Execution timeline
Median run · hover to scrub all charts · click a node to focus
Time by stage
Click a stage to filter the views
VRAM
Whole device · GB
GPU
Nodes
Configurations
Same workflow, same inputs, new seed per run · click to switch
Synthetic numbers for previewing the layout. Run python -m perfmon bench on your GPU for real measurements.
Median run · hover to scrub all charts · click a node to focus
Click a stage to filter the views
Whole device · GB
Same workflow, same inputs, new seed per run · click to switch