Error: GPU is lost when increasing the load

[ahmed.asad19]

Hello everyone,
I have a Nvidia 1070ti (8gb) on my system with following specification:
16GB RAM, Ubuntu 16.04, Nvidia-driver 418.87.01 and Cuda 10.1 installed.
Recently, I am getting “GPU is Lost” error when I am using nvidia-smi.

I have tried reconnecting the GPU and reinstalling the drivers along with CUDA. The issue gets resolved but if the usage increase over 4GB, same error comes again.

What should I do? Is it a hardware or driver issue?

Screenshot of the error is attached.

[ryork]

Have a look at this thread : [url]https://devtalk.nvidia.com/default/topic/1064726/cuda-setup-and-installation/quadro-rtx-6000-gpu-cards-disappearing/[/url]

You might be having the same problem.