be the TCP-IP stack. If I can somehow get away with a hack to use only the WiFly, that would be very cool.  8) I'm now wondering if given the advantage of WiFi, if my standalone units, could simply run the software process, server side over the network (The  WiFly does do telnet). Therefore I'd have no need for a processor other than the one in the WiFly unit. Thoughts?

The wifly just takes data and pushes it over wifi, that's all. You would still need something else to communicate with the wifly and the fpga.

Thank's inh. Thought the thing had processing power of it's own. Just don't understand how it can run an OS & software without a processor.  :-\ Have to look at something more substantial I guess.

Karmicads- check out this earlier post:
https://bitcointalk.org/index.php?topic=37904.msg469658#msg469658 (https://bitcointalk.org/index.php?topic=37904.msg469658#msg469658)

He had a similar question. I'm *pretty sure* his board choice would work, but nobody has tried it yet.

Keep in mind you also have to get 5-12V power to the board somehow. Using an external computer Power supply would work pretty well, you just need to cross a wire or use a dummy plug like Cablesaur is selling (see above)

 :o Egad! That's a bit of overkill isn't it? I meant the motherboard of course.  :-\ I don't need a PCI slot, or video, or most of that stuff. Just enough processing power and memory to run whatever the X6000 software requires and a USB serial interface. Can you think of anything else I'd need on-board? I was thinking more along the lines of an ARM embedded system or something minimalistic. For that price I could buy a tablet PC running android. Don't mean to be difficult, but this isn't likely to be a one off thing. I rather fancy the idea of running the software remotely on a server if possible.


No probs. This baby's going to be bolted to a set of solar panels. So low voltage from go to whoa.  ;)

Hi all,

We've decided to temporarily halt pre-orders at this time. The response has been huge, with close to 70 total pre-orders! This exceeded our expectations, and we now need to focus on filling all of these orders and keeping you guys (and ladies) happy.

Stay tuned for us to get the first working prototype running. Once we get the technical details and logistics settled, we'll get back to you with a firm price point.

For those who haven't pre-ordered yet, watch this space. As soon as the pre-orders are handled, we'll let orders come in again.

If you hook some of these mining boards up to solar panels, I definitely want pics  8) Very cool.

Or better yet sell us a kit to replicate your setup :)

Alternatively, you can buy the connector and a crimping tool and build your own. (Plus, that way if you're electronically inclined it should be fairly trivial to rig up remote power-cycling - an Arduino, a NPN transistor and some simple software ought to do the trick.)

That reminds me - one of these days I keep meaning to write some deeply-embedded mining control software for ARM microcontrollers. For your purposes something like a Fonera 2.0 or some other hackable routing hardware with a USB port might be better. Depends what the developers come up with and how hard it is to communicate with.

I'm working on testing out the software on a Dockstar in the coming days. If it works, I'll definitely write up a guide on how to do this.

It'll just be the one for prototype/demo first up. But yeah, I post some pics for sure.

Well, I could do that I guess, but the whole thing is intended for a larger product/service package deal, which I can't go into until I have some non-disclosure agreements written up. It's a standalone unit just for the solar powered mining part though, so I don't see why a separate kit couldn't be feasible. Thanks for that idea Pipesnake, I'll look into it. You'll be the first to know.  ;)

Can anybody please explain what the objective of this is? I have an excellent crimping tool and reasonable electronics knowledge, but I don't get what the problem is or what this dummy plug is supposed to do.

EDIT: Wait! It's a female isn't it? so you don't have to plug the PSU into a motherboard. Now I think I get it. Er.. but then again, what's the Arduino, transistor and software stuff for?  :-\ Maybe I'm still lost.  ::)


And then I might kiss your boots.  :-*


I'll check that out. Otherwise I might have to collude with these guys and see if they can come up with some custom design for me. Trouble is, my skills/knowledge aren't meeting in the middle. I know a little of programming/computing/networking and a little of electronics theory, but all the microcontrollers & hardcore digital system stuff is getting me in over my head. I'm just working it out and learning as I go.  :)

About using an external power supply to power our FPGA boards--


I was confused about what that jumper does, too. It just sends the PSU a signal, telling it to turn on. Usually the motherboard does this. If you run the power supply externally, the motherboard isn't around, so you need to have a jumper to another source. (or, well, ground in this case...)
http://www.overclock.net/faqs/96712-how-jump-start-power-supply-psu.html (http://www.overclock.net/faqs/96712-how-jump-start-power-supply-psu.html)

*Disclaimer: Paper clip method not recommended, due to lack of insulation and possibly not enough current capacity.

