How about something like https://bitcointalk.org/index.php?topic=48863.240 !?
1GH/s
20w
$500
But real this time not a scam.
What do you think ? Impossible ?
Topic: X6500 Custom FPGA Miner - page 50. (Read 219954 times)
Thanks fizzisist, the adapter is for power supply (2.5mm) to splitter (2.1mm). I'll get spending!
Any delivery date projections currently?
Any delivery date projections currently?
Nope, I don't see a reason why that wouldn't work. Actually, the board has a 2.1 mm (inner diameter) x 5.5 mm (outer diameter) power connector, so you don't even need the second "this." Great find!
For the GLBSE system, I will use this ATX power supply, although it was cheaper when I bought it. There are a million different power supplies out there, so look around. The two main things I looked for were efficiency and a large current limit on the 5V rail. You can power the boards with either the 5V or the 12V (actually, anything between 5-14V) but it should be a little more efficient on the 5V rail. I'll probably power 6 boards on the 5V line and the rest on the 12V. You'll need to make a little adapter to jump the 12V over to the right pin, but I believe Cablesaurus is going to be selling those.
The early payment discount offer is about to end!
We've had a few awesome customers pay in full in BTC, taking advantage of the $20/board discount. We're now ready to get these boards built, which means we'll be ending the early payment discount very soon. If you pre-ordered and want to get this deal, put in your payment soon! You can go to the Cablesaurus.com product page to finish your order now.
If you didn't get in on the pre-order phase, or want to pay with Paypal, you'll have to wait just a little bit longer. We'll open up for new orders very soon and accept Paypal payments as soon as possible.
We've had a few awesome customers pay in full in BTC, taking advantage of the $20/board discount. We're now ready to get these boards built, which means we'll be ending the early payment discount very soon. If you pre-ordered and want to get this deal, put in your payment soon! You can go to the Cablesaurus.com product page to finish your order now.
If you didn't get in on the pre-order phase, or want to pay with Paypal, you'll have to wait just a little bit longer. We'll open up for new orders very soon and accept Paypal payments as soon as possible.
Making a transfer to Dwolla now! Hope I can get in before it closes!
The early payment discount offer is about to end!
We've had a few awesome customers pay in full in BTC, taking advantage of the $20/board discount. We're now ready to get these boards built, which means we'll be ending the early payment discount very soon. If you pre-ordered and want to get this deal, put in your payment soon! You can go to the Cablesaurus.com product page to finish your order now.
If you didn't get in on the pre-order phase, or want to pay with Paypal, you'll have to wait just a little bit longer. We'll open up for new orders very soon and accept Paypal payments as soon as possible.
We've had a few awesome customers pay in full in BTC, taking advantage of the $20/board discount. We're now ready to get these boards built, which means we'll be ending the early payment discount very soon. If you pre-ordered and want to get this deal, put in your payment soon! You can go to the Cablesaurus.com product page to finish your order now.
If you didn't get in on the pre-order phase, or want to pay with Paypal, you'll have to wait just a little bit longer. We'll open up for new orders very soon and accept Paypal payments as soon as possible.
Are this synthesizer or router results? I have one design which can achieve 292 MHz according to XST, but router delivers not more than 170 MHz.
Sorry, I was a bit unclear there. The 250 MHz is post place-and-route; you're right that it doesn't make much sense to quote anything else.
I'd certainly hope so; did a test synthesis run for the Kintex-7 70T a while ago with a variant of FPGAMiner's code I had lying around and ISE claimed it'd reach 250MH/s.
Are this synthesizer or router results? I have one design which can achieve 292 MHz according to XST, but router delivers not more than 170 MHz.
Quote
It maybe only a premilenary, so final price may be different.
That are no fantasy prices. You can order the parts from there (but you have to wait for it).
AFAIK (from my distributor) Xilinx defines the prices of its distributors.
