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
Same throughput with much less heat. Not sure what that extra 175w is going towards but it's diminishing returns.
Cap the power to 400W you won’t see much impact
Same throughput with much less heat. Not sure what that extra 175w is going towards but it's diminishing returns.