An FPGA based SHA-256 Processor (whitepaper)

[btc_artist]

http://www.ee.usyd.edu.au/people/philip.leong/UserFiles/File/papers/sha_fpl02.pdf (PDF warning)

Just ran across this. What do you make of it?

[1714231772]

1714231772

[1714231772]

1714231772

[1714231772]

1714231772

[DeathAndTaxes]

His performance seems to be far below what our "local FPGA heros" have achieved.

To be fair the paper is on a streaming hash engine (i.e. hashing a 20KB document).  Bitcoin is a somewhat unique (easier) form of hashing.  For each set of 2^32 attempts, all the data remains the same except the nonce.  That scenario never exists outside of Bitcoin.  Also Bitcoin is able to "cheat" because we are only interested in the most significant digits (to see if hash is below the target) which allows Bitcoin specific implementations to cut rounds off the end.

So comparing to a processor designed for general purpose hashing to Bitcoin isn't exactly fair but he acheived 86MB/s.  Bitcoin block is 512 bytes and it requires a double hash so that is ~0.08 MH/s or 18 KH/s.  The best homegrown FPGA designs beat that by a factor of 2500x.

[btc_artist]

our "local FPGA heros" have achieved.

Are you referring to the FPGA board sold at Cablesaurus?

[DeathAndTaxes]

our "local FPGA heros" have achieved.

Are you referring to the FPGA board sold at Cablesaurus?

Not just the X6500.  Any FPGA results have vastly exceeded those performance specs.  Ztek, rph, Icarus & X6500 are all a couple magnitudes higher performance than the results in the paper.