HWMonitor Restores RTX 50 Hotspot Temperature Readings

NVIDIA’s RTX 50 (Blackwell) graphics cards contain a hotspot temperature sensor that was hidden from third‑party monitoring tools when the series launched. CPUID’s HWMonitor utility version 1.65 re‑exposes this data, showing both the GPU die hotspot and a second sensor for the power‑delivery VRM. The restoration lets users see temperature spikes that can indicate poor cooler contact, thermal paste issues, or thermal throttling that would be missed by the average GPU temperature reading.

Overclockers, including “Madness,” tested a shunt‑modded water‑cooled RTX 5090 drawing more than 900 W. HWMonitor reported a hotspot of about 68 °C while the average GPU temperature was ~51 °C, a 17 °C delta that helps diagnose cooling efficiency. The sensor has always existed in hardware, but NVIDIA only provided access to it via its internal MODS diagnostic software. HWMonitor’s implementation may involve reverse‑engineered or undocumented methods, though the exact technique is not disclosed. The update is expected to prompt other monitoring utilities (e.g., HWiNFO, GPU‑Z) to add similar support.