Black screen when running CUDA 11.2 code on a Windows 10 system with two RTX 2080tis

We have a Windows 10 system that has two RTX 2080ti GPUs with no additional dedicated GPU for display. We have the latest cuda toolkit 11.2 installed with driver version 461.09. When running cuda code (even the default nvidia code samples) on the system we observe that the display(monitor) blacks out during the entire run of the kernel code. After the kernel run, the display wakes up and we observe that the kernel code mostly fails with a 999 unknown error.

What could be the cause for the black screen and the consistent kernel failure?

Note:

  1. We observe the black screen even with kernels that have extremely small runtimes (for example the default “deviceQuery” solution).
  2. We are not able to reproduce the black screen and the error on a gaming laptop which has gtx 1660ti card installed. (We are not sure if the issue is specific to the RTX 2080ti family.)
1 Like