Shared app · 24 h window

2026-10-271× RTX 5090, power-capped 403 W (inference backend)

jobs completed

206 jobs · 0 failed

Setup

The shared local app fronts image and video generation on the 5090 backend (403 W cap). I pulled one 24 h slice from the job logs.

What I measured

  • Completed jobs vs failures
  • Active people in that window (distinct users, not concurrent peak)
  • Median queue wait

Numbers

206 jobs, 0 failed, 3 people, 0 s median wait for that day.

Caveats

Small user count, single day, real usage not a synthetic benchmark. A bad day would show up here — this slice happened to be clean.

What’s next

Rolling weekly summary cards so one good day is not the whole story.

Method
Aggregate counters from a single 24 h window in production logs; image + video workloads.
Caveats
One day; three people; not a load test; median wait is from the same window.

← All benchmarks