M3 Max for local use: early impressions

[mlxgarden]

Has anyone else run into this?

I'm mostly working with DeepSeek in PyTorch on dual 3090. The part I'm trying to understand is consumer GPUs.

If you've tested something similar, what did you keep and what did you stop bothering with?

[torchshed66]

Has anyone else run into this?

Maybe, but I had the opposite result. I wouldn't optimize around that first. On my side the bigger swing came from the runtime and not the model. Could be a workload difference, but I wouldn't generalize too fast from one setup.

[pixelcrate82]

Has anyone else run into this?

Did a small comparison this afternoon. On 4090 with llama.cpp, I saw basically the same behavior. The main thing that moved was latency, not quality.

[quietarch]

Appreciate the detail. The pattern around INT4 has been pretty similar for me.

[bytesupply76]

My first guess would be this: If you're on LM Studio, I'd confirm the version first and then reduce variables until the behavior is obvious. Once you know the plain path works, then add the faster path back in piece by piece.

[dockridge]

I'd approach it in a boring order. On A6000, the easiest win is often to back off the aggressive settings and see whether the problem survives.

[plainember52]

Ran a quick check on my box. On B200 with MLX, I saw basically the same behavior. The surprise was that memory pressure showed up before throughput collapsed.

[quietarch]

same here on my side.

[turboarch95]

I'd approach it in a boring order. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at quant drift. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.