Current capacity is negligible, all it has to do is overcome an internal pull-up resistor in the PSU. A little electrical tape should take care of insulation. No need to buy any kind of dummy plug just to turn on a PSU.

In that case why not just pop the cover and bridge the pads with a bit of hookup wire. Alternatively, I think there are two wire bridging crimps you can use with normal pliers. They fold over the two wires and have little metal jaws, that cut through the insulation and make contact with the wires inside. Neat, simple and cheap.

Edit: I found a Pic:

 :o Egad!! I just looked at that software from Xilinx it's 4.62 Gig. Good grief, what is all that code for? It says it's a design suite. Is that really necessary to run the mining environment for the end user? I thought the idea of a FGPA was that it does in hardware what would otherwise have to implemented in software.

I'm still not convinced that I couldn't get away with running whatever software is needed remotely, over a network connection. Why can't a WiFi / Ethernet link support the FPGA board with software being run remotely? Assuming we have serial interfacing on both devices, the need for a dedicated CPU/memory/board should be eliminated - well, I imagine, at least. I want to connect these units over a VPN/WAN, so although they are stand alone (physically), server support and remote process serving should not be an issue. As far as I'm concerned the software could be deployed in a virtual appliance. In fact, that would be all the better for data center management (upgrades, security etc.). The end user environment IMHO, should be as simple and abstracted from system management as possible. It should also avoid physical on site maintenance of individual units at individual locations. So I fully intend to leverage cloud based solutions and virtualize whatever is possible.

Also, I forgot to inquire. Would this FGPA miner afford any advantage towards R&D of an asic based solution? Is there any/much benefit of having a good FGPA based design to leverage a future asic device? Just out of curiosity. 

The X6000 and X6500 boards will not require an ISE installation at all. I am writing the software to control these latest boards, and it will be very light weight.

It most certainly does. In fact, there are many companies that can take an FPGA design and convert it to ASIC.

We started a new thread for details about the next generation FPGA boards (the X6000/X6500s):

https://bitcointalk.org/index.php?topic=40058.0

This will hopefully be the best place to look for updated information on these boards in the future (and keep it separate from the X5000 information).

UPDATE4 Aug 30th 13:30 - Received package containing FPGA miner "built with care". Also received Platform Cable same day :)

now what? I'm afraid!

Go bravely Molecular, you pioneer.  Just remember that the whole bitcoin world is watching you and counting on those trembling hands. :o

Feel safer now?  :P 

Hmm, [mumbling, cig in mouth], just let me get this, .... power ... hmgrbl ... plug ... which way w.. BRATZL BRUTZ ZAPPP

Where did you hear that?

Sounds like their parts aren't available from retail distributors yet (nor is there any timeline for that to happen).

You should put your company on the GLBSE and distribute shares to raise funds.  I'm sure alot of investors will support you.  I had no idea people were doing such things, I must visit this forum more often.
 

you mean this (see below)

Yes exactly that.  Something very exciting to invest in.

https://bitcointalk.org/index.php?topic=9047.msg472048#msg472048

https://i.imgur.com/tL4Ht.jpg

Looks awesome, right. Green LED and all ;) It get's better:


YESSSS!

I got it working, on LINUX! fucking hell yea!

Li, newMeat, should I write up a little linux guide. It's not been trivial ^^

EDIT: changed image to include eligius screenshot of hashrate

UPDATE5 Aug 31st 16:53 - After initial disappointment that I might have to use windows, I got the X5000 to work on LINUX (see above post with pic)

One question: is it safe to plug in/out the power plug while the PSU is on? I'd really hate to have to power off my system for this (my X5000 currently runs on it's own PSU but I want to switch to the one used by my wall-pc)

Oh yeah! I forgot to mention. DON'T LET THE SMOKE OUT! These gizmo's really work better if you keep the smoke inside of them. ;D

Yeah, "hot-swapping" should be fine.

