Don't Assume Your Cloud T4 Runs at Spec: Gemma 4 Decode Falls 7% as the Card Warms

Post · https://dev.to/gde/dont-assume-your-cloud-t4-runs-at-spec-gemma-4-decode-falls-7-as-the-card-warms-3fd3

The same Tesla T4 on the same Google Cloud VM decoded Gemma 4 E2B at 115 tokens/s at 45 °C and 107 at 77 °C. The card sits at its 70 W limit, a hotter chip draws more power at the same clock, so the clock drops. No thermal throttle fires, and no spec sheet mentions it.

First seen by Oriel on Oct 17, 2026, 18:23 UTC.

Metric trends

Last 30 days · Up to 120 recorded samples per source and metric · First and latest retained

Reactions

Total public reactions reported by the source; distinct from points and likes.

DEV Community: 10, latest observation Oct 20, 2026, 08:39 UTC; +1 over observed period since Oct 17, 2026, 18:23 UTC

Comments

Comment count reported by the source.

DEV Community: 4, latest observation Oct 20, 2026, 08:39 UTC; +2 over observed period since Oct 17, 2026, 18:23 UTC

Aliases

Source	External ID	URL
DEV Community	4863183	https://dev.to/gde/dont-assume-your-cloud-t4-runs-at-spec-gemma-4-decode-falls-7-as-the-card-warms-3fd3