Nope, I was watching it for 10 mins before I had to leave. Will check on in later.
After programming the FPGA, did the red LED on the back of the FPGA card turn off? What exactly do you mean by it says it's OFF? |
|
|
|
Anybody know the estimated shipping date?  with BFL's track record probably a year from now =p just think of it this way the last time it took what like 9 months or more(I don't remember) to get their first units out.... and they just had to design and test the pcb/software this time they have to design and test the actual ASIC, pcb and software |
|
|
|
And furthermore, there is exactly one entity outside of Intel itself that is allowed to use Intel's fabs - Achronix. And Intel has stated that they are using less then 1% of the fab capacity.
Actually, for the record, more than one company gained access to Intel fabs so far - at least three: - Netronome - Tabula - Achronix Also BlueSilicon which is owned by Lattice |
|
|
|
Just for the record, Im not entirely pulling numbers out of my arse. I cant tell you everything I know, but I can tell you what google will tell you if you search for SHA256 performance comparisons between 65nm FPGAs and 130nm ASICs, like this paper here: http://eprint.iacr.org/2010/536.pdfOf course, no one said BFL will use 130nm, they might have picked 90nm or who knows, perhaps even 65nm, in which case you could be looking at closer to 5TH per wafer. But even if they went for 180nm or Im wrong by an order or magnitude, the basic premises remains the same: the variable cost per TH is several orders of magnitude lower than the current market price, and if BFL rides the price / difficulty curve from the current top to the eventual bottom, as a miner you will have a serious problem keeping up. It is true. P4man and me are both scared that BFL will corner the mining market ...  I almost had a stroke when I read this thread yesterday. Looking at BFL's past history it looks like they will deliver, unfortunately for all of us running GPUs and FPGAs now. ( not Singles ) Bowing down to the BFL mining overlords ... all this can be stopped by a community effort and mining can be decentralized again. NDA ASIC = not good plan for us and Bitcoin in general. What's stopping BFL having a ride with the blockchain and 51% ? NOTHING. This needs to be fixed. Algo change is in order IMHO. I wish BFL all success in the world but I don't like where this is going for BTC  In fact, ASIC is the only thing that can prevent 51% attack. Why? Today there are banks, financial institutions, intelligence organizations that have super-computers, consisting of thousand and thousands of CPUs or GPUs. These grids are used to run financial analysis, mostly Monte-Carlo. Imagine "Societe General" in France for example. They have a GPU array of 15,000. If for financial interests or certain other reasons, they decided to switch to BitCoin for a week. This grid alone can give them over 15TH in processing power... this will not be 51% attack, it will be 120% attack. This is just one bank, activating only one of its grids. As long as the algorithm used in BitCoin can be effectively implemented in CPU or GPUs, network will always be under threat. Scrypt won't be good either, since many institutions have grids of several thousand CPUs. The only thing that can stop it is the ASIC. Imagine tomorrow the processing power of the BitCoin network is 2,000TH. This is something super-computers cannot attack, simply because their combined processing power cannot reach even 10% of it, let alone 51% attack.... Regards, BF Labs Inc. First off 15,000 GPU != 15TH Where the hell did you get that number? as far as I know there are no GPU's that do 1 Gh/s and most likely they don't have 15,000 high end video cards but lower end video cards. Second whats to stop any entity with money to produce their own ASIC's? if you have enough money I'm sure you could get some made on intels new 22nm tech and they would outperform any Bitcoin ASIC made for the public. Second whats to stop a entity from just buying enough hashing power from BFL to attack the network? Third what are you going to do if the proof of work system changes? ASIC's will be hard coded for the current proof of work system. GPU's and FPGA's can just be reprogrammed to work with the new system. What would you do in that situation completely redesign your ASIC and offer replacements? There are ways that ASIC's can be very bad for bitcoin especially if one company has a monopoly over the whole market and abuses that power (which is illegal in the US where BFL is based) |
|
|
|
Manufacturers of regulator chips usually headline the highest efficiency achieveable which is usually at the smallest input/output difference. It's not unusual for a 12V to 1.2V regulator circuit to be around 75-80% even if the headline says 90%+. Best efficiency is rarely at maximum current either.
