+- hashcat Forum (https://hashcat.net/forum)
+-- Forum: Support (https://hashcat.net/forum/forum-3.html)
+--- Forum: hashcat (https://hashcat.net/forum/forum-45.html)
+--- Thread: amd vs nvidia (/thread-9232.html)
amd vs nvidia - bla - 05-18-2020
I have two machines, one is running nvidia gtx 1080 the other amd radeon vii.
Hashcat on nvidia machine takes around 20% processor power while hashcat is running (distributed across the cores). The same command on radeon machine takes almost no processor power. Both are running on the same operating system: win10. I use only a single gpu for hashing (no legacy etc) on both.
I have read in an old thread (https://hashcat.net/forum/thread-6712.html) that this can be associated with how nvidia run things. Is that so? Is there a way to fix this?
Edit: using hashcat-5.1.0%2B1803
RE: amd vs nvidia - philsmd - 05-18-2020
are you talking about a busy CPU core ? This only happens with Nvidia running OpenCL kernels.
You could try to use the new beta with CUDA: https://hashcat.net/beta/ and cuda from https://developer.nvidia.com/cuda-downloads
RE: amd vs nvidia - bla - 05-18-2020
Well not a core coz the load is distributed on all cores but yes.
This is exactly the setup I'm using, latest beta hashcat and cuda_10.2.89_441.22_win10.exe
RE: amd vs nvidia - bla - 05-18-2020
Also:
It seems I'm able to use cuda or opencl (cuda is device 1 while opencl device 2) on nvidia machine. Cuda gives 10% better hash performance (-m 22000). The cpu load is the same using -d1/-d2
I'm getting https://hashcat.net/faq/wrongdriver warning on nvidia and no warning on amd.
The same attack requires a different memory amount it seems
Host memory required for this attack: 1725 MB on nvidia
Host memory required for this attack: 5047 MB on amd