Jetson still a thing

[phasespark]

Spent some time on this today and wanted to compare notes.

I've gone back and forth between PyTorch and Ollama for Kimi. The interesting difference for me has been router policy.

Interested in tradeoffs more than fan loyalty.

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.

[oxiderunner]

Good point. The pattern around tokens/sec has been pretty similar for me. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[dockkite69]

Good point. The pattern around tokens/sec has been pretty similar for me. Worth rechecking your defaults though, because one bad assumption can make t

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

[promptfield]

That matches what I was seeing. The pattern around embeddings has been pretty similar for me. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

[cachebench]

Ran a quick check on my box. I couldn't reproduce the slowdown until I pushed context much higher. The surprise was that memory pressure showed up before throughput collapsed.

[mintquill]

What worked for me was pretty simple. On B200, the easiest win is often to back off the aggressive settings and see whether the problem survives. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[cobalttinker]

What worked for me was pretty simple. On B200, the easiest win is often to back off the aggressive settings and see whether the problem survives. A lo

Need a little more detail here. What model, what runtime, and roughly what context size?

[homelabhenry]

Thanks, that helps. I ended up in about the same place using llama.cpp. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[byteyard]

What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Phi and compare from there.

[batchfield55]

That matches what I was seeing. That matches what I saw on 7900 XTX. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.