oclhashcat 21 speed reduction

[c4p0ne]

I just upgraded from 18 all the way to 21 and i've dropped down to a FRACTION of the speed. Now (with one nvidia 275 and one 260) i can barely do 100m hashes per sec with BOTH cards where with 18 i was doing over 900m hashes per sec.

wth happened????????????

[c4p0ne]

Welp, just upgraded to the latest bleeding edge nvidia drivers 257.21 (CUDA 3.1.1) and still get about 112m md5 per second. Abysmal compared to 900m+. Perhaps I'm doing something wrong, some new config option?

I should mention my hashlist is hovering around 4 million hashes (md5).

[atom]

please paste speed of v0.18 with this command:

oclHashcat.exe example.hash -1 ?l?d ?1?1?1?1 ?1?1?1?1 -n 8

and of v0.21 with this command:

oclHashcat.exe example.hash -1 ?l?d ?1?1?1?1 ?1?1?1?1 -n 8 --gpu-loops 1024

[c4p0ne]

please paste speed of v0.18 with this command:

oclHashcat.exe example.hash -1 ?l?d ?1?1?1?1 ?1?1?1?1 -n 8

and of v0.21 with this command:

oclHashcat.exe example.hash -1 ?l?d ?1?1?1?1 ?1?1?1?1 -n 8 --gpu-loops 1024

The speed it exactly the same for both (>1000m/sec) which is good. I've discovered that it is the size of my md5 hashlist that is slowing oclhashcat down TREMENDOUSLY. 3 million hashes slow it down to a crawl of 100m/sec when i chop the list in half to 1.5 million hashes, the speed jumps up to 500m/sec, etc... This was not the case with .18.

[Rolf]

Latest OCLHC is optimized to handle up to 0.5M hashes.

[c4p0ne]

Latest OCLHC is optimized to handle up to 0.5M hashes.

Hmm, it was becoming apparent that the latest version was optimized for lower numbers of hashes. Guess that gives me a reason to keep the old versions around then....

[atom]

fixed in v0.23