Posts: 341
Threads: 51
Joined: Sep 2026
Reputation:
0
Did a bit of testing and thought I'd post the result here.
I tried this with DeepSeek using llama.cpp on A6000. The headline result was better than I expected, but long context still ended up being the limiting factor.
Happy to post more details if anyone wants them.
This is mostly for local work rather than public serving.
measure twice, benchmark three times
Posts: 341
Threads: 51
Joined: Sep 2026
Reputation:
0
I tested this tonight. The issue went away once I changed one setting and restarted the stack cleanly. The surprise was that memory pressure showed up before throughput collapsed.
measure twice, benchmark three times
Posts: 341
Threads: 51
Joined: Sep 2026
Reputation:
0
Ran a quick check on my box. The issue went away once I changed one setting and restarted the stack cleanly. The main thing that moved was latency, not quality.
measure twice, benchmark three times
Posts: 341
Threads: 51
Joined: Sep 2026
Reputation:
0
Appreciate the detail. The pattern around quant drift has been pretty similar for me. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
measure twice, benchmark three times
Posts: 50
Threads: 2
Joined: Sep 2026
Reputation:
0
subscribed, curious where this lands.
learning in public
Posts: 341
Threads: 51
Joined: Sep 2026
Reputation:
0
A couple questions before I guess too hard. Is this local-only or are you routing anything externally?
measure twice, benchmark three times
Posts: 35
Threads: 3
Joined: Sep 2026
Reputation:
0
My first guess would be this: If you're on SGLang, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.
Posts: 341
Threads: 51
Joined: Sep 2026
Reputation:
0
Quote:My first guess would be this: If you're on SGLang, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of the
I don't think that's the whole story. I've seen people blame consumer GPUs when the real issue was much more mundane. Worth testing the simpler explanation before adopting the more interesting one.
measure twice, benchmark three times
Posts: 30
Threads: 4
Joined: Sep 2026
Reputation:
0
My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at speculative decoding. Once you know the plain path works, then add the faster path back in piece by piece.
Posts: 341
Threads: 51
Joined: Sep 2026
Reputation:
0
I tested this tonight. My result was mixed: first run looked bad, second run looked normal. I expected the model choice to matter more than it did.
measure twice, benchmark three times
