Bitcoin Forum

Hello,

As the title says I would like to fund an ASIC mining board and make it available to the community. I believe this will not only strengthen the mining network against potential "ASIC institutional attacks" but will also help it scale faster and better with the increasing transactions number. I intend to funnel income from my main company, a DNA testing laboratory, to fund the ASIC board.

I would like to get volunteers who are good with hardware design to help this project out. So, if you are interested in the design of this board, please let me know.


My background: I have completed a bachelor degree in science (then 1 year out of 2 of a masters degree in electrical engineering), and started a high-tech dna testing laboratory 3 years after graduation. I am setting aside almost all of the profits generated by my company to fund an ASIC mining board. The reason I am saying this is because I want to build credibility to this project.

Please join me in this project if you want to help strenghen the bitcoin mining network and make it the strongest hashing network the world has ever seen!

What would we need todo?

If you can get ArtForz to talk to you, then you just might be able to buy his design off of him. 

Well, I need people to build the design and compile it to a TMSC process http://www.tsmc.com/english/default.htm (http://www.tsmc.com/english/default.htm). I understand gate logic, transistor and so on - since my background is in physics. But I have little experience with the software and the design on a large scale. I need help with the design. Someone to make a schematic in Cadence and compile it with the TMSC libraries, so that it is ready for lithography.

ArtForz has done all of that, and even commissioned a small run of asics for his (not insignificant) mining cluster.  By reports, each card produces 200 Mhashs per second while only using 8 watts.  Most of this is hearsay, of course, but he was discussing ASICs back in September.

I have read everything there is about ArtForz, so I know. But he is not exactly making the ASICs available to the community. This is the reason I am doing it.

Cool idea.  I'd buy one if it were cheap enough... perhaps a kickstarter project would help offset your costs. If you're planning to make it open-source or public domain, there's no disadvantage I can foresee.

Also: is there some reason you chose TSMC right off the bat or was that an example?  I believe ArtForz used a structured asic approach which is significantly cheaper in smaller quantities...

Search for the term FPGA - this has already been discussed at length in there. The FPGA people already have working implementations. The big impediment to ASIC implementation is the cost not the talent on these forums.

TSMC is simply an example.

I think there as been a misunderstanding. If not, then I apologize. But, I do not intend to simply release the design, as an "open source" file. I intend to finance the construction of a batch of ASICs chip based on a design, and sell them at cost-to-produce. So that miners can use them.

Just commenting so I can follow the thread.

I would donate some bitcoins for this and would probably buy them.

Count me in then. I am open for quite some boards.

Don't you dare say FPGA in front of ArtforZ, he will scold you into oblivion.

Since nobody posted a link to this tweet yet - seems relevant: https://twitter.com/#!/MikePFrank/status/78409140699017216

Note that I have no idea how the terms VLSI and ASIC relate to each other, only that both are talking about chips.  :D

does anyone have a rough estimate of cost right now? I though the cost to produce makes it a lot more expensive than GPU mining...

The cost to develop it certainly does. 

Hi ahtremblay. This is my first post in the forum. I'm new to the BC realm, but have been lurking a while.

The idea sounds great. However, do you mean a printed circuit card with the chip, or just the ASIC chip?
Have you any idea on what comparitive cost might be of the card compared to a GPU?   i.e. $/Mhash

Also, what kind of initial target hashrate might be desired? One could always throw more chips on the card to increase the rate... it would be more cost efficient than a chip per card. Or, perhaps, it can be designed so that multiple chips can be installed on one card (by the user)...somewhat like multiple GPU's can be installed on a MoBo.

I have a rudimentary design right now. It is the circuit in gate array format that I designed. It is not a final design by any means, but it allows me to identify the key bottlenecks of the system, calculate gate propagation, maximum clock speed, etc. So what I have from this is; 8MHz clock, 1 hash per clock, and 100 pipelines. This could fit on low volume production run for 27$ / ASIC from TSMC. You then need to add the cost of shipping, putting it on a PCB board and add a controller, + design cost for the lithography. So basically, 800 Megahash/s, for 57$ or so (+ design costs spread across the production line). And that is without the optimization to the main bottleneck (a 7 input adder) that I am working on. So it is encouraging.

You can easily add 4-5 ASIC on each PCB board improve the Hash/$ ratio. So, with 5 ASIC per board, you would get 4GIGAHASH/s for 165$ (+ design cost). (the price of a 5850). Essentialy it makes gpu mining obselite.

This sounds legitimately true from a few random sources speaking about ASIC and possible performance numbers comparing to graphics cards.
It will be interesting what you can manage to produce, sounds very economic. The ASIC stereotype is that it'll provide more speed just because it is chips on a board, why do you think that ASIC is going to work successfully with your plan to fit it to PCI(e) slots.

What about power consumption? Sounds amazing.

I don't see him here :P

This sounds great but my understanding is that those design costs can be quite significant.... 800k to 2 mil USD.  Is that about right or can it be done cheaper? 

Following this thread with interest! :)

It depends on the technology used. There is a rather large one-time, non-recurring expense, and then some price/unit for the actual manufacturing. Of course the more you can pay upfront, the better technology and lower per unit cost you get later. I made some research yesterday (and I was just about to open such a topic :) ), but unfortunately the manufacturers are pretty secretive about their prices.

But it seems to me that the lower end starts somewhere around $100k one-time cost which gets you a structured asic. This will give higher performance than an average fpga, but far from a real asic, and also the unit cost won't be that low. (but much cheaper than an fpga) As far as I know, ArtForz is about to order such a design.

