Bitcoin Forum

http://www.achronix.com/products/speedster22ihd.html
http://www.achronix.com/products/speedster22ihd/hd-summary.html

I wonder what the approximate mh/w here is.

They have access to Intel 22 nm foundries. The deal probably works somewhat like this: Achronix has a minimum guaranteed number of monthly wafers, and an option to run more wafers in case Intel does not need the full fab capacity. For both parties, this is a win-win situation.

 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.

So, someone please answer me: Why is there not an Achronix FPGA sitting on my desk mining at like 1 ghash on a single chip on like 10 watt?

Because they haven't been released yet. They do have in-house prototypes, but still a lot of validation ahead of them prior to release.

And propably lack of free software. I didn't see any free oferings from them. And tools like that costs 1000's of $.
Oh, I've read somewhere that they have bought 1% of Intel 22nm fab production.

Is BFL obsolete already ?

Arguably they were obsolete before they even came out. BFL has been purposely finding solutions that are very cheap per mhash, not highest efficiency, and generally thats been done with hardware no one wanted anymore.

So then GPUs are obsolete too?

The only thing that is obsolete on bfls is that u had to wait so long for them to arrive.

I consider buying one of those units before thinking of building a 7970 Miner.

They have fair products. Don't expect them to lean back and sleep. It's a fast moving game and nobody stays on top very long.

Obsolete?  Man, if so, show me another product that does more for less!  Otherwise, you'd be hard pressed to define them as obsolete.

update - their site says "now shipping" does that mean anything useful to the hobbyist? Anyone ordered a dev board :P ?

Thanks for the heads up, just about to send an email for a HD1000 Dev Board.

I already have 6 Spartan6x150's on their way, but damn.

Ive also contacted them.


Whats the direct link??

Their Contact Form: http://www.achronix.com/company/contact-us.html

I'm already in contact with a rep from them named Ken, setting up to by a dev board, then possibly some HD1500 22i's.

Does anyone have a price quote for the dev board yet?
And/or the FPGA itself?

Nothing back from their rep just yet, but looking at the specs on the chips, they would make killer hashers, but I expect a price point north of $1,500/chip based on the specs.

I guess it depends entirely on how well they hash to see if they're worth the money.

Any devs willing to write mining software for these chips?

I feel like that's a bridge better left crossed after we at least get a price point.

how many hash could this chips run? 22nm should be fast and low cost power consumption, lets hope the price is reasonable lol ......... :)

Any guess on hash rates right now would be pure speculation based solely on the chips specs.  Can I say they will be better than Spartan6's? Most certainly.  Will they be worth the price? I guess we will find out, huh?

Wild speculation: They can probably fit 5 hashing cores and run faster than a spartan 6...maybe 1 to 2 gh/s in total. Could be way off. Hard to guess about the power usage.

They might be worth it, if they're cheap and the btc price stays high.

Well, you'd use rolled cores n this. So you potentially have hundreds of cores on this chip.

Achronix's response on price points.

This chip is not going to hash well. For each 10 LUTs there is only two 2bit hardware adders with carry chains. Ordinary LUTs don't have carry chains. In Spartan 6 at least half of LUTs had carry chains, and this is very important to build fast adders, and those adders comsumes most of the logic resourses when used in SHA256 algorithm... Cheaper and better (in term of price/performance ratio) should be Atrix7's.

hi guys,

i ve discussed with one of their representative and he said that basically it would be possible to run 7 instances of the spartan equivalent instances in their chip.beside that the clock speed is around 600/700mhz so it would get better speed than the spartan6(wich is running at 50mhz if i m not wrong)

so basically he said the speed of this fgpa could be X87 times better than the spartan6

any of you who have technical background in fgpa can confirm this?

and another question, is there any chips similar to the achronix one with the other competitors,altera,xylinx or lattice?

thank you guys

200mhz is the current figure (without manual placement).

It's unlikely the rep meant 600/700mhz for the hashing designs; he probably meant that the fabric itself can run at 600/700mhz.  In other words, the registers and block ram will run at 700mhz, but any complex logic will run much slower.  For reference, the Kintex 7 fabric runs around 500mhz, but that doesn't mean a CLB based miner would run at 500mhz (more like 300).

Looking at just the LUT count, I would agree with the rep, and I imagine they did the same (look at the LUT count).  However, as Dexter770221 mentioned, performance could be much worse if their FPGAs are sparse in terms of carry chains.

At $3k per chip, I don't personally see a good performance/$ here.  It just sounds a lot like a Virtex/Stratix chip.

so x87 Would not be realistic?

what would be the realistic speed we could probably get out of it?

i was having in mind to make run these chips in on pcb with multiple FGPA on it and get some interesting output

what do you think?

thanks

http://investorshangout.com/post/614814/SANTA-CLARA-CA-Marketwire--Feb-20-2013--Ac

"The HD1000 development kit is available for purchase immediately at a price of $13,000, including ACE development software."

 :o

Is their place and route software expensive (not counting 3rd party synthesis and timing analysis tools)

It's included in their dev kit, which is $13000. Other vendors charge about 3 grand for their software, the Achronix FPGA costs about 3 grand - thus, I'd say, Achronix charges you more for the software, basically 9 grand or so. Do the math.

one of their representative in europe said me that the software will be free for now

the only issue is that their chips have not finished all the due tests so you can have the chips but you will be the one who will test it with all the bugs etc.

so this is not ideal fgpa to use right not ,not only the dev board +chip is expensive but because there is lack of docs because it s a new product(first people have just start to dev on the chips last end feb/beginning march)


i think kintex7 could be more interesting as multiple instances equivalent to spartan 6 could be implemented.

does any one have the kintex7 and could do the implementation to confirm this?

i could be interested in several if they give 2/3 giga per chips and even hire soemone to work on pcb for this for big quantity

thanks

The XC7K480T would probably do that, but I would assume that it's not cheap...

3800 usd for one chip

need to check in big quantitiy but i think it won t do