Posting this in case it saves someone else a little time.

I'm not trying to build a perfect architecture diagram. I just want a sane starting point around MLX and 7900 XTX.

I'm especially interested in mistakes worth avoiding.

Thought I'd post a small recipe because a lot of threads ask similar things.

Maybe useful as a starting point: MLX on Threadripper box, keeping quant drift in mind from the start.

Feel free to tear it apart if you see something dumb.

I care more about stability than squeezing the last possible tokens/sec out of it.

Did a bit of testing and thought I'd post the result here.

I tried this with Kimi using vLLM on GB200. The headline result was better than I expected, but VRAM headroom still ended up being the limiting factor.

Happy to post more details if anyone wants them.

Thought I'd post a small recipe because a lot of threads ask similar things.

Just sharing a setup that ended up working reasonably well: Mistral, PyTorch, A6000, with a lot of attention on consumer GPUs.

Maybe useful as a starting point for newer folks.

I care more about stability than squeezing the last possible tokens/sec out of it.

This keeps coming up in conversation so I figured I'd start a thread.

The more I work with local setups, the more it seems like tokens/sec decides everything downstream.

Curious whether that lines up with what other people are seeing.

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.

Did a bit of testing and thought I'd post the result here.

After a few runs on mini PC, I ended up with decent performance, but the setup only really made sense once I adjusted around tail latency.

Curious if anyone else landed in the same place.

Thought I'd post a small recipe because a lot of threads ask similar things.

Maybe useful as a starting point: TensorRT-LLM on 7900 XTX, keeping MoE serving in mind from the start.

If someone wants the exact settings I can clean them up and post them.

I care more about stability than squeezing the last possible tokens/sec out of it.

Did a bit of testing and thought I'd post the result here.

Short version: PyTorch behaved well for me, but only after I stopped ignoring VRAM headroom.

Happy to post more details if anyone wants them.

Did a bit of testing and thought I'd post the result here.

I tried this with Mixtral using SGLang on GB200. The headline result was better than I expected, but quant drift still ended up being the limiting factor.

Curious if anyone else landed in the same place.

This keeps coming up in conversation so I figured I'd start a thread.

It feels like the conversation has shifted hard toward inference and serving details lately, and honestly that makes sense once you start dealing with INT4.

Maybe obvious to some of you, but it clicked for me this week.

I expected the biggest difference to come from the model, but the stack around it mattered at least as much.