Slow Speed With Tesla V100
#1
So i just get a google cloud instance for test with k80, i install drivers with these

https://killtacknine.com/test-driving-go...h-hashcat/


After these for test, i remove the K80 and put a v100 and just run hashcat and get those numbers(ridicullus)

Code:
627 MH/s
[Image: xpD8hOg.png]



Compared to those benchmarks, i think im doing something bad

https://hashcat.net/forum/thread-6972.html

Code:
Hashmode: 2611 - vBulletin < v3.8.5

Speed.Dev.#1.....: 15978.1 MH/s (84.10ms)

Command i use

Code:
screen -L -dmS hashcat ./hashcat64.bin -m 2611 --username --potfile-path 2611.pot hashes.txt wordlist.txt



Running hashcat GPU status

[Image: R0uiQ3g.png]




Also i can do something for better performance? like what is the best attack mode for a lot hashes at a time with 300GB wordlist? im testing in Different databases hashes.
#2
You can use -w4 for slightly better speed. Other than that, you are comparing a straight wordlist attack with over 3m hashes with a bruteforce on one single hash. Performance will definitely be worse.

Besides, if your wordlist is 300gb it's probably crap.
#3
(09-09-2018, 04:04 PM)undeath Wrote: You can use -w4 for slightly better speed. Other than that, you are comparing a straight wordlist attack with over 3m hashes with a bruteforce on one single hash. Performance will definitely be worse.

Besides, if your wordlist is 300gb it's probably crap.

for trying to crack the most % of a DB isnt the best to have more passwords?

I get around 27Million lines with k80 in 12 hours(96.44%) just like these, difference? it was md5, and these is vbulletin(pretty worst).

And what about attack mode.
#4
If you want to compare the speed of two cards at least use the same hash list and the same attack. Everything else is not going to give much meaningful information.

300gb is crap because it will mainly consist of somehow generated garbage, hardly better than brute force. Use clever attacks that involve proper wordlists with rules and other mangling.