I really like topics about optimization like this one. Great job, OP!

Looks absolutely stellar!

Another step forward for local AI / local LLM inferencing on (very!) resource-constrained consumer GPU hardware!