Quote
7 series FPGA have almost identical architecture to Virtexes 6. Theres only different CLB/DSP/BRAM slices ratio. In this topic:
https://bitcointalksearch.org/topic/m.454555
iidx managed to put 3 unrolled cores (and almost 4) to XC6VLX240T. So it should be possible to put 2 in Kintex 160T. Spartan 6 have some limitations thats why 2 cores don't fit.
https://bitcointalksearch.org/topic/m.454555
iidx managed to put 3 unrolled cores (and almost 4) to XC6VLX240T. So it should be possible to put 2 in Kintex 160T. Spartan 6 have some limitations thats why 2 cores don't fit.
This was an optimized 1 stage per sha256 round design which achieved 125MHz. Speed of V6 and K7 is approximately equal.
I've looked few days ago for some prices on this devices but I didn't find any. Strange, there's only price for 325T. It maybe only a premilenary, so final price may be different.
7 series FPGA have almost identical architecture to Virtexes 6. Theres only different CLB/DSP/BRAM slices ratio. In this topic:
https://bitcointalksearch.org/topic/m.454555
iidx managed to put 3 unrolled cores (and almost 4) to XC6VLX240T. So it should be possible to put 2 in Kintex 160T. Spartan 6 have some limitations thats why 2 cores don't fit.
7 series FPGA have almost identical architecture to Virtexes 6. Theres only different CLB/DSP/BRAM slices ratio. In this topic:
https://bitcointalksearch.org/topic/m.454555
iidx managed to put 3 unrolled cores (and almost 4) to XC6VLX240T. So it should be possible to put 2 in Kintex 160T. Spartan 6 have some limitations thats why 2 cores don't fit.
With 28 milion bucks in pocket, I think better way is to make an ASICs. Buy one V7T test VHDL code and order 100.000 ASICs... More profitable, IMHO...
I can't wait to lay my hands on some Kintexes... 160T should have comparable price with LX150 but with good design it should reach 500MH/s.
I'd certainly hope so; did a test synthesis run for the Kintex-7 70T a while ago with a variant of FPGAMiner's code I had lying around and ISE claimed it'd reach 250MH/s. While that's obviously with a preliminary version of the tools, I'd be disappointed if the 160T couldn't reach at least 500 MH/s.I can't wait to lay my hands on some Kintexes... 160T should have comparable price with LX150 but with good design it should reach 500MH/s.
The Kintex-7 architecture seems to be an awful lot better suited to Bitcoin mining than Spartan-6.
Edit: 70T, not 75T.
XC7K160T will cost about 3 to 4 times as much as XC6SLX150 and should reach about 300 MH/s.
In some earlier topic someone who works in Xilinx said that 325T will be cost somwhere around 500$. Somwhere on Xilinx site I've read that prices will be comparable to Spartan 6. You have reach almost 200 MHz. Kintexes shouldn't have problem to double that. Using DSP's should also free up resources and thus allow to put two fully unrolled cores. So, 400MH/s should be possible without any hassle, and more with optimized disign.EDIT:
I've found it.
https://bitcointalksearch.org/topic/introduction-vhdl-fpga-mining-in-kintex-xc7k325tl-2-devices-post-1-29169
And that was in July. Many things have change since then, designs are better now (for example you have managed to almost double hashrate from S6).
At Digikey 325T devices cost about $1200. Digikey is usually a little bit expensive. Extrapolated 100 unit price should be about $800.
Even if all DSP's are used LUT utilization with 2 pilpelines would be about 130%. One 300MHz pipeline is realistic.
XC7K160T will cost about 3 to 4 times as much as XC6SLX150 and should reach about 300 MH/s.
In some earlier topic someone who works in Xilinx said that 325T will be cost somwhere around 500$. Somwhere on Xilinx site I've read that prices will be comparable to Spartan 6. You have reach almost 200 MHz. Kintexes shouldn't have problem to double that. Using DSP's should also free up resources and thus allow to put two fully unrolled cores. So, 400MH/s should be possible without any hassle, and more with optimized disign.EDIT:
I've found it.
https://bitcointalksearch.org/topic/introduction-vhdl-fpga-mining-in-kintex-xc7k325tl-2-devices-post-1-29169
And that was in July. Many things have change since then, designs are better now (for example you have managed to almost double hashrate from S6).
