02-27-2012, 05:19 AM
I was using cudaHashcat-plus64.bin 0.07 to crack NTLM hashes with my own wordlist, and found it returns inconsistent results. It doesn't seem to be dependent on the hashfile, but does seem to be somewhat dependent on the wordlist file.
This was tested on two machines. The first is i7 2600, 8gb ram, GTX 295, Fedora 16, and nvidia driver 290.10. The second is Q6600, 8gb ram, GTX 260, Fedora 16, and nvidia driver 290.10.
To reproduce, download the files at the url below. First uncompress test-wordlist.txt.gz. Then run ./tesh.sh under Linux. Watch the output. It should give you the number in the loop and three passwords each time. Out of fifteen runs I had three cases of only two passwords.
http://higgs.cygnusx-1.org/~edgan/oclHashcat/
This was tested on two machines. The first is i7 2600, 8gb ram, GTX 295, Fedora 16, and nvidia driver 290.10. The second is Q6600, 8gb ram, GTX 260, Fedora 16, and nvidia driver 290.10.
To reproduce, download the files at the url below. First uncompress test-wordlist.txt.gz. Then run ./tesh.sh under Linux. Watch the output. It should give you the number in the loop and three passwords each time. Out of fifteen runs I had three cases of only two passwords.
http://higgs.cygnusx-1.org/~edgan/oclHashcat/