User achieves 100+ t/s on Qwen3.6-27B with tensor split

A Reddit user reports over 100 tokens per second on Qwen3.6-27B Q8 by using tensor split-mode across an RTX 5090 and 3090 Ti. The switch from layer split doubled throughput from ~70 t/s.