You total legend!!  ;D  and 100 MH/s no less. Sweeeeeeet.  8)

Good job, please do share with the community what you've done to make it work. I'm sure it'll be an interesting read at least.

OMG it's so cute!!  I can't wait!!

 :D

Yeah, well. Lots of luck in a 15 minutes measuring window. It goes from 80 to 130 if you look at 15 minute average. That's not the X5000, of course, that's just fortuna playing with it ;)

Allright, here we go: Linux guide to X5000 mining:

Step 1: start download of Xilinx ISE (it's development environment. There might be other ways to program the FPGA, but for now, I used this, as Li suggested in the windows version of the guide). The file is huge (4.6GB), so we start this now to avoid coffee breaks later.

Step 2: Get the USB Platform Cable to work (I use a USB platform cable here, there are others, this guide is for USB Platform cable, I have a "Model DLC9G")

Step 2 A: Plug the "Cable" (it's called cable, although it's a box, and I will just call it "Cable" or "Platform Cable") into a USB port. Better try without a hub first. Check if the USB portion of the device works:


Step 2 B: Driver for Platform Cable

Before we start, lets make a directory to work in:



My gentoo linux did not have any sort of driver coming with it and there is no module in vanilla kernel 2.6, at least not in mine.

The Xilinx ISE actually comes with source for a kernel mode driver, called "windrvr". I tried to compile this but I failed miserably.

I used this: http://rmdir.de/~michael/xilinx/. It's a usermode driver that uses libusb (or parport for parallel port cables) and emulated windrvr afaik. Download, unpack, compile it and check it created the lib we need:


Now, this software comes with a script that copies firmware-files from the Xilinx ISE install directory, this means we have to install the Xilinx ISE now:


I checked the option "install driver" here. First I ran this as user and after unzipping files for an hour, the installer told me I'm not root and therefor drivers could not be installed. I did the whole shebang again as root, and this time it told me that driver installation failed. Since we do not use the windrvr that comes with the ISE, screw it. You can try, though, maybe it works for you.

I chose /opt/Xilinx as install directory. You can choose any dir, just replace that in the following

Now let's run the above-mentioned setup-script:

This installs some udev-rules in /etc/udev/rules.d/xusbdfwu.rules by copying from xilinx folder somewhere. They already mod these xilinx-rules for newer udev versions, but it seems udev changed again, so I had to edit that file:


change all occurences of "$tempnode" to "%N" (uppercase N)

you also need a tool called fxload to upload the firmware to the device, install it (this example is gentoo)


now if you replug the USB Platoform Cable, the firmware should load and the LED on the device go from "maybe very dimly red" to "BRIGHT RED", when the firmware is loaded. If x5000 is connected, it will go green.

If it doesn't work for some reason, set udev_log="debug" in /etc/udev/udev.conf to see what's going on in the syslog when you replug.

You can also load the firmware manually (look in the rules file and lsusb which one by device id):


At this point, the driver should be working and the LED on cable should be either GREEN or RED depending on wether x5000 is plugged in or not.

To check if driver works libusb and the firmware works, you can use another tool, urJTAG (http://urjtag.org/):


maybe we can even program with this tool, I don't know. Anyone?

Step 3: program the FPGA using ISE:

Now you can start impact, which can be used to program the X5000 fpga, preloading the usb driver:

impact should start up and ask you some stuff about project folder. Just answer as you see fit. It also asks you "Configure devices using Boundary-Scan (JTAG)". You can try "Automatically connect to a cable...." here, but it never worked for me. Instead, I select "Enter a boundary-chain manually".

The gui starts now. Important: Do  "Menu: Output -> Reset Cable". This helped me to get the cable working at all. It did not work for me without doing this.

No you can select "boundary Scan" on the left and then Right Click on "Right click to Add Device....". and select "Cable Setup...". This is also in the "Output Menu".

If it says: "warning, windrvr not installed", you probably did something wrong with the LD_PRELOAD above.

Select "Platform Cble USB/II" (at least for me that's what I wanted) and then check out "Advanced USB Cable Setup", it should show an entry in the list of usb cables/ports. If it doesn't, exit that stuff and try "Output -> Reset Cables" again.

If it says: "this function requires a target to be set first", it's probably working ;). Load the xxx.bit file from the miner.zip by right-clicking on the white "boundary scan" areay and selecting "Add xilinx device...". Then try again with the Cable Setup.

