when RAG was overkill

[homelabhenry]

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

One thing I keep noticing is that people underestimate batch size until they actually try to run this stuff for more than one afternoon.

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.

[quantshed]

What exact setup are you running? Are you testing this on dual 3090 or something else?

[sanescope]

What exact setup are you running? Are you testing this on dual 3090 or something else?

I don't think that's the whole story. I get why that seems right, but on 5090 the bottleneck felt elsewhere. Could be a workload difference, but I wouldn't generalize too fast from one setup.

[southspark]

What exact setup are you running? Are you testing this on dual 3090 or something else?

I tested this tonight. The issue went away once I changed one setting and restarted the stack cleanly.

[deepseeked]

I'd approach it in a boring order. If you're on MLX, I'd confirm the version first and then reduce variables until the behavior is obvious. Turns out 'just one more test' was a lie.

[fastbench]

following this because I'm in the same boat.

[plainscope50]

Interesting, I hadn't tried that. I ended up in about the same place using llama.cpp. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

[tensorkite]

Interesting, I hadn't tried that. I ended up in about the same place using llama.cpp. I got better results once I stopped chasing one metric and just

I tested this tonight. On 7900 XTX with SGLang, I saw basically the same behavior.

[quiettoken]

Need a little more detail here. Are you testing this on 4090 or something else?

What exact setup are you running? Do you see the same thing with a smaller batch or shorter prompt?

[papermixer4]

Interesting, I hadn't tried that. I ended up in about the same place using vLLM.