oclHashcat-plus v0.09b52 speed issues
#1
Hi atom, been testing this today against v0.08 with a clean install of the cat 12.8 drivers on Win7 x64.

Results:

0.08
NTLM 10 hashes: 3060 mil/s
NTLM 1.6mil hashes: 1158 mil/s
MD5 1.6mil hashes: 791 mil/s

0.09b52
NTLM 10 hashes: 1568 mil/s
NTLM 1.6mil hashes: 340 mil/s
MD5 1.6mil hashes: 623 mil/s

What going on there? NTLM is 4 times slower with 1.6 mil hashes.

Cheers dude.
#2
Hey Blandy,

yeah, I can reproduce it:

v0.08
Quote:root@sf:~/xy/oclHashcat-plus-0.08# ./oclHashcat-plus64.bin -m 1000 /root/testhashes/10 -a 3 -1 ?l?d?s?u ?1?1?1?1?1?1?1?1
Speed........: 6350.6M c/s Real, 6270.6M c/s GPU

v0.09b52 (notice the --bf-min added)
Quote:root@sf:~/xy/oclHashcat-plus-0.09# ./oclHashcat-plus64.bin -m 1000 /root/testhashes/10 -a 3 -1 ?l?d?s?u ?1?1?1?1?1?1?1?1 --bf-min 8
Speed........: 4479.6M c/s Real, 4561.5M c/s GPU

WTF I did not change anything to the code. Maybe new drivers require some new workarounds...

Thanks!
#3
Thanks dude, yeah sorry I didnt post the oclhc-plus command-lines used. I did used the --bf-min=6 when using charset -1 ?d?l?u?s with mask ?1?1?1?1?1?1

All tests using same -n 320 and masks / lengths.

Let me know when u have new BETA and I'll test for you.
#4
NV has done something to either the driver, or to the cuda 5.0 toolkit.
Litecat, v0.10 release vs 0.11b20 is always a bit slower in all algos, except LM and old oracle 7-10g.
Rolling back to good old 296.10.
#5
Testing using 12.8 drivers:

- did a run against a list of 6k ipb (2811) and with latest beta, simple dic
attack, using only rockyou.dic file as dic, took 3 minutes;

- then running exactly the same attack, but using 0.07, took 2 minutes 20 seconds.
#6
Many people report bad speed with cat 12.8, have you tried with cat 12.6?
#7
Just did an uninstall of 12.6 believing that 12.8 was giving
good news. At the moment can't go back to 12.6 and, who
knows, maybe someone around here is still using 12.6 and
can give some feedback.
I'll try, as soon as I can, but can take a while for work
reasons.
#8
continue here pls: https://hashcat.net/forum/thread-1523.html