The setup The starting line was 43 tokens per second decode on vanilla llama.cpp. The...
Three Months of Speed-Up Experiments on a 3090 Ti: Autoregressive DFlash MTP for Qwen3.6-27B
Related
I Built a Voice AI Tutor in 200 Lines of Code (and Zero Backend)
Day 35 of TechFromZero. Every voice assistant you've ever used is the same three Lego bricks. Let's snap them together in a single afternoon — using only free, browser-native APIs.
Building llama.cpp from source on a Dell Precision T5820 with an RTX 3090 Ti (after seven power cycles)
I pulled a Quadro M4000 out of a used Dell Precision T5820, dropped in an RTX 3090 Ti, and turned the...
Three Months of Speed-Up Experiments on a 3090 Ti: Autoregressive DFlash MTP for Qwen3.6-27B
The setup The starting line was 43 tokens per second decode on vanilla llama.cpp. The...