Yohan
Yep, very true note how a converter might say 95% efficiency but doing 12v to 1.2v conversion at full load theyre more like 75-85% a lot of manufacturers show graphs or a table of efficiency at different voltages and loads in their data sheets |
|
|
|
|
My thoughts are most people use bitcoin charts as it is a very good and free chart site. Most people arn't going to pay a subscription for something that they can get for free unless you offer something that is better than what they can get for free.
|
|
|
|
Just wondering when is the next units shipping as I ordered a couple of weeks ago and was wondering?
We're still waiting on getting them from the manufacturer. The boards will ship shortly after we receive them. I see you ordered 1 backplane and 1 Spartan 6 daughter board 9 days ago on 5/14/12. Note that we guarantee our boards will be shipped out within 30 days of the purchase date or we will refund 9% of the final cost. |
|
|
|
I don't think this is feasible. The FPGA is actually the part that is most tricky to solder onto such a board (and easiest to mess up), so it seems backward to me to ship kits without FPGAs.
Yeah now that I think about it I would probably include the FPGAs but there would be no price discount on them. A kit like this isnt meant for someone with no experience doing surface mount soldering. Someone who has experience doing BGA reworks and repairs would have no problem soldering these FPGAs. I don't see any problem in allowing people to get discounted boards by assembling them themselves. Of course we would put a big disclaimer warning people that soldering these small parts and BGAs is not easy and should be done by someone with experience and we offer no warranty at all on the kits. |
|
|
|
|
Achronix was founded in 2004 by researchers at Cornell University who wanted to push the performance limits of FPGAs and change this estimated $3bn market, challenging the market leaders Xilinx and Altera.
FPGAs are, as the name suggests, malleable and can be rejiggered to change their basic functions in ways that an ASIC cannot. It might take $30m or $40m to develop an ASIC to do a particular job — say, support the Ethernet or InfiniBand protocols — and if you make a mistake, you cannot erase and go back.
For very high volume products — with hundreds of thousands to millions of units where the cost per unit has to be low — you want an ASIC. But in places where you need a chip that might only require thousands to tens of thousands of units to satisfy an entire market, an FPGA, while more expensive to buy, is better because it is less expensive to make and is correctable in a way that an ASIC is not.
According to Greg Martin, a spokesman for the FPGA maker, Achronix can compete with Xilinx and Altera because it has, at 1.5GHz in its current Speedster1 line, the fastest such chips on the market. And by moving to Intel's 22nm technology, the company could have ramped up the clock speed to 3GHz.
By the way, the goal is to bring the cost of that 1 million LUT FPGA down to around $400 a pop when they start shipping in the fourth quarter of 2011. FPGAs sell for $1,000 and higher today, depending on features.
|
|
|
|
Hey man,
I would hope there would be interest in DIY kits. I mean if I understand what you are offering correctly I think that would be a great idea. I understood it as you would basically put all the parts together and send them in a kit correct? I know it would not be soildered but I think that would be a great idea. Maybe you could make a new post to gauge interest in a project like that. I would prefer a design similar to the one that you had designed. Something that could be added onto when more cash is acquired.
Now would you offer these kits at a lower price if there was enough interest? Let's say 100 kit orders were placed, would you be able to offer it at lets say $500 for the starter back plane, all the chips, and the PCB board? Now I know that you would only be able to include 2-3 FPGA chips at that price but I am looking for an option that allows me to start small and work my way up. I really love the design of the miner you designed for btcfpga.com. I would like to see maybe an 8x or 16x back-plane as well that way I could only buy single daughter board kits for lets say around $150-$175 bucks.
I would like to hear some of the ideas you have concerning a DIY project kit. I mean theoretically if you ordered enough FPGA chips at once I am sure you can get the cost down as well as the PCB back planes and daughter boards. Just would love to your ideas on a kit and the cost factor. If it is something that could compete with the ButterFly Labs single I think there would be a hell of a lot more people buying these kits. I mean I would like to pay around $600 bucks for a complete 4x FPGA setup. Also I think you should include a stencil for the FPGA soldering so that newbies could attempt this with some research. I would hope that they would have enough sense to practice before using the actual chip but that would be on them and not you.
