+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: LLM Models (https://www.datasculpting.ai/inference/forumdisplay.php?fid=12)
+--- Forum: Proprietary / Frontier (https://www.datasculpting.ai/inference/forumdisplay.php?fid=48)
+--- Thread: 70B still worth the pain (/showthread.php?tid=1468)
Pages:
1
2
Re: 70B still worth the pain - northgarden37 - 10-18-2026
Tried to reproduce it just now. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed.
Re: 70B still worth the pain - rocmpath - 10-19-2026
Thanks, that helps. I ended up in about the same place using MLX. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
Re: 70B still worth the pain - plainscope50 - 10-19-2026
worth checking the driver first.
Re: 70B still worth the pain - marbletrack - 10-19-2026
Quote:worth checking the driver first.
Ran a quick check on my box. The issue went away once I changed one setting and restarted the stack cleanly. I expected the model choice to matter more than it did.
Re: 70B still worth the pain - localpatch - 10-19-2026
What worked for me was pretty simple. On 7900 XTX, the easiest win is often to back off the aggressive settings and see whether the problem survives. Not glamorous advice, but repeatable baselines save a lot of time.
Re: 70B still worth the pain - tinycontext - 10-19-2026
Quote:Quote:worth checking the driver first.
I think the bottleneck is somewhere else. I wouldn't optimize around that first. On my side the bigger swing came from the runtime and not the model. Not saying you're wrong, just that I wouldn't lock that in as the lesson yet.