For a real ASIC however you will need to pay a lot more upfront. (a million bucks? couldn't find exact quotes anywhere), but you could absolutely own the network with dirt cheap ICs. The technology is the simplest possible: only digital gates, no analog stuff, you don't even need memory blocks, also no complicated IP is neccessary, everything can be done and tested on FPGAs. The thing could easily run at least 100-200 MHz, even with an older technology (you don't want to pay for a 45nm process), and probably a few full pipelines could fit in. That means several hundres of MHash/sec on each chip, possibly more. The chip can be in a very simple case, with low density pinouts, so it could be soldered very easily, even with DIY methods. But that would take a lot of time with for example a thousand chips :D  If you had the money to manufacture the IC, making the PCB is pocket change. And bamm, you have a lot of panels, each of them with around a hundred chips, each chips putting out 100s of MH/sec. There is a topic somewhere discussing whether the CIA could commit 50% of the full network power. Well, they certainly could.

There is a free fpga design out there already, that's a good starting point. What it lacks is a mean to distribute the work and collect the result, as it currently has some hack relying on the debug features of the fpga and its dev panel. But it wouldn't take really much time to come up with a full design that can be quickly converted to any format the manufacturer needs.

Also, if you want to be safe, you probably want a design that can be somewhat reconfigured, so if the mining project fails, you still have an IC that can be used for other tasks requiring fast SHA-256 computation.

So you would release  the design as open source and also sell them at almost the cost-to-produce?

I would just like to repeat here, that for the sake of safety of our money it would be really nice to build into protocol hashing based not only on sha256, but maybe 2-3 other algorithms. This makes bitcoins much much safer. It could start at block number 300k or something like that.

It would of course make ASICs quite useless (unless somebody produce them for this new version).

Somebody selling hardware dedicated just for mining is inevitable with bitcoin growth, so this is one of the last moments for such a big change (we don't want blockchain splits because of miners that invested a lot in ASICs later on). Such change could make bitcoin solid as a rock. Even broken sha256 would not be a threat.

Screw PCIe ASIC miners, what we really need (and will eventually get, if bitcoin doesn't tank) is a USB miner.  Put your wallet on the USB miner, plug it in to a computer and it starts hashing automatically. 

If I had the time, money or expertise I'd be working on something like this.  Unfortunately I have none of the above - I'm merely an ideas man.   :P

spend up to millions of dollars to fit one asic chip on a usb stick (and get limited hashing and zero scaling ability)  vs. build a PCIe card or array that can fit tons of chips and scale  hugely)?  

I would go with the latter.

I've only seen ahtremblay talk about PCB, i.e. circuit boards. Frankly, fitting these on pc slot-in cards would be daft. External units connected by usb, or preferrably ethernet is the only way to go..

USB hubs vs. PCIe splitters... just look up the prices!

Ethernet or USB (2.0 or 3.0) would be the way to go for sure (with Ethernet, you could do PoE and USB also allows electricity transport...)

Anyways, as long as I just read about "I designed this and that and it does xxx MH/s" and noone releases even an FPGA implementation of that, I personally don't really trust it.
Doing all of Bitcoins math in just 1 clock cycle for example takes quite a lot of gates, hm?

I apologize for locking this thread by accident. It is now unlocked.

Absolutely impossible.

Maybe, maybe not.  Those who wanted to mine on GPU's didn't think to highly of the claims of a single ATI 5830 getting 100 times the hash rate that a high end cpu could get either.  I'm going to reserve judgement until I see some physical hardware.

Hello, care to elaborate a little more?

I have developed in Xilinx ISE and from my experience that's way off. How many gates do you have? To have 1 hash/clock it must be insanely big, and you have to multiply it by 1000 (as you said). Hoe did you guess the production cost?

Even without knowing anything about algorithms Bitcoins and other stuff, the fact is: Commercial solutions are slower and more expensive, how a person with no experience make something better, faster and cheaper?

The performance numbers sound reasonable. I have no idea why he only gets 8MHz, but this sounds like it can be improved. For a full pipeline you need roughly 100K flipflops plus all the logic. 100 times that would be 10M flipflops plus logic, which would be a rather big ASIC, but it's certainly possibly.

I had a co-worker try an Altera HardCopy synthesis and that one ended up with 20 pipelines running at >200MHz each, yielding >4GH/s per chip.

I don't trust these cost estimates at all though. The ASIC price estimate might be sensible for >10000 volumes, but the board, host interface and assembly cost estimated at $30 seems to be impossible. The VRM alone will cost that much.

I never said I had not experience. In fact, I said I am a physicist with 1 year of masters in electrical engineering. I said I understand, transistor, gate logic and most if the solid state physics stuff related to electronics, since thats what I did for a few years at the university. I said, I had no experience design on a massive scale using current software and compiling the data to the TSMC format. Thats a lot different than no experience or no knowledge of the issue. The reason I know the price per units is because I have a quote from TSMC.

Anyways, I have hired a firm to help out with the design from start to finish. So things will go smoothly now.

I think you might be assuming too much about his level of experience and knowledge on the matter, lets not rush to conclusions.

I'll defiantly be watching this thread, but in a way it would be kind of a shame to see mining move away from the GPU and onto dedicated hardware, it would mean the end of the garage miner.  It seems like the people who would be most interested in buying these would be small companies and start ups, only the die-hard miners would buy hardware that could only be used for mining bitcoins, even if they were cheaper than GPUs.

But on the other hand, I wouldn't want to see this type of thing only be available to a select few either. I'm kind of torn, but honestly I would probably buy one.  

Good choice given the nature of the thread thus far, I wish you the best with this and think you may yield some good results depending on the firm you hired.

bookmarked

Commercial bitcoin mining solutions? I didn't know there was a solution built specifically for bitcoin mining already instead of an adaptation of generally available commerical products built for differing reasons. Please enlighten us..

