Posts: 49
Threads: 9
Joined: Aug 2026
Reputation:
0
Spent some time on this today and wanted to compare notes.
I'm comparing two approaches here: keep it simple on 4090, or squeeze more out of the stack and deal with the complexity.
Trying to avoid rebuilding the whole thing twice.
I care more about stability than squeezing the last possible tokens/sec out of it.
Posts: 68
Threads: 9
Joined: Sep 2026
Reputation:
0
My first guess would be this: On dual 3090, the easiest win is often to back off the aggressive settings and see whether the problem survives.
Posts: 25
Threads: 4
Joined: Aug 2026
Reputation:
0
My first guess would be this: If you're on MLX, I'd confirm the version first and then reduce variables until the behavior is obvious.
small models can still be good
Posts: 38
Threads: 8
Joined: Sep 2026
Reputation:
0
Tried to reproduce it just now. The issue went away once I changed one setting and restarted the stack cleanly.
Posts: 52
Threads: 6
Joined: Sep 2026
Reputation:
0
subscribed, curious where this lands. The logs were somehow calmer than I was.
Posts: 20
Threads: 1
Joined: Sep 2026
Reputation:
0
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.
still benchmarking everything
Posts: 44
Threads: 3
Joined: Oct 2026
Reputation:
0
I'd start with the simpler stack.
small models can still be good
Posts: 35
Threads: 7
Joined: Sep 2026
Reputation:
0
Good point. I don't think your conclusion is far off. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable. Turns out 'just one more test' was a lie.
curious and caffeinated
Posts: 36
Threads: 8
Joined: Aug 2026
Reputation:
0
What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Llama and compare from there. Once you know the plain path works, then add the faster path back in piece by piece.