Errr I don't think you realize there is no price discount on the FPGAs even in large QTY.... see for yourself Digikey and Avnet are the only authorized distributors of Xilinx chips... http://search.digikey.com/us/en/products/XC6SLX150-2FGG484C/122-1750-ND/2339792http://search.digikey.com/us/en/products/XC6SLX150-3FGG484C/122-1725-ND/2339800 http://avnetexpress.avnet.com/store/em/EMController/FPGA/Xilinx/XC6SLX150-2FGG484C/_/R-12249118/A-12249118/An-0?action=part&catalogId=500201&langId=-1&storeId=500201&listIndex=-1http://avnetexpress.avnet.com/store/em/EMController/FPGA/Xilinx/XC6SLX150-3FGG484C/_/R-12261582/A-12261582/An-0?action=part&catalogId=500201&langId=-1&storeId=500201&listIndex=-1The only other options are figuring out how to get the chips directly from Xilinx and if I can get price discounts like that I figure there would be a min qty ordering them like that though. Say they required you to buy at least 1000 chips to get the chips straight from them I don't have over $100k to drop on FPGAs so I can get them at discount. This is just a example but I will look into it and see if this is possible. The other option is finding some asian unauthorized dealer that could easily send you fake / old chips that have been remarked to look like the chips you ordered, used chips, broken chips, or just straight rip you off on thousands of dollars worth of chips and not send anything and there wouldn't be anything you could do about it. Anyway what I was planning was maybe making some kits that would include all the parts to make the backplane and daughter boards + the PCBs idk if I would include the FPGAs in with this or not and I'm not sure what the pricing would be yet but definitely cheaper than buying it already soldered and cheaper than buying the parts yourself in low qty. I might make the kits separate also like a backplane kit and the daughter board kit. Oh and a stencil isnt needed the FPGAs come with the solder balls already attached you just have to put a very thin layer of flux on the pcb and get the FPGA perfectly aligned on there and heat it up with your preferred DIY BGA soldering method(toaster oven, skillet, hot air gun, or whatever you can think up that will do the job =p) following Xilinx's soldering guidelines for the Spartan 6 as closely as possible to make the solder balls bond to the PCB. |
|
|
|
It's vcool that your going to be licensing it but will it include customers who receive there boards before eldentyrell new bitstream is released. Or will you just ask for a small fee from them hopefully cheaper than eldentyrell fee due to your license?
Yeah I'm sure we will work something out. I'm not sure how this is going to work as eldentyrell hasn't given out any details of pricing or licensing. I don't even know if he will be offering licensing to the FPGA board makers. It would be stupid not to though.... so I figure he will. |
|
|
|
Looking at "eldentyrell" thread it looks like the MH/J goes down a bit when it's pushed. So the boards with fans it looks like it would to need up to 50A. Reading what he saying is that three Sparten6 at 245MH/s would use 39W so with the boards also powering fans maybe three Sparten'6 per board would be best unless the boards can operate at 58W instead of 45W at 800MH/s with the fans. 58W tho for 980MH/s is still good tho beating the BFL-Single on both hashrate and efficiency. Or with just three Sparten6 modules with fans per board/backplane on the new bitstream at 44W for 735MH/s saving $250 per board or a fourth board for every three you buy. Tho maths could be wrong and "eldentyrell" could change his figures.
"eldentyrell" is going to be selling this bitstream is this FPGA company thinking of licensing it?