EDIT: I misread the initial quote however I'll leave my reply as is because the general point is the same, nothing has been purpose built that is generally available to the community, in that sense it could be done.

I have a background in hardware driven SHA-* hashing for different reasons and there is a fair amount of good research available online as well as good performance analysis of different approaches, should be an interesting project if the details are released to the community (vs. Art's close hold)

I think an open source hardware design is a great idea. The FPGA design work that is going on is just awesome. Provides a level playing field for anyone who wants to get in to the custom hardware game. At the end of the day, the innovation should not be in mining, but in the software and services to make Bitcoin more readily accessible and integrated with other ecommerce systems.

As least USB 2.0 is limited to 2.5 Watts (5V, 500mA) I think USB 3.0 may have a high current mode, but am not sure.

Power over Ethernet is limited to about 15 watts (48V, 350mA).

In the FPGA thread they are talking 8 watts per chip. While less than the ~200Watts drawn by the GPU miners, it is still significant enough to kill the scalability you cite.

He can't get over due to hazards . The propagation delay will be way too big to get any higher frequency.

You're not seeing the big picture (or at least a different picture to the one I see).

Imagine a world where btc is the default currency used in online transactions.  Mining difficulty is such that the idea of an individual building a dedicated miner is borderline lunacy - the btc made by mining is pretty much exactly equal to the electricity cost of doing so.

So who mines, given that mining is necessary for the integrity of the system?  The answer - everybody.  Your btc wallet is a keychain usb stick with a small btc client that mines whenever powered and internet connected.  Maybe only gets a few hundred MHash/sec (we're talking dedicated custom built hardware), but there are hundreds of millions of these things around the world.  We're talking Peta-hashes now.

And who pays for the portable btc wallet/miner hardware/development?  The pool owners, that's who.  Amazon gives you a free btc wallet with every purchase.  Google give you one free if you sign up for adwords.  Buy a new Toyota and it comes with a btc wallet (and a USB plug to stick it in your car, so you can pay for fuel with btc at the pump).  And the pool owners in general keep the proceeds of mining (we're probably talking transaction fee only by this stage).  Or maybe they give you a discount on your next purchase, or something like a frequent flyer system.

Anyway, that's one vision of a possible future with btc.  But one that I think is plausible if btc "takes off" and lives up to its potential.

This can't be true if we manage to get >100MHz even on FPGAs. 8MHz sounds like the algorithm isn't pipelined at all, and calculates a hash in a single clock cycle, which is a waste, as latency doesn't matter here. What matters is throughput, and if you can get more than 20 times the throughput with just a couple of additional flipflops, you shouldn't bother about 128 clocks of latency.

Just posting to follow this thread

seem interesting. :D

My issue with all of this is, that the hadware spec won't be Open Source, only the finished ("compiled") chips should be sold.

Sorry, but this approach creates a centralized entity again. If it's really so hard and expensive anyways to produce ASICs, you could open source the hardware plan anyways, as there won't be someone else to copy it who has not a few million bucks to play with.

I won't develop (or help developing) anything that in the end will only output a centralized miner chip that noone can audit (who says this design I implement doesn't use 10% of it's power to submit my private shares?) or improve. It is my believe that as much about Bitcoin as possible should be open source - this includes miners in software AND hardware!

Is this a delusion here?

Chip production is not something that scales down. Want open source - sure. Pay for tapeout yourself. I think you overlook three really critical poitns here:

* One, multiple such designs can coexist.
* Second, this is economy of scale. Startup costs for production are HIGH, so you neeed tp produce a LOT to make it financially viable. Get the design and be surprised by 50.000 USD costs just to get the tape out - not having product a chip, plus the factory then calling for a minimum run. Not having enough money? Go buy chips.
* Third, the chip can not sumit anything (10% private shares - nice joke), it still has to go through an open source DRIVER. hich sort of makes sure it gets properly used.

But really, anyone asking for not selling the chips is delusional - I think not a sinbgle miner here can seriously finance a fully production run for special chips sensibly.

I, personally, would hbe happy to buy special boards that contain q lot of lower wattage processing power.

Yeah, open source the VHDL code, or even the die layout. And you still can't verify that the chips you would buy can't have been tampered with.
The only way to verify that would be building the chips yourself - which you can't do for obvious reasons.

Do you build your own GPUs and graphics cards today? You certainly don't. And a graphics card (oh, and it has a closed source firmware/driver!) could tamper with the data more easily than a pure sha256 bruteforcing accelerator ASIC (like it's being planned here) could.
The worst that the ASIC could do is not finding some shares that would have been valid. This can easily be detected, and the only way in which the vendor could profit from that is a slightly lower difficulty.

This of course assumes that the network interface isn't on the ASIC itself (which doesn't make sense for cost reasons anyway), but instead running on some PC (for PCIe accelerator cards) or on some ARM SoC with open source firmware (for standalone mining boards).

How about Thunderbolt?

Apple's usage is limited to 10W on laptops, but since it is based on PCIE (it IS pcie)) it could deliver 75W on desktops.

http://www.intel.com/technology/io/thunderbolt/index.htm

Following.

I just want to say that I find this whole ASIC-Thing very interesting and a great engagement. If I'd be sure that I could get those babies running, I would throw say a thousand bucks in the ring.

As for high initial costs we could open a website which tracks the spendings from each member to the project and then calculate the board price as "sum of expenses / boards delivered".
Then at first only a small amount could be produced (for say 1500$ per unit). Later when a second (equally large) order needs to be placed in china, resulting in a price of 500$ because no more design etc has to be done, the overall price would be 1000$ bucks, which every board would cost then. then the members who payed 1500$ initially would have like 500$ cash-back or discount on further boards. they would practically a board for free.
As I stated above both orders were equally large, so you notice that this calculation makes no sense. but you get the idea, hmm? ;)

I still would think the very best Idea would be to open source the layout and design in the first place.
If you want to produce them as well, go on, people who are not able to produce their own will buy yours.
So why don't just go both ways?
Greets, M

well, is it really a call for investors?
There would need to be some hardware guys with a reputation or an escrow service.
they see cash, they start to develop.
asic is done, op tells it is done. people set up a pledge if it is released, he releases it and receives the pledge.
sounds like a neat idea, doesn't it?
greets, m

Casper,

Where have I asked for money? I asked for help developing the software by contributing free time to the project. Granted, 1 forum member has privately expressed interest to invest and I have started talks with this person. I said, I will finance this project. If anything, I should be the one worried about getting scammed. I received a bunch of emails from people will little to no experience or knowledge offering help for a fee. If anything, I should be worried that their design won't work, making me waste my money.

If I say that I will finance the design of an ASIC board and release the design for free. As well as order a batch based on the design and also release that affordable. Of course people will say its a good idea... since there is no disadvantages or risk to anyone. Why would anyone say its a bad idea.

The software looks like the most simple part of such a project, IMO.

You are a conspiracy theorist nut.  My profile has an email, I just choose not to show it publicly.  I don't have a website of my own.  You want my TFN? (That's tax file number, closest Aussie equivalent to an SSN I believe).

Tool.

Casper, great, you've made your point. Now please settle down and prepare to either accept your kudos for being right or to give your apologies if you are wrong.

This can very well be interpreted as "not just open source, but also affordable physical chips, as doing that on your own isn't affordable", and if this is the intention here, that's perfectly fine.

He's basically taking all the financial risk, and only asks for some help on the design/software side. The only scam I could imagine is that he builds the chips with our help and then decides to use them all for his own mining instead of selling them at affordable prices as promised.

To get back to the technical aspects:
Why do you only get 8MHz out of that circuit? How is it constructed? All the SHA256 rounds in one huge tree of combinational logic? Or is it pipelined?
Do I understand correctly that each SHA256 unit on it gets a throughput of one hash per clock? How deep is the pipeline, i.e. how many clocks is the delay between input data arrival and result availability?

Throw that on GLBSE!!!! PLEASE!

I would gadly invest on that, even if the return is 0 :D

The hardware effort is already in the open source realm: http://forum.bitcoin.org/index.php?topic=9047.0;all
Maybe it will be required to make extra hardware mechanism to ensure integrity of the deployed computational power.

Once the chip exists, it only a matter to make it available to hardware developers (e.g.: make it buyable from digikey.com) that will be able to cook up hardware around the chip. Right now all the money is going to AMD...I think it would be more decentralized if we could all build our own hardware.

I say let's make a project on www.kickstarter.com (in the technology/open hardware section) to get as much cash as possible and let's build the damn thing. Projects on kickstarter have gathered as much as 300-400k$ in the past...so it's possible to put some real money togther. People could fund the project by pre-ordering their board. Funds should go to and be managed by bitcoin founders/bitcoin.org...the chip's IP too.

We better get our skates on as a community before someone else does it.  Somebody somewhere will be working on a GPU killer, and keeping it for themselves.

This, times 10.

I have posted this elsewhere in the forum, but perhaps because I'm a new user, people haven't noticed. I'm actively working with a group of experienced product managers and ASIC engineers to build a Bitcoin ASIC, and expect first units to arrive in September or October. This is a well financed project, and we will likely eclipse what the community can do through Kick Starter. This is not to say that the community should not build its own ASIC - I just don't want people to be pouring their life savings into Kick Starter only to be surprised that some startup has surpassed them.

We know that our product will be valueless unless we support the Bitcoin community and encourage the kind of grass roots engineering that has taken the system so far in such a short time. To this end, we are planning to make as much of the design open as we possibly can; for example, the device drivers will be completely open source to encourage innovation. For obvious competitive reasons, the digital design will remain proprietary.

I don't want to say more about the product at this time; however, if you are interested in getting on our announcement list or if you have suggestions for us, please visit the form at the following address:

http://www.largecoin.com

Thanks,
The LargeCoin Team

i'm skipping the gpu fad. never bought a computer that came with a video card. dont mind buying a custom fgpa or asic.
anyone not accounting for moores law is retarded.
also, THANK YOU Mr. Tremblay, this community needs to work like a well oiled machine if bitcoin is to succeed.
also, WHEN CAN MY 3D PRINTER CREATE ASIC FISHES N CHIPS?

Could you elaborate on some questions, which I believe are not interfering with your competetive reasons, but might be helpful for the community to prepare?
- Will the ASIC itself be sold by your company (or e.g. through digikey)?
- Will ready-to-use boards containing your ASIC be sold by as well, or should the community handle that?
- Have you already decided on the ASICs external interface?
- Are you planning to sell PCIe mining boards based on your ASIC?
- If yes, will the ASIC itself have a PCIe interface, or will the ASICs have a more simple interface, with multiple ASICs connecting to some controller that handles the PCIe interface? (I would strongly encourage the latter)
- Are you planning to sell standalone mining boards, equipped with ethernet and some small CPU running linux, doing the highlevel work? (Not having a PCIe interface for the individual ASICs would pay off here, I'd really like to have an ARM-based backplane connecting to like 8 miner boards containing 8 or 16 ASICs each, even if that thing might cost a fair bit of money.)
- Will there be a public datasheet for the device, which contains physical/electrical/thermal specifications along with a specification of the software interface (I2C register map or something)? (I consider this way more useful than e.g. an open source linux driver. This is work that the community can do for you if you provide the specs.)

