hashcat Forum

I have a multigpu 7970 system and i've noticed an almost order of magnitude drop in the number of #/sec between 0.15 and 1.01.

Previously (using -n 80 -u 320) i'd get ~130k #/sec per card and now I get about 17k/sec.

Has anyone else experienced this? I'm running Catalyst 13.2 drivers as well

why would you use -u 320 for wpa? that doesn't make any sense. use -u 4096

(01-06-2014, 07:45 AM)epixoip Wrote: [ -> ]why would you use -u 320 for wpa? that doesn't make any sense. use -u 4096

Code:

epixoip@butters:~/oclHashcat-1.01$ ./oclHashcat64.bin -m 2500 -b
oclHashcat v1.01 starting in benchmark-mode...

Device #1: Tahiti, 2967MB, 925Mhz, 32MCU
Device #2: Tahiti, 2967MB, 925Mhz, 32MCU
Device #3: Tahiti, 2967MB, 925Mhz, 32MCU

Hashtype: WPA/WPA2
Workload: 4096 loops, 32 accel

Speed.GPU.#1.:   130.9 kH/s
Speed.GPU.#2.:   131.0 kH/s
Speed.GPU.#3.:   131.1 kH/s
Speed.GPU.#*.:   392.9 kH/s

thanks for the quick reply,
so using -u 4096 in benchmark mode i get figures like yours however when I use it with my wordlist that's when it drops to ~17k/sec.

apologies if this has been covered before...

did you prepare your wordlists?

(01-06-2014, 09:49 AM)epixoip Wrote: [ -> ]did you prepare your wordlists?

do you mean by running them through splitlen and then recombining them?
if so yes and it reduced the #/sec from 17k/sec to ~13k/sec.

hm, that's really weird. are you giving them enough work to do? maybe you can post the output of `s' during a run.

(01-07-2014, 05:20 AM)epixoip Wrote: [ -> ]hm, that's really weird. are you giving them enough work to do? maybe you can post the output of `s' during a run.

and here's oclHashcat plus 0.15

It's pretty strange, with the exact same wordlist and hccap file I get ~130k#/sec per card so i've got no idea what the story is.

i did notice on the oclhashcat wiki page it says to use Catalyst 13.4 but as far as I can tell 13.2 is the latest relase for linux so maybe that just applies to Windows?

Oh and those screen dumps remind of something else, what happened to the "Hash.Target" field in 1.0? I found the mac address and essid description useful

The screenshot of oclHashcat 1.01 shows that you are trying to crack more than one network (.hccap file w/ several essid - note the file size of the hccap file + "Recovered......: 0/10"). That's also the reason why not every MAC is listed there.

As far as I understood, you said that the speed of 1.01 and plus 0.15 is the same w/ same input (wordlist + hccap file), therefore it could be that indeed the fact that you are trying to crack 10 network at the same time, could be the reason for the slower speed.

ah yep, that's it. he's cracking 10 different hashes with 10 different salts. 13099 * 10 = 130990 H/s which is the expected speed.

(01-07-2014, 07:42 AM)philsmd Wrote: [ -> ]The screenshot of oclHashcat 1.01 shows that you are trying to crack more than one network (.hccap file w/ several essid - note the file size of the hccap file + "Recovered......: 0/10"). That's also the reason why not every MAC is listed there.

As far as I understood, you said that the speed of 1.01 and plus 0.15 is the same w/ same input (wordlist + hccap file), therefore it could be that indeed the fact that you are trying to crack 10 network at the same time, could be the reason for the slower speed.

thanks for picking that up on that, that capture is the wpa2 test capture in the aircrack suite and for some reason when i use cap2hccap to create the hccap file from the packet capture it puts multiple copies of the struct in the resulting hccap file.

using aircrack -J solves that.

love your work.