GeekBench 5.2.0 Pro fails to initialize on Linux 5.4.46 with vfio-pci

jeff.geekbenchsupport's Avatar


13 Jun, 2020 10:27 PM

I have GeekBench 5.2.0 Pro installed on my host hypervisor. Recently, it won't initialize if my nvidia GPUs are bound to the vfio-pci driver. It seems that GeekBench tries to initialize the nvidia driver and then fails to recover when it won't start. Every invocation of geekbench litters dmesg with the failed nvidia init messages. It's fine for GeekBench to attempt to initialize the driver, but it should be able to proceed when that process fails.

Log from geekbench --help --cpu is attached. The errors appear as follows:

[0613/162216:WARNING:src/halogen/cuda/cuda_common.h(29)] CUDAError: cuInit(0) returned 100
[0613/162217:WARNING:src/halogen/cl/exception.h(35)] halogen::RuntimeError: clGetPlatformIDs(0, 0, &num_platforms) returned -1001
[0613/162218:ERROR:src/interface/console/consolemain.cpp(768)] Geekbench encountered an internal error and cannot continue. Please contact [email blocked] for assistance.

Internal error message: basic_string::_M_construct null not valid.

Reply to this discussion

Internal reply

Formatting help / Preview (switch to plain text) No formatting (switch to Markdown)

Attaching KB article:


Already uploaded files

  • gb-help-fail.log 1.37 KB

Attached Files

You can attach files up to 10MB

If you don't have an account yet, we need to confirm you're human and not a machine trying to post spam.

Keyboard shortcuts


? Show this help
ESC Blurs the current field

Comment Form

r Focus the comment reply box
^ + ↩ Submit the comment

You can use Command ⌘ instead of Control ^ on Mac