Tag
The author shares six months of measurements on a four-RTX 3090 local setup, revealing that data parallelism often outperforms tensor parallelism for models fitting on fewer cards, with up to 3.4x throughput difference.
User tests ternary quantized Qwen3.6 27B on an RTX 3090, achieving 60 tk/s with two slots and 100k KV cache using 21GB VRAM, with good quality and stable tool calls.
User benchmarks Qwen3.6-27B on an RTX 3090 using llama.cpp, achieving 35 tok/s generation and 1247 tok/s prompt processing.