Causes
- Triple-fan cooling system obstructed by dust buildup on the intake vents or the bottom-mounted air intake mesh — a common issue after several weeks of operation in desktop environments
- Default power profile in the BIOS sets the EVO-X3's peak TDP to 140W without adequate fan curve tuning for sustained AI/LLM workloads that keep the CPU/GPU saturated for extended periods
- Insufficient clearance around the chassis — the side and rear exhaust vents require at least 10 cm of space on all sides; placing the unit in an enclosed cabinet or against a wall traps hot air
- Thermal paste pump-out effect on the Strix Halo APU after repeated thermal cycling, reducing heat transfer efficiency between the die and the vapor chamber cold plate
- Aggressive AMD PBO (Precision Boost Overdrive) settings in the BIOS or Ryzen Master causing the APU to draw more current than the cooling system can dissipate in sustained loads
How to Fix
- Shut down the EVO-X3, unplug the power adapter, and inspect the bottom intake mesh and all exhaust vents. Use a can of compressed air or a low-speed vacuum to remove any accumulated dust. Pay special attention to the fan blades visible through the top ventilation grille.
- Relocate the EVO-X3 to an open area with at least 10 cm (4 inches) of clearance on all sides. Avoid placing it inside a desk cubby, entertainment center, or directly against a wall. If stacking with other equipment, leave space between units.
- Enter the BIOS by pressing Del during boot. Navigate to Advanced → AMD Overclocking → Precision Boost Overdrive and set it to Auto or Disabled instead of "Enabled" or "Advanced." This prevents the APU from exceeding its 140W thermal budget under sustained load.
- In Windows, open AMD Software: Adrenalin Edition → Performance → Tuning. Under GPU Tuning, set a manual fan curve that reaches 80% speed at 75°C and 100% at 85°C. Under Power Tuning, reduce the max frequency by 200 MHz or set a power limit of −10% to maintain stable boost clocks without throttling.
- Download HWMonitor or HWiNFO64 and run a 15-minute stress test (Cinebench R23 for CPU, FurMark for GPU). Monitor the CPU/GPU hotspot temperatures. If the hotspot exceeds 105°C and clock speeds drop below the base frequency, thermal throttling is active.
- If throttling persists after the above steps, consider repasting the APU. The EVO-X3 uses a vapor chamber cooler with factory thermal paste — carefully remove the bottom cover and the cooling assembly, clean the old paste with isopropyl alcohol, and apply a high-performance paste such as Honeywell PTM7950 or Thermal Grizzly Kryonaut Extreme.
- For users running Linux, install power-profiles-daemon and switch to the power-saver profile during sustained workloads, or configure amd-pstate with a lower frequency limit using
cpupower frequency-set -u 3500000to cap the boost clock.