GPU Watchdog

reading device…
live
▲

—
Resident in VRAM
nothing loaded
Recent model swaps
none
Temperature
—°C
slowdown — · shutdown —
Utilisation
—%
SM clock —
VRAM
—MiB
of — MiB
Power
—W
no model resident

Thermal & loadlast 30 min · live

temp °C util % throttled

Usage by appattributed from container source IP

Apps

AppReqsOut tok Avgp95tok/sGPU s

Models

ModelReqsOut tok tok/sCold starts

Recent requestsnewest first

WhenAppModelEndpoint DurationInOut tok/sTemp
waiting for the first request through the proxy…