A handy metric is needed for gauging if GPUs are being used optimally(theregister.com)

8 pointsby cyberhostanno scorso2 comments

A handy metric is needed for gauging if GPUs are being used optimally

https://www.theregister.com/2025/05/20/gpu_metric/

2 comments

MFU is probably the best but requires application logic. You can export metrics at the infra level like SM efficiency. We explain it a bit how we used it to do some optimization.

https://www.trainy.ai/blog/gpu-utilization-misleading

thundergolferanno scorso

MFU is indeed very useful. Today we found that while scaling Karpathy’s nanoGPT to multiple H100 nodes the MFU calculation itself was dropping MFU performance![1]

Commenting it out improved iter performance by almost 30%

1. https://github.com/modal-labs/multinode-training-guide/blob/...