Dual 3090 setup: 400 pp t/s to 1600 pp t/s on Qwen 3.6 27B... with slightly lower tps.

First of all, my setup: Ryzen 9 5950x DDR4 3200Mhz 64gb (2x32) Dual 3090s, no NVLINK Runtime: llama.cpp Nvidia Drivers 610 Windows 11 25H2 Qwen 3.6 27B Q8 I've been using llama-server with --split-mode tensor for a couple months now, since it gave a pretty nice 10%-20% boost in overall tps, specia

中文全文翻译正在补齐，可先查看下方英文原文。

First of all, my setup: Ryzen 9 5950x DDR4 3200Mhz 64gb (2x32) Dual 3090s, no NVLINK Runtime: llama.cpp Nvidia Drivers 610 Windows 11 25H2 Qwen 3.6 27B Q8 I've been using llama-server with --split-mode tensor for a couple months now, since it gave a pretty nice 10%-20% boost in overall tps, specia