I can't wait to lay my hands on some Kintexes... 160T should have comparable price with LX150 but with good design it should reach 500MH/s.
XC7K160T will cost about 3 to 4 times as much as XC6SLX150 and should reach about 300 MH/s. 
With 28 milion bucks in pocket, I think better way is to make an ASICs. Buy one V7T test VHDL code and order 100.000 ASICs... More profitable, IMHO...
I can't wait to lay my hands on some Kintexes... 160T should have comparable price with LX150 but with good design it should reach 500MH/s.
Well in case of HFT they need that time to market advantage and reconfigurability high end FPGAs offer. I can't wait to lay my hands on some Kintexes... 160T should have comparable price with LX150 but with good design it should reach 500MH/s.
I think the best from all this is that the Spartan-6 will eventually be sold for wholesale prices, we have to keep our eyes open on ebay for that
With 28 milion bucks in pocket, I think better way is to make an ASICs. Buy one V7T test VHDL code and order 100.000 ASICs... More profitable, IMHO...
I can't wait to lay my hands on some Kintexes... 160T should have comparable price with LX150 but with good design it should reach 500MH/s.
I can't wait to lay my hands on some Kintexes... 160T should have comparable price with LX150 but with good design it should reach 500MH/s.
I wonder what would be possible with that chip 10GH/s? This price is nuts 30k$. In a whole year I can earn only 10k$.
Nice discount on 1000+ units - 1300$. I wonder who buys 28 milion worth FPGAs...
Well the people who need it are the high frequency trading firms in top locations, they pretty much have to buy them in order not to loose money on those who do. 10GH/s? This price is nuts 30k$. In a whole year I can earn only 10k$.Nice discount on 1000+ units - 1300$. I wonder who buys 28 milion worth FPGAs...
I wonder what would be possible with that chip 10GH/s? This price is nuts 30k$. In a whole year I can earn only 10k$.
Nice discount on 1000+ units - 1300$. I wonder who buys 28 milion worth FPGAs...
10GH/s? This price is nuts 30k$. In a whole year I can earn only 10k$.Nice discount on 1000+ units - 1300$. I wonder who buys 28 milion worth FPGAs...
Looks like the Virtex_7 released today. You guys thinking of incorporating that into V3?
http://www.xilinx.com/products/silicon-devices/fpga/virtex-7/index.htm
I'm sure it's really spendy. Anybody know pricing?
http://www.xilinx.com/products/silicon-devices/fpga/virtex-7/index.htm
I'm sure it's really spendy. Anybody know pricing?
Here's the Avnet link: http://avnetexpress.avnet.com/store/em/EMController/Miscellaneous/Xilinx/XC7V2000T-1FH1761C/_/R-5001699400122/A-5001699400122/An-0?action=part&catalogId=500201&langId=-1&storeId=500201&listIndex=-1
$30,000 for one! Is that right? 16 week lead time, too. If someone really wants one, we'll build a board!
Looks like the Virtex_7 released today. You guys thinking of incorporating that into V3?
http://www.xilinx.com/products/silicon-devices/fpga/virtex-7/index.htm
I'm sure it's really spendy. Anybody know pricing?
http://www.xilinx.com/products/silicon-devices/fpga/virtex-7/index.htm
I'm sure it's really spendy. Anybody know pricing?
So wtf... have you forensically analysed the code yet? 
By the way, it looks like the main thing of interest is the way he breaks down the calculation of W into multiple stages. (This isn't really surprising; the W calculation was what was driving me nuts and I think the same is true of FPGAMiner.) The design only actually has 2 pipeline stages per SHA-256 round.Edit: Also, the map and place and route options - ouch!
Jump to: