RTX 5080 (PCIe Gen 4) + Ryzen 3900X
8 records across 2 models (quants: UD-IQ4_XS, UD-Q4_K_M).
| hardware | model | quant | backend | tok/s | pp tok/s | W | ttft s | ctx/tg | date | class | flags | source |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| RTX 5080 (PCIe Gen 4) + Ryzen 3900X | Qwen3-Coder-Next | UD-IQ4_XS | llama.cpp (CUDA) | 50 | – | – | – | 256000 | 2026-09-03 | community | – | source |
| RTX 5080 (PCIe Gen 4) + Ryzen 3900X | Qwen3-Coder-Next | UD-IQ4_XS | llama.cpp (CUDA) | 30 | – | – | – | 256000 | 2026-09-03 | community | – | source |
| RTX 5080 (PCIe Gen 4) + Ryzen 3900X | Qwen3-Coder-Next | UD-IQ4_XS | llama.cpp (CUDA) | 27 | – | – | – | 256000 | 2026-10-05 | community | – | source |
| RTX 5080 (PCIe Gen 4) + Ryzen 3900X | Qwen3-Coder-Next | UD-IQ4_XS | llama.cpp (CUDA) | 17 | – | – | – | 256000 | 2026-10-05 | community | – | source |
| RTX 5080 (PCIe Gen 4) + Ryzen 3900X | Qwen3.6-35B-A3B 35B-A3B | UD-Q4_K_M | llama.cpp (CUDA) | 80 | – | – | – | 256000 | 2026-09-03 | community | – | source |
| RTX 5080 (PCIe Gen 4) + Ryzen 3900X | Qwen3.6-35B-A3B 35B-A3B | UD-Q4_K_M | llama.cpp (CUDA) | 66 | – | – | – | 256000 | 2026-10-05 | community | – | source |
| RTX 5080 (PCIe Gen 4) + Ryzen 3900X | Qwen3.6-35B-A3B 35B-A3B | UD-Q4_K_M | llama.cpp (CUDA) | 50 | – | – | – | 256000 | 2026-09-03 | community | – | source |
| RTX 5080 (PCIe Gen 4) + Ryzen 3900X | Qwen3.6-35B-A3B 35B-A3B | UD-Q4_K_M | llama.cpp (CUDA) | 40 | – | – | – | 256000 | 2026-10-05 | community | – | source |
All rows cite their source. Provenance: community 40018 sourced 1085