Subscribe. Sorry for not contributing anything, but I haven't a clue about hardware stuff (my geekiness stops at the kernel).

Wow.
You sound exactly like one of those middle-aged drill sergeants from Vietnam.
Sure you weren't in the marines?

Do you go online with a large pocket calculator and 28k dial-up?

Sent you a few bitcoins.

We haven't determined this yet, but I would argue it's most likely that we'll be selling complete mining units rather than individual processors. There are applications for Bitcoin technology outside of the public Bitcoin network, and we want to address these opportunities, which don't have an active hobbyist community behind them.


I can't comment on this yet, except to say that you'll be really delighted by the interface choice. It will permit very easy management of a mining cluster with "zero configuration".


Documentation will be provided - as I said in my post, we really want the community to get involved in building tools and services around the hardware.

It does of course make sense to sell ready-to-use mining units in one or another way. However, I'd expect that the community would build some alternative mining platforms based on your ASIC if you sell the raw chip as well. As you're probably making the big money with the chip and not the board, this might actually increase revenue, as it will probably increase the total number of ASICs sold. (Some enthusiasts would certainly build their own mining rigs based on your chip, and possibly sell those as well.)


If I interpret your answers from above correctly, you're focusing on a standalone mining solution, and not PCIe accelerator cards. I think that's the long-term way to go. This means that the external interface of it is most likely ethernet, with possibly some additional daisy chaining interface.
Oh, and as I said above, if you focus on this, just sell the raw ASICs as well, and let the community care about how to do a PCIe card solution if they want that.


That's nice to hear :)
Don't make the same mistakes as certain GPU vendors ;)

I'd just like to have the complete design open source, i.e. that I can produce my own chips with enough money.
I think this is what we will have in the end and certainly need as well to keep a dezentralized currency.
Greets, M

Cool - glad you like the idea. I can see why the GPU vendors guard their drivers, because there is a lot of intelligence in the driver. In our case, the management software and drivers are not giving away much. Just making an ASIC requires a huge NRE investment; and of course we know that the guts of the chip are not rocket science, save for a few careful optimizations. By contrast, the guts of a GPU are rocket science.

I wanted to share some of the results of the above-mentioned survey link. The following pie chart breaks down (in real-time -- thanks, Google Docs!) the community's feedback regarding what would cause them to move from GPU mining to ASIC mining.
https://spreadsheets0.google.com/a/ttul.org/spreadsheet/oimg?key=0Ap9K7zsrtYpcdEkyNDFTQkZHWktoOVNkMXlZWDdHUlE&oid=2&zx=xz0n45ktocp6

I am quite sure that the community is going to provide this regardless of whether our commercialization effort is a success, and having multiple options to choose from will only strengthen Bitcoin. A few things to keep in mind for the community effort, based on what we have learned from our many years of commercializing ASIC designs:

1. The NRE (non-recurring engineering costs) are truly stunning for ASIC production, unless you use simple techniques like a gate-array, which have performance and density drawbacks;

2. ASICs are a "wild west" business, and a great deal of haggling and negotiating is required to get a reasonable price out of the various service providers you'll need to work with to get chips made; and,

3. There are numerous physical pitfalls that can render a design completely useless, introducing delays and increased costs - for example, you may find that you've made a tiny error that can't be corrected in software.

So what's going on? Can I already buy some board? ;)

Watching this thread.

Thanks for letting us know, guys, but please use the notify button next time.

the notify button, also sends you a mail :(

Change your email settings to something it's not.  Try "null".  What other reason would you have your real email address registered at this forum, if not to receive notices?

So is this actually happening? Or was it just noise? Or what?

-MarkM-

I for one am sure I'll be buying the first aisc that the community produces capable of this. Although I'm somewhat amazed no aisc for something like sha hashing it's available already. Then again cpus are probably enough for servers that encrypt their data using it... Out of curiosity has anyone thought about what chips the self encrypting hard drives use?

Highly interested ;D ;D ;D

An ASIC would be interesting, provided it does not cost a lot. Also, selling the ASICs would be good business, since if the ASIC outperforms GPU (which it will, most likely), everyone will start using them and those who still use GPUs will fall behind, since the difficulty will increase, much in the same way as with GPU vs CPU mining in the past.

If the cost is comparable to a GPU (or at least a PC with the GPU), including shipping to Lithuania, I'll most likely buy some. It's always fun to play with electronics.

I've done the back-of-the-envelope math on this. It is doable. The problem is the up front cost is around $2,000,000 to do it right.

The target ASIC would use the fully-unrolled design that's already used by the FPGA miner. The design goal would be to fit 4 fully-unrolled miners on a chip. The target frequency would be around 350Mhz, one double-hash per miner per clock. That would mean 1.4GHash/s per chip if all goals are met. The chip would draw around 50W. You could put four on a card and connect it to a USB port.

The cost per ASIC would be under $100 plus a share of the development costs. So it all comes down to how many chips you can make/sell. That would give you a 5.6GHash/s mining peripheral for around $650 including baseboard, cooling, power supply, and so on, but not including the cost of development.

To have it make business sense to borrow money to fund the development (assuming you could even arrange it), you'd have to be able to sell the miners for $1,500 each or so and you'd have to be able to sell 5,000 units. I think people would pay $1,500 for 5.6GHash/s at 250W, but I'm not sure 5,000 units would sell, especially as early sales raise the difficulty. (And if that happened, the total network hashing power would jump to three times what it is now!)