The cable works and a nice Xilinx image popped up in the white area? Then you can program the fpga by right-clicking the image and selecting "program". Took 11 seconds for me, the bar didn't go all the way to 100%.

Step 4: mine!

go to the directory where you unpacked Miner.zip and make a script, mine.sh



note that this script quite sucks, you probably have to change a couple of paths.

edit the configfile... and start to mine


you might get some errors. If tcl complains about missing packages, try installing tcllib.

Ok, that's it. Hope it helps someone.

• User Mode Platform Cable Driver: http://rmdir.de/~michael/xilinx/
• Info about udev rules $TEMPNODE: http://ubuntuforums.org/showthread.php?t=1313572 (maybe outdated, but still interesting)
• Linux udev manpage: http://linux.die.net/man/7/udev
• urJTAG: http://urjtag.org/

I think it's fitting that molecular is about to become a Hero Member.    :D

He's right about the variation on hash rate from our boards. They are pretty much hard-wired to run at 100 MHz right now, doing one hash per cycle. Any variation on that is due to latency communicating with the pool or something like that.

Lol. Thanks for the flowers, but not all of my 498 posts are as long and contentrich as the above one ;)

That's a fantastic guide, molecular.

Can't wait....

Any ideas if the mining scripts will run on a mac?

Nope, because none of the usual FPGA tools support Macs. (Someone I vaguely know was complaining about this on Twitter the other day.) It's possible someone will write an alternative that doesn't need them and works on Macs at some point but I wouldn't hold your breath...

Edit: Of course the new board version and the scripts for that might, once they're finished, but...

With the X6000/X6500, you don't need to program the FPGA chip, right? So we just need the mining script to work on the Mac. Does that script support the Mac? Anyone know? I'd love to be able to plug this into my iMac and start mining.

Don't know osx a lot, but let me do some guessing from my experience with linux.

The script itself is in tcl, that should be no problem. Also probably libusb should be right there in osx, right, so it's a matter of compiling that "libusb-driver"? You'd also have to see about the firmware upload. Don't know about the Xilinx ISE or what else could be used to smolder the logic onto the chip.

So: I wouldn't count on it and it might require some fiddling, but chances are it's doable, I'd say. You'd best need someone who knows his way around osx.

https://i.imgur.com/C6Yi5.png

Maybe I can wake someone up ;)

I'm on deck! I have my preorder in. Are hardware schematics available to study?
 :)

Hey kazoo,

Thanks for the order! first of all. We haven't released any schematics yet. I don't imagine we will anytime soon, because FPGA's are turning into a hot field. Wouldn't want to give away our secret sauce. But, if you send me a PM, I'll answer almost any hardware question you want.

Where can I buy these boards? What range of boards can you offer?

Nice work, molecular! Do you mind if I put your guide up on fpgamining.com?


It still needs to be tested, of course, but I see no reason that the X6000/X6500s won't work on Mac, Linux, and Windows.


This particular board is done, and the next generation is about to come out. Here's the thread for it: https://bitcointalk.org/index.php?topic=40058.0

Do you have the capability to make a custom board once you get the FPGA design to 150+mhash/s?

I'd like 8 chips on a single width 1x pci-e slot . I'd be willing to pay 1$ per 0.75 mhash .. About 1600$ per card as long as you can get the design above 150mhash/s average.

I'd want 6 cards.

Also, Would you be able to build a board if I sent you the chips? I managed to find a Virtex 5 LX240T for 100$ and 4 Spartan 6 LX150s for < 150$.

Sure, I mean: no, I don't mind.

Great! It is done: fpgamining.com/index.php?title=X5000_Guide (http://fpgamining.com/index.php?title=X5000_Guide)

cool!

A quick guide how to synthesize a project in ISE WebPack 13.2 would also be very helpful. And for Altera users how to do it in Quartus...
I'm trying for 4 hours at least and nothing. Errors, errors... I've tried to synthesize LX150_makomk_Test. Becuse WebPack have limits I've change LOOP_LOG2 to 1. Any help?
TIA.

For ISE you simply load the xise project file, make sure fpgaminer_top is selected in the Hierarchy in the upper-left, and then double click "Generate Programming File" in Processes in the lower-left.

