04-17-2013, 05:10 AM
I'm working up a set of prefix and suffix based dictionary cracks, and my very first pass was 1, 2, and 3 digit prefixes/suffixes.
Using oclHashcat-plus64 0.14, on an AMD HD6950 with Catalyst 13.1, Windows 7 x64 SP1, with fairly large dictionaries against a 14 million hash unsalted MD5 list, I was seeing between 384KH/s and 10.6MH/s (higher speeds with 3 character masks) with attack modes 6 and 7 (mask+dict, dict+mask).
When I switched over to using Maskprocessor to generate rules files, the 1 character rules saw 177MH/s and 2+ saw 260-277MH/s, which is much more in line with what I was expecting.
Note that Markov/bruteforce mode also shows 274MH/s on the same size list (up from 138MH/s on the original set of 25 million words), so that appears to be the top speed I can expect at this time.
My comment: If the current implementation of modes 6 and 7 are so slow, then perhaps it would be better to instead keep the command line argument the same, but internally include a copy of the Maskprocessor code and generate an internal rules list, then use the rules engine?
Using oclHashcat-plus64 0.14, on an AMD HD6950 with Catalyst 13.1, Windows 7 x64 SP1, with fairly large dictionaries against a 14 million hash unsalted MD5 list, I was seeing between 384KH/s and 10.6MH/s (higher speeds with 3 character masks) with attack modes 6 and 7 (mask+dict, dict+mask).
When I switched over to using Maskprocessor to generate rules files, the 1 character rules saw 177MH/s and 2+ saw 260-277MH/s, which is much more in line with what I was expecting.
Note that Markov/bruteforce mode also shows 274MH/s on the same size list (up from 138MH/s on the original set of 25 million words), so that appears to be the top speed I can expect at this time.
My comment: If the current implementation of modes 6 and 7 are so slow, then perhaps it would be better to instead keep the command line argument the same, but internally include a copy of the Maskprocessor code and generate an internal rules list, then use the rules engine?