+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: Inference Engines (https://www.datasculpting.ai/inference/forumdisplay.php?fid=9)
+--- Forum: Other Engines (https://www.datasculpting.ai/inference/forumdisplay.php?fid=39)
+--- Thread: throughput per watt reality check (/showthread.php?tid=1040)
Pages:
1
2
throughput per watt reality check - signalcraft14 - 10-12-2026
Wanted to compare notes with people doing similar things.
I'm mostly working with Mixtral in MLX on B200. The part I'm trying to understand is tokens/sec.
Would appreciate practical advice rather than theory.
I care more about stability than squeezing the last possible tokens/sec out of it.
Re: throughput per watt reality check - bytestack - 10-12-2026
Tried to reproduce it just now. The issue went away once I changed one setting and restarted the stack cleanly.
Re: throughput per watt reality check - fastfield4 - 10-15-2026
Maybe, but I had the opposite result. The reported numbers can look compelling, but in ordinary usage I didn't see the same benefit. Worth testing the simpler explanation before adopting the more interesting one.
I'm happy to test one more variant if you want a second data point.
Re: throughput per watt reality check - bytestack - 10-15-2026
I'd start with the simpler stack.
Re: throughput per watt reality check - cobalttinker - 10-15-2026
same here on my side.
Re: throughput per watt reality check - edgekite - 10-16-2026
My first guess would be this: On 4090, the easiest win is often to back off the aggressive settings and see whether the problem survives.
Re: throughput per watt reality check - relaysupply90 - 10-16-2026
A couple things I'd check first. Before rebuilding anything, I'd test one clean baseline with Mistral and compare from there. Once you know the plain path works, then add the faster path back in piece by piece.
Re: throughput per watt reality check - pixelcrate82 - 10-16-2026
Did a small comparison this afternoon. On B200 with SGLang, I saw basically the same behavior. I expected the model choice to matter more than it did.
Re: throughput per watt reality check - fusedhall56 - 10-17-2026
My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at small local stacks. Not glamorous advice, but repeatable baselines save a lot of time.
Re: throughput per watt reality check - deepseeked - 10-17-2026
Quote:My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at small local stacks. Not glamo
A couple things I'd check first. Before rebuilding anything, I'd test one clean baseline with FLUX and compare from there.