My 5090 runs at full TDP(pretty much exactly 575W) when running inference through LM Studio.
Cap the power to 400W you won’t see much impact
alt Hacker News
Cap the power to 400W you won’t see much impact