First off the fans don't run off the power converter they tie directly to the 12v supply from the molex so we dont have to worry about the power used by those. Using the number he posted 18.4 MH/s per watt at 245 MH/s you would be using 13.31 watts per FPGA or 53.24 watts for the whole board another thing to note is he says this is for the whole board including power loss in the power converters these spartan 6 chips run off of 2 power supplies 1 3.3v and 1 1.25v(actually 3 but VCCAUX and VCCIO can be combined on the 3.3v) so to get the actual wattage used by the 1.25v power converter it would be 53.24 watts - power lost in converting 12v to 1.25v - power lost in converting 12v to 3.3v - power used by the 3.3v supply - power lost from copper trace resistance I'm not sure what the end number would be but I know theres not very much power used from the 3.3v A lot of power is lost in converting 12v to 1.25v I think most power converters will loose 10-20% power doing a 12v to 1.25v conversion at full load so lets just say 15% you come out with 45.9 watts then you still have to subtract the power used by the 3.3v supply I'll just take a wild guess and say 3 watts(may be more may be less idk?) 45.9 - 3 = 42.9watts used by the 1.25v supply the power converter can output 40 Amps and actually more than that but its not recommended. 1.25v @ 40 Amps = 50 watts when 42.9 are needed These numbers really arn't accurate at all though just a rough estimate Another thing eldentyrell stated: I expect the power numbers to get slightly better when I switch from SRL16s to RAM32M's, but otherwise it's probably safe to assume that the current MH/J numbers are an upper bound on the final numbers. Note that the final MH/s is not yet known, so the final power consumption is not known either.
"eldentyrell" is going to be selling this bitstream is this FPGA company thinking of licensing it?
Yes we are. I'm not sure how eldentyrell is going to be selling it.... I sent him a email asking about licensing and power requirements but he wouldn't give up any info, he just told me to wait until later this month when he announces it. I guess we'll just have to wait and see. |
|
|
|
So as long as it's at or under 40A all the Sparten6's should hash at 250MH/s each taking the boards to 1GH/s each. Or if it needs to much power just use three of the Sparten6 modules per board at 750MH/s.
Yep. The power converter can operate over 40A, up to 70A, but care should be taken to keep it cool and operating over 40A would make the lifespan of the power converter shorter |
|
|
|
Will these boards be compatible with "eldentyrell" new 250MH/s bitstream for the Sparten6? If so this will take the boards to 1GH/s will the heatsinks/fans be sufficient?
I can't really give you a answer as eldentyrell hasn't given any numbers on power consumption yet..... However the power converter is made for 40 Amps but will go up to 70 before it enters "Hiccup mode". You could probably run it at 50-60 Amps pretty safely and I really doubt you would need more than that. Going past the 40 Amp mark would probably also require that the power converter gets lots of airflow so it doesn't overheat. As a comparison ZTEX uses 8A per FPGA which would be 32A for 4 This is just a guess but I think the new bitstream will use around 10A - 12A per FPGA A nice thing about this board design is if the board cant supply enough power for 4 FPGA's on the new bitstream you could always just unplug one of the FPGAs. Another option is new backplanes with larger power converters could be made and you could just plug your FPGAs into the new backplane. |
|
|
|
On first look my only concerns are the first one I'm not sure what the exact dimensions are with the fan and without..... A lot of these coolers are unclear about this. you may have to take the fan off and put it beside the board blowing on the heatsink to get more cards in. Oh and no they shouldn't be too small to cool a FPGA Hm, they looks like they only have a small contact area in the middle? That might be a problem...
The first one does look like it has a small contact area in the middle. Should be fine though as long as it covers the whole chip. The second one just looks like theres a square of thermal paste from previous use. Everyones been asking for more pictures I'm sorry I can't provide some real pictures of the REV4 just yet as were still waiting on getting them from the manufacturer.... but I've got some 3D renders..... I'll post up some pics of my REV3 board in a bit also. Theres only a few small changes between REV3 and REV4.      |
|
|
|
