Limiting GPU max frequency cuts power from 47W to 23W on DGX Spark cluster with minimal decode impact
MaziyarPanahi · x · 2026-08-17
A developer shows that capping GPU max frequency via sudo nvidia-smi -lgc on a 2×DGX Spark cluster running DeepSeek V4 Flash reduces power from 47W to 23W and temperature from 72°C to 61°C, with minimal impact on decode speed (memory-bound) but some impact on prefill.
More from Infra
- Score Studio enables local inference with decentralized storage — markjeffrey · 2026-08-17
- Nvidia to Invest Up to $3B in Lancium, Owner of Stargate Campus Power Infra — Beth_Kindig · 2026-08-17
- Bun's Web APIs get up to 4x faster in next version — ctjlewis · 2026-08-17
- GLM 5.3 Coming to AI Gateway: Top Score in DeepsecBench — evilrabbit_ · 2026-08-17
- Red Hat: Building a production-grade operational layer for AI agents — blaizedsouza · 2026-08-17
- Day 81: Building distributed AI pipelines with observability — blaizedsouza · 2026-08-17