For Quartus you can just load the project and press Ctrl+L. There should also be a menu options under Processing.

Let me know what errors you are getting or what the specific trouble is. Please note that the LX150_makomk_Test project won't compile on WebPack. You'd have to target a different device (LX75 or below).

Fantastic work, molecular! I freed up a Linux machine last week for my lab, so I can certainly give your guide a try :)

Okay, here's the deal with UrJTAG. I wanted to completely replace the ISE requirement on these X5000 boards by using UrJTAG which works on multiple platforms, at least Linux and Windows. Using UrJTAG would completely eliminate the need to use ISE for mining, because it can both program the FPGA and communicate with it to push and pull work.

Now, UrJTAG states explicitly in their documentation that support for Xilinx Platform Cables is experimental and slow. I figured I didn't have any other options, so I gave it a try anyway. Long story short, I can program an FPGA using UrJTAG and my Platform Cable, but it takes ten or more minutes to do so. I also have terrible trouble with the drivers, which need to be re-installed every time I boot my system.

I have not given up on using UrJTAG. I dug into its code for xpc_ext. It's fairly simple code, and it looks like the reason why it's so slow is because they're using a slow, GPIO bit-bang mode of the platform cable, whereas ISE would normally use special JTAG specific modes. So with some hard work it might be possible to fix UrJTAG's support of xpc_ext, and I have that on my TODO list. Don't expect it next or week or anything like that; I have a lot to clear off my plate and the work is non-trivial. But it would be kind of fun to get working :)

I'm synthesize it for LX75.
Errors:
HDLCompiler:1654 - "...\LX150_makomk_Test\hdl\sha256_transform.v" Line 290: Instantiating <e0_blk> from unknown module <e0>
HDLCompiler:1654 - "...\LX150_makomk_Test\hdl\sha256_transform.v" Line 266: Instantiating <s0_blk> from unknown module <s0>
... and 4 warnings about identifiers being never used...
TIA.

So many good and bad news in one post ;)

Awesome that urJTAG can, in fact, be used for programming, I was really hoping that. Too bad it's using this bit-bang mode. 10 minutes is a tad too slow.

I also noticed another problem (not urJTAG-related): Once I plug in some other usb device (a hub in this case) on another port of the computer the miner is connected to (haven't verified if it's the same usb host), the "cable" gets "stuck" (for lack of other words). When starting the miner, it says "Cable locked, unlocking...." (or similar), but it does not succeed. The workaround is to use "Reset Cable" in Impact. Any ideas about that? Does it happen to you?

Holy shit, I thought I fried my x5000 for a minute there.

I fed it 12V. The tiny fan of the X5000 immediately gave me an audible clue (lawnmower-style) that I fed it more Voltage than it expected. Took me about 2 seconds to realize that and unplug it.

It survived!!!

Now I have a problem: I've got quite a weird setup for the PC I want to use the X5000 with:

https://i.imgur.com/t3yUI.jpg
large image: https://i.imgur.com/tDusj.jpg

Problem is: I'm using a pico-psu similar to this one: http://www.mini-box.com/s.nl/it.A/id.417/.f (you can see it at the top of the motherboard, plugged into the atx power connector) which is operating close to it's maximum capacity. Using one of the molex plugs coming out of it (this is not shown in the picture, which is older) for the X5000 (sharing it with the harddrive) causes the harddrive to behave erratically, so I'm guessing this is too much for the tiny thing and it's probably dropping Voltage.

So I thought I could connect the X5000 directly to the primary PSU delivering the 12V for the pico psu, assuming the red wire on molex was 12V instead of 5V for some reason (WRONG!)

Would've been quite elegant.

What are the input Voltage specs for the X5000? I assume it's not a good idea to operate it with 12V input voltage, right? (the fan could be adapted)

I'd hate to have to use yet another PSU for providing 5V to the X5000. Any ideas?

molecular, that's a sweet little setup. I think that's a smart idea to bypass the picoPSU, and I'm surprised it doesn't seem to be working. I didn't actually help design the X5000, but I'm 90% sure it's running on 12V. I believe it should accept between 6 and 20V. Are you sure that the primary PSU is outputting 12V? Maybe it's a bit higher? It could just be the fan that doesn't like the voltage. Are you sure you had the voltage going to the right pin?