The schematics were taken down awhile ago Well I see this project is the design used by btcfpga.com. What am I wondering if the complete plans for this have ever been released. I mean I see so CAD designs of the PCB boards but I cannot submit those to a company to produce them. What I am trying to do is basically build my own FPGA boards with this design because I really like this style of miner. I am really hoping that the original poster will release the designs for his PCB boards so I can get some made. I would really love to try and build my own FPGA miner. I am willing to risk screwing up a little bit as well to learn how to build these efficiently. I know that I am asking a lot but would it be possible to get a complete parts list and the PCB board designs so I can submit them for manufacturing. I really think that I would be able to build my own boards at home. Here is a link to a site showing you how to solider a FPGA at home. http://www.youtube.com/watch?v=OkjOwuSEzKUI think I could follow this man's video pretty well. I would order some stencils as well so that I could make applying the paste and everything very neatly. I mean if we could get a few of us that would like to DIY a FPGA miner to order the PCB boards together I am sure we can get a decent price. I have done research and have found companies that will make me a single PCB board at a time but it will be a premium cost. I am trying to get the cost as low as possible. If I could get the Spartan 6 chips for about $150 each I would like to think that I could build a complete unit for around $700.00 which would be saving me like $400 dollars from the units that can be purchased at BTCFPGA.com. Like I said I simply would like to be able to build my own FPGA miner for a fun hobby/project. I would not be reselling them in anyway. They would be for personal use only. I am hoping the original poster will release the designs to me. I would be willing to pay a little bit for the designs privately if that was possible. Original poster please just PM with the cost details for the PCB board design ready to send to the manufacturer and a complete parts list. The designs were taken down awhile ago... I prototyped this board and had some bugs with it and have worked them out in future revisions. I don't think you realize how much more this stuff costs in low quantity..... you wouldn't be able to make these for $700 a unit in low quantity.... you would need to make over 100 to get to around $700..... for example say you wanted to make 3 units the parts for each backplane would be around $80 -$90 + the price of the PCB's which would be around $180 for 5 then you need 12 spartan 6 plugin boards the parts for those would be around $185-$205 depending on what speed grade of FPGA you get each plus you would need 12 4layer PCB's which are pretty expensive in low qty if you got 12 of them made it would be around $400-$450. So now lets add this all up.... Parts for backplane: $85 ea * 3 = $255 PCBs for backplane 5 for $180 Parts for Spartan 6 plugin board = $185 ea * 12 = $2220 PCBs for 12 Spartan 6 plugin board = $400 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ TOTAL: $3055 PRICE PER UNIT: $1018.33 Oh and it will be more than that you have to include other things you'll need such as taxes on the parts, shipping, solder, soldering wick, flux, a good soldering Iron, and more....... The key to getting it cheap is QTY! I may be able to make a kit or something..... actually the original plan was to just sell kits but I figured most people wouldn't be able to or wouldn't want to solder the 484 ball BGA and all the tiny 0402 capacitors and resistors + the ARM chip that only has 0.008 in. between pins..... I would be happy to make a kit though as long as there is enough interest. |
|
|
|
You should also note: eldentyrell will be releasing his new bitstream for the spartan 6 pretty soon which will give any board that is compatible with it a nice ~50+ mh/s per FPGA performance boost  |
|
|
|
It'd be nice if you could find a heatpipe-type solution so people can put quieter cooling on their rigs.
Also, the URL in your sig is 404.
The main problem in this design for cooling is the amount of space between each fpga card.... You may be able to find a small waterblock something like this: 
So, that mean passive heatsinks wont cool enough? cuz those little fans really REALLY suck, they die nearly instantly LOL...
It all depends on what speed you run the FPGAs.... even with the big passive heatsinks like the Zalman northbridge shown below you still need fans to provide airflow of some sort when running your FPGAs at a high clock speed. with those black heatsinks you see in the pictures, to keep the FPGAs cooled we had to set 2 80mm fans side by side blowing on the FPGAs  |
|
|
|
So, Im curious will the fpga's themselves come w/ the heatsink pictured? Will this be suffecient enough cooling to perform @ 200mh/s / fpga?
Or will I need to go find HSF's for them as well?
No, We've updated the boards to allow for bigger northbridge / older GPU heatsinks, the new design also includes 4 fan headers. We plan on putting a heatsink/fan combo on each fpga that will dissipate around 25 watts of heat which is more than enough cooling for these FPGA's. Heres some examples: http://www.newegg.com/Product/Product.aspx?Item=N82E16835119079http://www.vantecusa.com/en/product/view_detail/93
What do you base that information on? Are you piggybacking on nelisky's x6500 cgminer porting effort? Or who's gonna port that?
The interface is similar to Icarus and BFL, as long I have time I will code it myself... but I may find someone else that already knows cgminers source very well to do it, I'm not sure yet. |
|
|
|
I Like the fact that you are now offering it so that you can upgrade the board and purchase with less on. But where is the option on your web site to purchase the the cards individually once you have bought the back board?
Phil
Not yet I was going to add that today whenever I got the time.... |
|
|
|
|
Show Posts
