Ambient Temperature and Watt Benchmark Drift
Why we log inlet air every minute, and what happens when your farm runs hotter than our lab.
Every watt benchmark we publish includes a temperature column. Here is why that number matters more than most operators expect.
The Thermal Throttle Curve
Modern ASICs reduce clock frequency when junction temperature exceeds a firmware-defined threshold — typically between 75°C and 85°C depending on the model. When frequency drops, hashrate falls but fans keep spinning at full speed. The result: more watts per terahash, sometimes dramatically.
In our lab we target 24–26°C inlet air. At a client farm in Gumi last winter, we measured the same S19 Pro model at 31°C inlet during an on-site assessment. J/TH worsened from 27.1 (lab) to 29.8 (field) — a 10% efficiency penalty purely from ambient conditions.
What We Log
During a standard 4-hour run we record:
- Inlet air temperature at the unit’s intake grille
- Exhaust air temperature at the rear
- Chip junction temperature via firmware API where accessible
- Fan RPM at 15-minute intervals
If inlet temperature drifts more than 2°C during the run, we note it in the report and may extend the test window.
Practical Takeaway
Before comparing our lab numbers to your farm’s performance, adjust for the temperature delta. A rough rule: every 5°C above our test conditions adds approximately 3–5% to your real-world J/TH. For precise planning, book an on-site power assessment at your actual facility.