It's a mystery, but hopefully li_gangyi can shed some light on it soon.

Why mystery? If what you're saying about the input voltage requirements is correct, it all makes perfect sense. Note that the fan is connected directly to the input plug. It's a 5V fan. Does not surprise me its going berserker when fed 12V.

It would be totally awesome if I could put 12V. I'm not sure the fan is really necessary, the cooling block does not seem to get even slightly warm (to my finger, that is, so it's subjective). I could always adjust the fan's voltage using a resistor.

li_gangyi will soon clear up the fog, I assume.

Hi there,

Yeah it's 12V input to the board and 5v to the fan for the X5000. Anywhere between 6V and 19v should be fine.

Without the fan the heatsink does get quite warm, warm enough to warrant the fan.

Thanks, Li, for your answer. *whips out ohm-meter and soldering iron*

I'm pleased to report that my X5000 arrived today!

However... I've followed molecular's guide, I seem to get the board programmed, but when I try to ./mine.sh, I get:
Any suggestions?

I think I never saw this in my test.

Two suggestions:

• try the urJTAG tool. try to load the cable driver and detect the chain (whatever all that means)
• try keeping impact open after programming, then before starting mine.sh, use "Reset Cable" in the menu

Thanks. molecular, for the fast reply!

I changed
to
in miner.sh, and now have 4 shares accepted at bitcoins.lc!

Getting quite a few
Is that a problem?

Thanks

Awesome!


Only if it persists. I sometimes see this on startup, too. Sometimes it can't "unlock the cable", I then use impact "Reset Cable", then it usually works again. If all fails, I replug the USB cable.

If you gain any knowledge about this, let me know.

ok, finally got the x5000 powered by the main PSU (220V->12V) and got around to posting to "Pictures of you mining rigs" thread: https://bitcointalk.org/index.php?topic=7216.820

I'm now feeding only the 12V (black/yellow) of the molex connector and using a (I think 100 Ohm) resistor that I soldered to the molex connector on board to supply (whatever Voltage that boils down to) to the 5V fan. Seems to work... any thoughts?

It did settle down, but still see it occasionally.

I'm not too worried, as the pool is reporting > 100MH/s (114 at the moment, 30 minute average) and Flexible Mining Proxy reports between 93 & 97 MH/s (60 minute average).

I'll test against some other pools later, but for now I'm happy to just sit back & watch the bitcents trickle in!

On another note, this happens to me at times.
I think it's triggered by usb action of other devices (when I plug in/out other stuff). I'm not sure how this comes about. Have to restart mine.sh, then it works again.

Ah, the mystery to me was that the photo in the first post showed the fan connected to 12V, so I couldn't understand why the fan was going nuts. Nevermind!


That resistor could be dissipating a lot a power if you do that, probably almost a watt. Unless it's rated for that, it might suddenly or gradually die on you. A better option might be to take the 5V from the picoPSU, if you have a way to do it.

Not quite right either. It was a total fuckup by me. I put 12V on the 5V part of the molex. That's why it might make the impression that the fan is connected to 12V part of molex if you assume I connected the molex right, which I didn't. So the fan was connected to 12V via the 5V pins of the connector. That's why it went berserker.



I = 12V / 100 Ohm = 0.12A
0.12A * 7V = 0.82 VA

Did I calculate that right?

What happens when the resistor dies? Does it shorcut or does it break the connection?

Taking the 5V from the picu psu would be possible by investing more cable and soldering.

Is there any chance to get more than the 100 MHash/s out of a x5000? Can the changes to the "software" you made for x6*00 be "backported"?

Absolutely! I'm pretty sure the bitstreams are completely compatible. Could you give this one a try? http://fpgamining.com/bitstreams/lx150_125mhz.zip

I'm getting:
:(

Aaah! Tcl! I haven't even looked at that code before, and unfortunately fpgaminer is swamped with other stuff right now...

On the X6500, I'm using the firmware from LX150_makomk_speed_Test (https://github.com/progranism/Open-Source-FPGA-Bitcoin-Miner/tree/master/projects/LX150_makomk_speed_Test), with the clock at to 125 MHz. The software is completely different of course but the FPGA interface is still JTAG, just through an FTDI chip instead of the Xilinx platform cable. I don't know why that bitstream wouldn't work...

same result as thirdlight gets for me.

maybe the "core interface" or whatever changed?

can we use x6*00 software for x5000?

