RTX 3090 Inference Power Test: Performance Not Linear with Power Draw
haluxa · reddit · 2026-08-30
A user conducted an informal power consumption test for inference on an RTX 3090 using LMStudio and the Qwen3.8-27B-Q4KM model (20GB VRAM). Results indicate that inference speed is not linearly correlated with power draw, suggesting an optimal power range exists. The previously assumed 220W sweet spot may vary depending on the inference engine and model.
More from Infra
- Copper to optical: optical backplanes seen as necessary as racks move to optics — BenBajarin · 2026-10-16
- Polymarket prices 28% odds a US state enacts data center moratorium by year-end — Polymarket · 2026-10-16
- Pennsylvania residents reject $10,000-per-household offer for massive Amazon data center — Polymarket · 2026-10-16
- PyTorch now runs natively on AWS Trainium with zero code changes — PyTorch · 2026-10-16
- Chinese labs run on ~1/10 US compute; ByteDance pays 2-4x to hoover up SEA supply — robleclerc · 2026-10-16
- Zymtrace raises $12.2M and hires ML engineers to build agents that fix GPU bottlenecks — cfregly · 2026-10-16