100+ t/s on Qwen3.6-27B across RTX 5090 + 3090 Ti

A Reddit user reports ~100 tokens/second on Qwen3.6-27B at Q8 quantization using an RTX 5090 and RTX 3090 Ti in tensor split mode. Switching from layer split to tensor split boosted performance from ~70 to 100+ t/s.