There are at least two different ways that different versions of fpgaminer's Xilinx-targetted code communicate over JTAG, possibly even three by now, and I'm pretty sure that the X6x00 series use a different one to the original X5000 board. You should be able to find a different TCL script that speaks the newer protocol over the Xilinx platform cable somewhere, in projects/LX150_makomk_speed_Test/scripts/ I think after digging up the appropriate PM.

Hi all,

Apologies for taking so long to come up with a response, I've been real busy with work these few months.
I've managed to synthesize a bitstream that should net you 120Mhash. It'll require further testing, but it should work. I'm working with Fizzi to get it hosted on his site, but meanwhile until that is done I could email you guys individually if you'd PM me with your email address.

Here it is: http://fpgamining.com/bitstreams/X5000_120MHash.zip

Also, I received my USB cable for molecular to test out but unfortunately ordered the wrong 2 mm pitch connector. Should get the replacement and ship it out to him soon.

Thanks! Running for about 30 mins - BTCGuild is reporting less than 3% rejected & 110 MH/s.  :)

Awesome!

I programmed the 120mhash/s bitstream, it runs (can't measure hashrate well, using slush).

On a side-note: I previously used a 100 ohm resistor to power the fan using 12V. It was still quite loud and so I tried 470 ohm. That was too much. It was spinning (with a little help) at first, but it stopped at some point and the fpga got really hot (could've been up to 2 days). Luckily, it survived. I now have a switch to use 100 ohm or 200 ohm, since I couldn't find a poti of right resistance.

I'm looking forward to the cable and very much appreciate your going through the hassle for me, fizzisist.

Might be a good idea to just use a lower rated fan instead of trying to use a resistor, it won't work all that well because the fan is not a fixed load, it requires more initial power at turn on to start it spinning.

I know it's not the most desirable solution, but it works well enough. The fan is glued to the heatsink pretty tighlty, too.

I know you're super busy but am curious if we can go a little faster with the ztex-merge?

Hey guys,

most will laugh at this, but I turn off my fpga-miner at night because it's just too noisy. WHAT? you will say, but the x5000 original fan is this tiny screeching trouble-maker:

https://i.imgur.com/wO4wT.jpg (http://imgur.com/wO4wT)

Now, I'm not only jealous towards all the x6500-owners for the hashing-power, but also because of their much better cooling solutions (kudos for the improvements).

So I thought, fuck it, let's cool like the big boys and try this:

https://i.imgur.com/AF6l5.jpg (http://imgur.com/AF6l5)

Notice the yellow cable-binders I used? These also add little feet as a side-effect (sadly there are no holes in x5000 pcb).

As for the cooling-effect, well, it didn't work. The cooling block got hotter than with the old fan, very hot to the touch, after 10 mins.

So I tried this:

https://i.imgur.com/3IeMG.jpg (http://imgur.com/3IeMG)

and this works like a charm. The cooling block is actually "cold" (well, at least not "warm") to the touch, while it was "hot" before with the small factory fan.

it's very silent (not much airflow at all). Now I can mine at night again ^^

That little fan is pretty noisy...

Need moar speed!!

Lets figure out how to get the X5000 working with MPBM

TheSeven wrote this when I asked him about incorporating.  It's a little above my tech level:

"Not by default, but you might be able to hack the x6500 module up sufficiently to make it work with the x5000.
Does the x5000 use the same ping assignment as x6500 FPGA0?
If yes, you should only need to remove the second FPGA discovery and instantiation."

Warning: this topic has not been posted in for at least 120 days.
This has been mining for over a year now  ;D

again, Warning: this topic has not been posted in for at least 120 days.

I'm giving my x5000 to another person for purposes other than mining.

Does any one have one of these units laying around.

I am interested for collectors purposes only :)

I have to look for it, but I have one. Incidentally it has a custom fan duct (made from cardboard). A photo of it was used in a BFL ad in Bitcoin Magazine, showing the photo and words along the lines of: "something like this, just better".

Heh, do you remember which issue?  ;D

haha
what it is ?

not exactly. Should've been around first half of 2013.