Can CUDA MPS limit the GPU memory usage of a client process?

Accelerated Computing CUDA CUDA Programming and Performance
#1

Hi,

I know for Volta GPUs setting CUDA_MPS_ACTIVE_THREAD_PERCENTAGE in a MPS client’s environment will limit the client’s thread usage. Is there a similar way to limit the GPU memory usage for a client process?

thanks,
Xiaoning

1 Like
#2

Anybody knows? Thanks.