07-02-2016, 05:01 PM
07-04-2016, 12:55 PM
I though I've explained it in all level of detail in the release notes. Hashcat worksaround the problem as good as it's possible, but it's not solveable by hashcat in total. For -a 0 based attacks you will have a CPU load of < 20%
07-04-2016, 04:29 PM
What I'm wondering about is your solutions vs. comments out there about "for Maxwells, Nvidia drivers between 364.47 through 365.19 have a buggy PTX to GPU machine code compiler which causes them to miscalculate in GPGPU loads including CUDA and OpenCL loads." Vs. your guideline "NVidia users require NVidia drivers 346.59 or later (recommended 367.27 or later)."
Does the Nvidia driver 368.22 having "fixed" the miscalculation bug benefit hashcat v3? Partially or totally?
Seen at: https://forums.geforce.com/default/topic...-23-16-/8/ :
Does the Nvidia driver 368.22 having "fixed" the miscalculation bug benefit hashcat v3? Partially or totally?
Seen at: https://forums.geforce.com/default/topic...-23-16-/8/ :
Quote:RE: OpenCL/CUDA R364 miscalculations
Hey guys, I think this 368.22 driver does fix it! Testing is still preliminary, but so far, it has passed all my tests (and I'm the main person who worked with NVIDIA to solve it).
In short, I think the OpenCL/CUDA miscalculation fix did make it into 368.22, and it works! I'll chime back in if I find any problems with it, as I continue my testing.
Thanks,
Jacob
07-05-2016, 08:50 AM
No that's not related. I've tested with the latest 368 drivers, it's still buggy.