The MPS document says changing the CUDA_MPS_ACTIVE_THREAD_PERCENTAGE variable does not affect the existing client process and/or client CUDA contexts. I am wondering why MPS is designed in this way? Why couldn’t the MPS change the active thread percentage while the process or the context is running? Is it because of the scheduling algorithm behind it?
Related topics
| Topic | Replies | Views | Activity | |
|---|---|---|---|---|
| Can I dynamically change CUDA_MPS_ACTIVE_THREAD_PERCENTAGE to a running MPS process? | 3 | 629 | April 11, 2026 | |
| Multi-Process Service setting CUDA_MPS_ACTIVE_THREAD_PERCENTAGE variable while application is running | 1 | 712 | May 8, 2025 | |
| Improving MPS performance using Volta MPS Execution Resource Provisioning | 5 | 1477 | July 4, 2019 | |
| MPS set_default_active_thread_percentage not working as expected | 3 | 2233 | November 23, 2021 | |
| Misunderstand about MPS non-uniform partitioning | 1 | 93 | April 11, 2026 | |
| Mutli Process Service crashes on setting up the `CUDA_MPS_ACTIVE_THREAD_PERCENTAGE` when launching a huge number of processes (say around 40~48 ) | 0 | 778 | August 11, 2023 | |
| MPS: Limiting threads to different thresholds for multi-GPU processes | 1 | 772 | October 27, 2021 | |
| Intereference between client on MPS | 0 | 64 | October 25, 2024 | |
| Can CUDA MPS limit the GPU memory usage of a client process? | 1 | 790 | May 7, 2020 | |
| General question on MPS set_active_thread_percentage | 0 | 639 | December 14, 2020 |