It would probably be useful to make an entry-level unit with just one chip on the card, as the lower you keep the cost of entry the more sales you will probably be able to make.

-MarkM-

I doubt ;)

The main problem will be price - comared to the 7xxx AMD series.

But if htaht is ok, I would take some. Currently setting up a not too small operation.

Has anyone considered the engineering may be partly done I've seen a few things like these with hardware sha units.

They're designed for a different purpose. Here, the goal is not to do a SHA as quickly as possible or over a large amount of data but to perform a particular operation requiring two SHAs on a small amount of data as many times as possible concurrently. The FPGA miner already has the basic design -- it can do one double hash per clock cycle (per instance). It's just a question of how high the clock speed can be and how many of them can fit on an single ASIC. (I'm oversimplifying a bit, but the general point is sound.)

The current design uses an assembly line approach. While SHA step 5 is being done on none 12, SHA step 4 is being done on nonce 13, and SHA step 3 is being done on nonce 14. Nonces come in the front from a counter and once the pipeline is full, complete once per clock cycle. The actual SHA logical operations would be wires on the chip. Each SHA XOR would be an XOR gate. I don't think you can do better than that.

Posting so I can follow this thread...

JoelKatz, you seem to know a lot on this subject.  If I understand you correctly a $2 mill investment would resulted in a card with 4 ASICs 1.4GHash/s per chip at 50 W each, for a total 5.6 GHash/s 250 W  USB plug unit. Production cost would be $650 each. Would you care to comment on why this would cost up to $2 mill to do it right.

Did anyone do the math if this is done and sold for $1500 each if the math holds up and this would be a good investment for miners? Problem I instantly see is the difference between fist 1000 units sold  versus the last 1000 units sold, because in the meantime difficulty factor would double if not triple, while the product would still sell for the same price.

Correct. To do it right would require designing a fully-custom ASIC and having it manufactured on a high-end fabrication line capable of producing chips with hundreds of millions of transistors running at 350MHz. Every step in that process, and there are many of them, costs money. (1/4 of the cost is salary, by the way.)

Essentially, an ASIC would mean a chip produced in almost precisely the same manner as the way a high-end GPU was produced four years ago, except it would be designed for bitcoin mining from the ground up.

The numbers are solid, the problem is the uncertainty. Also, every unit you sell is weakening your market.

@JoelKatz

5.6ghash at 250 watts definitely makes business sense for miners as well at $1500.
I'd replace all my rigs, stop paying datacenter rent and run those at home.

But can a USB port (even a reinforced one) actually supply 50 watts let alone 250w? My biggest doubt.

Irrelevant. YOu can always put in a power supply. Actually no, USB can not power 50 watt.

That said, I would NOT stop my datacenter operation plans - I would rpelace cards with more asics ;)

If I understand him correctly, this card would have its own power supply, USB port would only be used for PC connection.

The USB port is strictly for work units to the ASICs and shares back to the PC. It would need its own power supply, cooling, and so on. The bandwidth needed to and from the unit is negligible. Essentially, it would need a stream of work units of 8KB every 12 seconds from the PC and return a share, about 300 bytes, about every second. A serial port would work as well. (Sorry, out of time. Numbers not double-checked. Hope I made no mistakes.)

Since you are counting $400 for 4 chips and $150 for the rest (for a total of $650) would it be posible to include a chipset that allowed a ethernet connection and a very simple webserver (like the chipset in a home router) instead of the usb chipset. This way you could have an independent device just connected to the ethernet network and configured by a simple web interface. Im not very informed about the prices but I dont think it would push the cost too much and it would avoid the need for a computer (saving energy and money). The question I guess is if that simple chipset could handle the mining software load.

It could be done. The chipset would just need to be able to talk to a mining controller, which could be the bitcoin client for solo mining. The work that needs to be done beyond the hashing is minimal. The equivalent of a $40 wireless router would do it.

Stack a bunch of these into a baseboard heater fan case, and allow me to control it with an external thermostat, and I would buy it today.

I've never really mined, other than 0.01 btc off an old GPU that took me 24 hours to generate.

But even I'd be interested in mining if I could buy 56 GHash/s for $15k.

Just need to figure out a way to direct the resulting heat into my swimming pool.

What about financing it through shares? I mean, creating a sort of a cooperative that pays dividends in cards, so people pay the money and at the end when it is developed they get payed in cards. That way you are making sure you will get a number of sales before starting the project. Sort of like testing the waters to know if the demand is there.

What about professional ;)

Willing to commit to a 10.000 USD per month purchase for 6 months, providing the conomics is better than the 7xxx graphcis card looks like. Bank guarantee can be provided ;)

Here is my idea.   Read it twice :)  It might still be somewhat messed up, but you will get the idea. I will also round up the numbers and explain why. If we go by JoelKatz numbers project upfront cost $2 mill, each unit would cost $650 to produce.

I will write on the example of 1000 investors. Each investor would invest $2500 for upfront cost and 5x$700  to receive 5 units of the product when it would be produced. Each of the 1000 investors would invest $6000 and receive 5 units of 5.x Gh/s at 250 W, so approx 25 Gh/s at 1250 W for $6000. This way we would have $2.5mill cash to develop and presold 5000 units at $700 each which is around production cost plus some spare.

When we would ship this 5000 units to initial investors, we would start taking pre-orders for a next batch of at least 1000 units at $2000 each for public / investors. There would be a time delay of at least  4 weeks between each batch of 1000 units. When this pre-order fills all 1000 units are shipped to their owners. Then we start taking pre-orders for at least 1000 units at $1750 each and repeat the process. Then at least 1000 units at $1500. We would stop lowering the price at $1000 per unit unless something drastically would change, like a new competitor or whatever.

