+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: RAG & Knowledge Systems (https://www.datasculpting.ai/inference/forumdisplay.php?fid=14)
+--- Forum: Vector Databases (https://www.datasculpting.ai/inference/forumdisplay.php?fid=60)
+--- Thread: Vector Databases discussion on M3 Max (/showthread.php?tid=207)
Vector Databases discussion on M3 Max - emberquill - 09-09-2026
Spent some time on this today and wanted to compare notes.
I'm comparing two approaches here: keep it simple on B200, or squeeze more out of the stack and deal with the complexity.
Trying to avoid rebuilding the whole thing twice.
Re: Vector Databases discussion on M3 Max - localtinker - 09-09-2026
I don't think that's the whole story. 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.
Re: Vector Databases discussion on M3 Max - batchfield55 - 09-09-2026
I'd approach it in a boring order. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at VRAM headroom. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.
I'm happy to test one more variant if you want a second data point.
Re: Vector Databases discussion on M3 Max - fastfield4 - 09-10-2026
subscribed, curious where this lands.
If you post your exact settings, people here can probably narrow it down pretty fast.
Re: Vector Databases discussion on M3 Max - minthall - 09-10-2026
Did a small comparison this afternoon. On 4090 with llama.cpp, I saw basically the same behavior.
Re: Vector Databases discussion on M3 Max - byteyard - 09-10-2026
I think the bottleneck is somewhere else. I get why that seems right, but on dual 3090 the bottleneck felt elsewhere.
Re: Vector Databases discussion on M3 Max - amberkite - 09-10-2026
Did a small comparison this afternoon. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed.