Core2Quad Q9300 2.5 GHz (one 32 bit thread): Cisco 5.64 MLinks/sec 7.48 MReduxes/sec 25.30 MHashes/sec [3xSSE2] MD5 5.65 MLinks/sec 7.35 MReduxes/sec 26.13 MHashes/sec [3xSSE2] Double Binary MD5 4.57 MLinks/sec 7.36 MReduxes/sec 12.95 MHashes/sec [3xSSE2] Double MD5 4.54 MLinks/sec 7.40 MReduxes/sec 12.22 MHashes/sec [3xSSE2] MD4/NTLM 6.07 MLinks/sec 7.40 MReduxes/sec 39.63 MHashes/sec [4xSSE2] SHA1 4.54 MLinks/sec 8.00 MReduxes/sec 12.79 MHashes/sec [1xSSSE3] MySQLSHA1/MySQL5 3.34 MLinks/sec 7.92 MReduxes/sec 6.35 MHashes/sec [1xSSSE3]There's a simple speed up by just commenting out code in keyspace.cpp for ((m_charaterEncoding & (CE_UTF16_LE | CE_UTF16_BE)) != 0) and ((m_charaterEncoding & (CE_UTF32_LE | CE_UTF32_BE)) != 0) but then NTLM will break. Next update will have a few changes for speeding up the reduction function such as three classes keyspace8, keyspace16, and keyspace32. Also needs variable length character support for UTF-8 and UTF-16. The key space used for benchmarking is loweralpha#1-7. This code can use key spaces such as Omni-6.
(WinRTGen speed info) Cisco [3xSSE2] 25.26 MHashes/sec (5.46x; 4.628 MHashes/sec) MD5 [3xSSE2] 26.14 MHashes/sec (4.65x; 5.618 MHashes/sec) Double Binary MD5 [3xSSE2] 12.90 MHashes/sec Double MD5 [3xSSE2] 12.29 MHashes/sec MD4/NTLM [4xSSE2] 39.60 MHashes/sec (5.94x/10.88x; 6.667 MHashes/sec/3.641 MHashes/sec) SHA1 [1xSSSE3] 12.82 MHashes/sec (5.38x; 2.383 MHashes/sec) MySQLSHA1/MySQL5 [1xSSSE3] 6.37 MHashes/sec (3.52x; 1.810 MHashes/sec)This is 3.52x to 10.88x (average 5.97x) faster than WinRTGen's hash speed. This is a little misleading since there's the reduction function. Hopefully it will have a link speed that is at least 2 times faster.