*If there would be more than 1000 orders we would fill all before start taking pre-orders for next batch at a lower price.

The time delay between each shipment, would need to be calculated more precisely ( I used an estimate ),  is for the difficulty factor changes. It needs to be calculated in the way that the guys who would buy $2000 unit  would have time to earn enough coins compared to the next batch units that would sell for $1750 and so on.


Where would all the profit go, if there would ever be any?   

Back to original 1000 investors. Company would buy bitcoins on the open market with all the profits and transfer it to the 1000 investors every quarter. Keep in mind that this example is for 1000 initial investors for $6000 each. If there would be more interest we would redo the math. If for example there would be 1500 investors, we would do $1500+ 4x$700. There has to be a minimum limit of at least 1000 investors so everyone could invest plus this would keep majority of the hashing power spread out. We don't want to get a small group to run ASICs while the rest would still use GPUs and would have to stop mining soon after.

Each investor would have 10000 shares in the company, which they could later on buy/sell on the open market for bitcoins, so everyone could be a part of this company. We would develop and add this exchange later on. This way we would get ASIC and everyone could get in on the action. Initial 5000 units would add 25 Th to the network, then each 1000 batch would add approx 5+ Th.



This is just a nutshell, community would work out every possible detail so it would be best for everyone involved. There is a lot more details I already know, but I don't want to write an essay here.

Sounds fine. The problem is finding 1000 investors. Does not sound like an easy task.

I would invest if A) it were feasible
                        B) the numbers were more concrete e.g cost defined

Same here, funds are at the ready :-)

Is it the money value that you think is the problem or sheer number of investors?

I think if the whole community can't come up with at least 1000 potential investors each risking $6000, then I think the whole bitcoin project isn't ready for ASIC yet. It needs more time for its natural growth.

We need to be careful here, I will give you an extreme example:

if only 50 people would invest each $100,000 they could put the whole bitcoin project in danger. Not in a way that they could ruin the whole thing, but just takeover mining and collect 40% of all coins in each round, then reinvest and keep growing, basically to the point where this small group of 50 people would hold 90+% of the market. By that time majority of the GPU miners would stop already, which could put the whole trust of the network into hands of 50 individuals, each only putting up $100,000 at the begining.



I understand it might not be an easy task, but we won't know unless we come up with detailed plans and see what community thinks. Right now we are only talking approx costs of the whole thing to get a frame work done.

I dont know if we will be able to find the investors. Maybe is easier than I think, who knows. The obvious move now is that someone capable and with some respect in the Bitcoin community ( **cough** JoelKatz **cough** ) picks up the project, writes down a viable plan both from the economical and engineering point of view and then throws the idea out to see if the investors are there or not.

Extending these numbers, 5000 units at 5.6GH/s each is a total of 28,000 GH/s
Given the current non-ASIC hash rate of about 11,000 GH/s, the new hardware would constitute 72% of the network
Assuming block generation levels out at 144 blocks/day and 50 BTC per block, that's 7200 BTC per day, 72% of which is 5169 BTC per day.
Divided among 5000 modules, each module would earn approx 1BTC per day.
At (say) $10 per BTC, that's $72,000 per day for the entire network
72% of that for the ASICS would return about $51,700 daily.
5000 modules at $650 each is $3.25 M, plus NRE of $2 M is $5.25 M.  It would take 101 days at $51,700 to recover the initial investment.

There is diminishing returns in producing more modules, as each additional module adds fewer BTC to the ASIC total.  According to my estimates, the highest return on investment occurs at about 2500 modules, paying off the initial investment in about 90 days instead of 101.

If a massive wave of zombies with GPUs drove up the total hash rate, that could hurt the return, but it would require quite a zombie army.  Or if BTC were to drop precipitously, that would obviously be a risk.  Owning more than half the network also hurts the perceived security, which could in turn hurt the value, which could be a reason to sell the hardware to diversify the ownership.  It would also be good to plan for, or at least consider when the coinbase drops to 25.

But aside from these risks, the idea seems sound.

I'm not an ASIC expert. I can manage engineering projects, design complete systems, write software, and design hardware above the chip level. I have some FPGA/PLD experience, but that was more than a decade ago. Someone with real ASIC-specific expertise would need to make a detailed plan and firm up my numbers (remember, these were back of the envelope calcuations).

I can tell just from my own expertise that if the ASIC meets its design goals, the rest of the system is not particularly difficult. Generating and moving the work units and collecting the shares is not difficult at all. Providing power and cooling is simple.

Also, I can't take on significant time commitments that don't provide certain revenue because I've already taken on more of them than I should have.

I'm probably in about the same boat as JoelKatz. I would not call myself an ASIC expert, but I've been involved in ASIC design projects for nearly 20 years, so I have a reasonable feel for the time and money it would take to get the job done in the context of a small engineering startup. I agree that it's about a $2 million, 1-year job to get one to production, using the open source FPGA miner as a starting point.

So, here's the thing. I have the technical expertise to participate in such a project, and I'd even enjoy it. However, I presently have a good job at a huge semiconductor company, with good compensation and a good expectation of stability over the next few years. If I were to pursue a Bitcoin mining ASIC project, then life might be good for up to a year, and then I'd find myself out of work. I have a mortgage to pay, and my bank doesn't take Bitcoins yet. And I don't have a nest egg stashed away that would let me do a one-time job like this and then be out of work for a while as I look for the next job. I'm interested in Bitcoin and would like to see it succeed, but I don't think it's in a state yet where I'd bank my career on it.

