latency vs throughput for small deployments on dual 3090

[rocmpath]

Wanted to compare notes with people doing similar things.

I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling tool calling.

Not looking for the perfect answer, mostly what worked in real use.

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day. At this point the real bottleneck is my patience.

[rocmpath]

Wanted to compare notes with people doing similar things.

Did a small comparison this afternoon. My result was mixed: first run looked bad, second run looked normal. The main thing that moved was latency, not quality.

[rocmhouse90]

Did a small comparison this afternoon. The issue went away once I changed one setting and restarted the stack cleanly. The main thing that moved was latency, not quality.

I'm happy to test one more variant if you want a second data point.

[kernelrunner]

Did a small comparison this afternoon. The issue went away once I changed one setting and restarted the stack cleanly. The main thing that moved was l

I'd start with the simpler stack.

There are a few different conversations hiding inside this thread, so being explicit about the goal helps.

[kernelmike]

[quote]Did a small comparison this afternoon. The issue went away once I changed one setting and restarted the stack cleanly. The main thing that move

What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at MoE serving. Not glamorous advice, but repeatable baselines save a lot of time.

[rocmhouse90]

[quote][quote]Did a small comparison this afternoon. The issue went away once I changed one setting and restarted the stack cleanly. The main thing th

A couple things I'd check first. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at KV cache pressure. Once you know the plain path works, then add the faster path back in piece by piece.

[sparrowshed]

I'd approach it in a boring order. Before rebuilding anything, I'd test one clean baseline with Kimi and compare from there.