Hash brute-forcing – CPU vs GPU – a quick overview


Software used:

  • Hashcat for CPUs
  • oclHashcat-lite for AMD graphic cards
  • cudaHashcat-lite for nVidia graphic cards

Results:

MD5

Equipment Speed (million brute-forces per second
HD 6950 4781.5
HD 5770 2897.7
HD 5750 2405.5
HD 4890 1743.5
GTX 560 Ti 1033.7
GTX 260 Core 216 718.1
8800 GT 489.8
HD 6310 86.5
i5-2500k (4 threads) 53.41
i5-2500k (2 threads) 28.80
i5-2500k (1 thread) 14.40

 

SHA1

Equipment Speed (million brute-forces per second
HD 6950 1675.1
HD 5770 1020.0
HD 5750 841.0
HD 4890 625.5
GTX 560 Ti 504.1
GTX 260 50.9
i5-2500k (4 threads) 45.78
HD 6310 32.1
8800 GT 31.7
i5-2500k (2 threads) 24.75
i5-2500k (1 thread) 12.51

 

AMD’s VLIW architecture owns nVidia’s Fermi and older architectures in this case. More parallelized and that’s what Cryptography is all about. Would be interesting to see how AMD’s Graphics Core Next fares up for this purpose.

What does 5 billion MD5 brute forces per second mean?

MD5 Hash of a 7 char string containing only lower-case alphabets and numbers cracked within 30 seconds. Think how fast 8 x HD 6990s would be.

Note: oclHashcat-lite didn’t work for nVidia cards. Don’t know why. Could be either – software or driver issue. Don’t consider this an Apples-to-Apples comparison. But still this is the best what each platforms can offer.