Re-purposing of FPGA boards made for mining

[sturle]

Bitcoin mining and other typical uses of FPGA differ in an important aspect.  While mining has very low demand for I/O, it is crucial for many other typical FPGA tasks.  E.g. HDMI dectyption, wide-band GSM demodulation and various crypto purposes.

Do anyone have experience in using mining FPGA boards for other purposes?  E.g. WPA cracking?  Fast DES decryption with unknown key?  Are some boards more fit for this than other boards?

[Bitznbitz]

I'm sure the NSA has a lot of information about this..

[Desolator]

I've got an idea! 

http://www.youtube.com/watch?v=zrg8nJ0bsUk

[regular]

are wpa cracking firmwares allowed to be released publicly?  Or are there laws against it?

[sturle]

are wpa cracking firmwares allowed to be released publicly?  Or are there laws against it?

In Germany you are not allowed to have software for cracking WPA keys.  I don't know of any other countries where it is forbidden, but IANAL.  Pyrit (WPA gracking software for GPUs) is hosted on Google code.

[stevegee58]

Haha good luck with making source code illegal.  Once the genie's out of the bottle that's it.  And it is out.

[emuLOAD]

Yeah but the catch is that while the genie may be out, you're in (prison) ^_^

Making things disappear of the internet is pretty damn hard, impossible maybe, but that won't stop a government trying and prosecuting who it wishes on the way.

[Ghostofkobra]

In Germany you are not allowed to have software for cracking WPA keys.  I don't know of any other countries where it is forbidden, but IANAL.  Pyrit (WPA gracking software for GPUs) is hosted on Google code.

Ha,

     iAnal

Sounds like some new Apple software

[stevegee58]

How would they find out?  Simply cracking a WPA key is undetectable.  Now, actively accessing someone's network with that key is another matter...

There must be a lower threshold for probable cause in Germany than in the US.

[emuLOAD]

The law sturle is referring to is not agains cracking WPA networks or accessing them, it has nothing to do with criminal or otherwise use of the software, it merely bans possession of the software (the law is in fact broader banning a whole load of tools related to "computer security"). They need not prove you used or had intention of using the software for whichever purpose, simply having the binary or code is against that law.

Is the law moronic, yeah of course. doesn't invalidate it though.

Not sure what probable cause has to do with this?

[lucif]

Really interested in getting BFL firware for WPA security test

[sturle]

Really interested in getting BFL firware for WPA security test

Any FPGA.  I hope there are reasonably priced FPGAs on the market by January. :-)

[exotime]

If it's a FPGA, you should be able to reprogram it to perform whatever algorithms you want, provided you have the know-how, or know someone with the know-how that likes money.

I doubt you'll find a "freely downloadable FPGA script to crack WPA" for your hardware. You'll probably have to write something yourself, or modify something you've already found.

[sturle]

If it's a FPGA, you should be able to reprogram it to perform whatever algorithms you want, provided you have the know-how, or know someone with the know-how that likes money.

The FPGA is one thing, how it is wired on the board is important as well.  E.g. HDMI decryption would need high speed I/O in both directions.  Up to 10.2 Gbit/s video + 36.86 Mbit/s audio.  Which of the mining boards can do that?  The application is very different from Bitcoin mining, where you don't need high speed I/O at all.  300 bit/s would be plenty.

WPA2 needs a little more I/O than Bitcoin mining.  Enough that I wonder if some boards may hit a bottleneck without adding some extra logic to the FPGA, but what do I know.  There are many applications with I/O needs in between.  E.g. filtering in a SDR.

[exotime]

I think you're overestimating how finely engineered the bitcoin FPGA's would be. It's quite unlikely that the IO bus has been tuned to the barest minimum before affecting hash rates, I'd think it's much more likely that they were a) the easiest to acquire/use, and b) the cheapest. Odds are the low end FPGA's from the same manufacturer use the same bus as the high end ones, it's just a different processing unit and oscillator.

You aren't decoding video here, you're still manipulating strings. WPA has a key length of 256 bits.

All the same, buy a few different ones, try them out and let us know!

[sturle]

I think you're overestimating how finely engineered the bitcoin FPGA's would be. It's quite unlikely that the IO bus has been tuned to the barest minimum before affecting hash rates, I'd think it's much more likely that they were a) the easiest to acquire/use, and b) the cheapest. Odds are the low end FPGA's from the same manufacturer use the same bus as the high end ones, it's just a different processing unit and oscillator.

I'm not worried about the FPGAs.  Those are quality components.  I am worried about how it is wired on the board.  If they chose to do it in the cheapest possible way, as you suggest, the boards will be unfit for almost anything but mining.

You aren't decoding video here, you're still manipulating strings. WPA has a key length of 256 bits.

With HDMI decryption you are decrypting live video at higher bitrate than the fastest Ethernet.  I haven't mentioned WPA as anything but one of many examples.  (And you obviously don't know anything about WPA either.)  I want to know the limits of the different mining boards.

[Zeek_W]

http://aktuell.ruhr-uni-bochum.de/mam/images/pi2011/fittosize_210_0_6179b3811f0390872c1beab167c880a0_man-in-the-middle.jpg

http://aktuell.ruhr-uni-bochum.de/pm2011/pm00386.html.en

[hardcore-fs]

I'm not worried about the FPGAs.  Those are quality components.  I am worried about how it is wired on the board.  If they chose to do it in the cheapest possible way, as you suggest, the boards will be unfit for almost anything but mining.

No matter what the state of the FPGA mining board, your application is going to need to have a PCB that is designed specifically for the signals you are trying to handle.

If it is LVD pairs, then no 'miner' board is going to be physically configured for the signal propagation OR layout/ ground-planes etc.
(look at some of the miners, they require just THREE lowspeed connection to the outside world (CLK_IN,TX,RX).
Even a six year old with a pack of crayola crayons ,could design a PCB layout for that.

You would be far better just buying scrap telecom equipment to recover the FPGA's ,re-ball and get them  mounted on the PCB you require.

Plus looking at the costs of some of these mining boards, unless you age going to pick them up for $50-$150 , it will still be close to almost the cost of a new FPGA.

If you are not interested in designing your own PCB, then I'm afraid it's going to require a development board from one of the manufacturers, certainly at the frequencies you are considering.


HC

[sturle]

No matter what the state of the FPGA mining board, your application is going to need to have a PCB that is designed specifically for the signals you are trying to handle.

If it is LVD pairs, then no 'miner' board is going to be physically configured for the signal propagation OR layout/ ground-planes etc.
(look at some of the miners, they require just THREE lowspeed connection to the outside world (CLK_IN,TX,RX).
Even a six year old with a pack of crayola crayons ,could design a PCB layout for that.

You would be far better just buying scrap telecom equipment to recover the FPGA's ,re-ball and get them  mounted on the PCB you require.

Plus looking at the costs of some of these mining boards, unless you age going to pick them up for $50-$150 , it will still be close to almost the cost of a new FPGA.

If you are not interested in designing your own PCB, then I'm afraid it's going to require a development board from one of the manufacturers, certainly at the frequencies you are considering.

Thank you!  Finally an informed answer!

As far as I understand some of the popular mining FPGA boards is basically re-branded general FPGA development boards.  E.g. Icarus.  This looks like it can be used for fairly high bandwidth purposes.  (HDMI is extreme, I don't expect any mining board to be able to handle full speed HDMI.)  Other boards are more custom mining specific designs, and may be designed by six year olds with crayons.  There must be a huge difference in what applications different popular FPGA mining boards can handle.

I guess t would be better to ask my question when the ASIC miners have been shipped and people have idled their FPGAs. :-)

[hardcore-fs]

Yep, many are just  development boards sometimes with the 'extra' stuff stripped off.
But even if you look at the basic setup of an FPGA development board, they are heavily engineered for reliability around the power sections.
But looking at some of these mining boards, it is a bloody miracle they even fly, but it does not stop there.

I've been diddling about with 'Open-Source-FPGA-Bitcoin-Miner' on a XUPV5 (i'm bored!!!)
After looking at one of the logic 'paths' it introduces a 4ns delay into the critical logic!!! (bad VHDL), and since it is a critical path it bolloxes up the FPGA design to the tune of -3ns per hash.

It took me just under 10 minutes to get that path down to 0.094ns in the gate delay and the net delay is now 0.694ns and from 74% utilization of the device LUTs to 62%

Which is like a 66% increase in speed for the net ,plus I have not even touched manual routing!!!

I was looking at "mining hardware comparison"  for the XUPV5,
but I cannot find the files "fpgaminer virtex5" to see if they can be improved, THESEVEN ain't responding either!!!


HC