If you are receiving one of the following errors:
Then you should apply one of the following patches depending on your OS.
While running hashcat with settings that let a single kernel launch run for a long time, users may experience crashes, followed by automatic recovery of the GPU(s) via driver reset. This is caused by kernel runtimes that exceed the 2-second limit. That limit is useful for preventing screen freezes, but not useful for hash cracking.
On the 7.1.2 release the usual way to reach it is -w 3 or -w 4. Workload profiles are retired on master, where autotune sizes the launch itself and picks a larger one than the old default, so the timeout matters there without your asking for it. -n, -u and -T are what size it by hand.
To disable WDDM recovery, create a file called wddm_timeout_patch.reg with the following contents:
Windows Registry Editor Version 5.00 [HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\GraphicsDrivers] "TdrLevel"=dword:00000000
Run this registry file as Administrator and reboot.
Add a line “Interactive” in /etc/X11/xorg.conf:
Section "Device"
Identifier "Device0"
...
Option "Interactive" "False"
EndSection
You might need to restart X11 or reboot your system entirely before this change works and the warning in hashcat disappears.
Newer versions of Ubuntu (and maybe other distributions as well) do not use xorg.conf any longer. Alternatively they use the folder /usr/share/X11/xorg.conf.d/ where you can put in snippets of the config. Create a new file, call it “20-nvidia.conf” and put in the following content:
Section "Device" Identifier "MyGPU" Driver "nvidia" Option "Interactive" "0" EndSection
You may need to restart the system afterwards.