There may be good ASIC engineers out there who are in a much better financial/career position than me to take on this sort of project. If there's also somebody with a couple million bucks they'd like to front on this, then GPUs might go the way of the CPU for Bitcoin mining by this time next year!

Regarding USB vs. Ethernet as the interface for these proposed ASICs, there are lots of options for embedded microcontrollers that would be suitable for managing a herd of ASIC miners. A little ARM processor, running Linux, with an Ethernet interface would be quite do-able. The end product could be a standalone appliance that just needs to be plugged into power and an Ethernet jack, configured to log in to your favorite pool, and then left alone. There are even all sorts of off the shelf embedded microcontroller boards that could be used for prototyping.

Fair enough.

Thanks for clearing up some of the math behind the feasibility - what do you think we as a mining community can do to prevent some company from rolling in and taking over? Should we try to stop that kind of behavior?

Maybe changing the bitcoin code? Developing our own ASIC? Distributing the bitcoins amongst a larger amount of people?

I definitely don't know what the answer is, but I've been getting the feeling that the invisible hand actually belongs to one or several people - and not the market (as it should).

It looks that way, especially with the cost of the hardware. What I mean is anybody can buy a decent video card, they are not that expensive. Also, they are useful for games, video encoding etc, so if mining does is not profitable, I can put the card in my main PC and play games on better graphic settings. It would not be a total loss (even though I would not have bought the card just for gaming). Even if I burn out the card trying to mine it's still ~$180, not $1000.

OTOH, the ASIC will cost at least $1000 (more than any one ATI video card) and while it will outperform any currently existing card, it it useful only for mining. If mining is not profitable, the $1000 is a total loss. And, $1000 is a lot of money. Of course, those who can afford to gamble $1k (or $2k) on this will make mining with a GPU not worth it (just like it is now with CPU mining), so the miners who have to pay for electricity will drop, leaving the majority of the hashing power in the hands of a few people who could pay a lot of cash up front for the devices.

I don't think that Bitcoin will be that distributed anymore, also, if the ASICs are made so that they can also forge transactions (for the manufacturer) it would be a big problem.

I think it is now evident that the mining algorithm probably should have been memory hard, not just processing hard. That is, the mining algorithm should have been designed to require at least 128MB of RAM. That would have made CPU and GPU mining practical and on a roughly even footing, and it would have made ASIC/FPGA mining impractical.

See 'scrypt' for an example of a memory hard hashing algorithm. It was designed to resist exactly this kind of 'attack'.
http://www.tarsnap.com/scrypt/scrypt-slides.pdf

I'll just leave this here

http://news.slashdot.org/story/11/07/11/2232253/JPMorgan-Rolls-Out-FPGA-Supercomputer (http://news.slashdot.org/story/11/07/11/2232253/JPMorgan-Rolls-Out-FPGA-Supercomputer)

Ctrl-F
Bitcoin

...there's always someone.

Personally I don't think I will join, at least not for now.

But I would advise you to wait for the HD 7000 generation gets released so you can measure what exactly you are up against, the graphics cards will jump TWO half nodes down to the latest state of the art 28nm process, this cannot be underestimated.
Furthermore, you will be going up against two microarchitectures, not just one. There is 2nd generation VLIW4 which you will be measured against. Then there is Graphics Core Next microarchitcture, which might turn out to be a very scary in addition to the two half node shrinks, but we don't have much info about this yet.

And as others have noted, the graphics cards are multipurpose, they can be used for other projects like F@H, they can be used for gaming, they can be given to family who can make use of them, they can be sold to other people as 2nd hand. This flexibility in itself also has a value, you need to somehow compensate for this to attract investors, I'm not sure if high hashrates alone is enough.

It won't hurt to be extremely careful when you're dealing with something like a 2 million dollar investment. But good luck you guys!

I think that as the other ASIC thread (The one that seemed to be a Chinese scam) is now pretty much dead, this one should be bumped up so that everyone can carry on with this discussion.

To the original OP, I think we should start compiling a list of interested ASIC investors here on the forum.  Perhaps the admins can send out an notice to let everyone know....

Regards,

As I said already  i am a willing investor. My current plan is fully scheuled out until end of the year, but this can be redirected.... and is not insignificant,unless I stop to invest into mining, whichwould be stupid given I am just signinf a 350 square meter lease only for mining ;)

I'd be interested.

Did interest for this just kind of die out or something? I have an investor that was asking me how easily someone could come up with an ASIC farm and why I wasn't running one myself recently.

The idea of the community doing this has pretty much died. But other groups are moving ahead. If you are interested in keeping up with LargeCoin, visit http://largecoin.com.

Why?  We're only seeing ongoing optimization towards what has been asked for.

We would see exactly the same thing happening with a "memory hard" algorithm.  Probably the trend would be to use big physical footprint chips with few logic resources, interfacing to many high density DRAM chips simultanously.  Memory octopuses so to say, keeping all DRAM busy at all times.  CPUs and GPUs would look pale against those, too, and you could THEN argue that a "CPU hard" algorithm should have been chosen instead ........ (?!)

Requiring more RAM would have no negative impact on the practicality of FPGA or ASIC mining. It's not terribly difficult to hang RAM off of an FPGA or ASIC. For that matter, the chip's memory bus architecture could be optimized for whatever the mining algorithm required, in contrast with a PC that has a general-purpose memory bus architecture.

No matter what algorithm you devise, a special-purpose machine can be designed which does it better than a general-purpose machine. It may or may not be cost-effective to design and produce that machine, but I don't think it's generally possible to devise an algorithm that is efficient on some CPU family but could not be done better with specialized hardware.

Posting in a necrobumped thread.   :D