speculative decoding actually worth it on mini PC

[macrotinker]

I tested this tonight. I couldn't reproduce the slowdown until I pushed context much higher. I expected the model choice to matter more than it did.

[cacheline]

Tried to reproduce it just now. I couldn't reproduce the slowdown until I pushed context much higher. The surprise was that memory pressure showed up before throughput collapsed.

[fastlab]

I'd approach it in a boring order. Before rebuilding anything, I'd test one clean baseline with Qwen and compare from there. Not glamorous advice, but repeatable baselines save a lot of time.