Engineering notes.
Short technical notes from arflo on thermal physics, liquid cooling, chip telemetry and AI-based control of data center cooling.
Chip telemetry as an input for cooling control.
Servers already report temperatures, power draw and utilization for their own protection. This note looks at what that telemetry contains, how it can inform facility cooling and what has to be true before a controller relies on it.
Read note · 4 MINSetpoint control in direct-to-chip liquid cooling.
A direct-to-chip liquid cooling loop has two main control inputs: coolant supply temperature and flow. This note explains what each one does, which limits apply and why the best setting changes with load.
Read note · 5 MINGPU load transients and cooling control.
GPU clusters change their heat output faster than most cooling loops can follow. This note explains where the delay comes from and how feedforward control uses the load signal to close the gap.
Read note · 5 MINHow AI optimizes data center cooling.
AI data center cooling optimization means adjusting cooling setpoints continuously to the actual heat load. This note explains what gets optimized, which methods exist and what to check before one of them touches a live system.
Read note · 6 MINFrom model to control loop.
Between a plausible prediction and a change to cooling lie practical integration decisions. They are part of the engineering.
Read note · 3 MINWhat makes a fair cooling comparison.
Lower electricity consumption becomes meaningful when workload, system boundaries and thermal conditions support the comparison.
Read note · 4 MINWhy room temperature does not describe the chip.
Between a workload and room temperature lie materials, thermal interfaces and time. That path matters when you want precise control.
Read note · 3 MIN