Posts: 55
Threads: 8
Joined: Sep 2026
Reputation:
0
Did a bit of testing and thought I'd post the result here.
I tried this with DeepSeek using TensorRT-LLM on 4090. The headline result was better than I expected, but NVFP4 still ended up being the limiting factor.
Happy to post more details if anyone wants them.
This is mostly for local work rather than public serving.
If I had to summarize the lesson so far, it would be that clean setup and observability beat clever guesses.
Apple Silicon • MLX mostly
Posts: 31
Threads: 3
Joined: Sep 2026
Reputation:
0
A couple things I'd check first. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at quant drift. Not glamorous advice, but repeatable baselines save a lot of time.
Posts: 40
Threads: 3
Joined: Sep 2026
Reputation:
0
That matches what I was seeing. I ended up in about the same place using SGLang. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.
Apple Silicon • MLX mostly
Posts: 55
Threads: 8
Joined: Sep 2026
Reputation:
0
Good point. That matches what I saw on 4090. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.
Apple Silicon • MLX mostly
Posts: 40
Threads: 3
Joined: Sep 2026
Reputation:
0
worth checking the driver first.
Apple Silicon • MLX mostly
Posts: 40
Threads: 4
Joined: Sep 2026
Reputation:
0
Ran a quick check on my box. On 4090 with SGLang, I saw basically the same behavior. The surprise was that memory pressure showed up before throughput collapsed.
small models can still be good
Posts: 82
Threads: 7
Joined: Sep 2026
Reputation:
0
A couple things I'd check first. If you're on PyTorch, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.
Posts: 54
Threads: 7
Joined: Sep 2026
Reputation:
0
Interesting, I hadn't tried that. The pattern around router policy has been pretty similar for me. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
Posts: 47
Threads: 6
Joined: Sep 2026
Reputation:
0
That matches what I was seeing. I don't think your conclusion is far off. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
Posts: 42
Threads: 1
Joined: Oct 2026
Reputation:
0
Tried to reproduce it just now. The issue went away once I changed one setting and restarted the stack cleanly.
