A whole bunch more comparison numbers in this section: <a href="https://www.macstories.net/stories/m5-ultra-mac-studio-review-the-dream-mac-for-local-ai-agents/#mx-pc" rel="nofollow">https://www.macstories.net/stories/m5-ultra-mac-studio-revie...
A basic llama-bench on Qwen 3.8 27B UD-Q4_K_M gives pp512 3920 tok/s / tg128 81 tok/s on a 500W RTX PRO 6000 (should be similar speeds to a 5090, chip is basically the same, just less VRAM). With MTP3 this is 140 tok/s on mtp-bench.
This is with llama.cpp. You can of course use vLLM/SGLang well on these cards and they're even faster. On vLLM w/ NVIDIA/Qwen3.8-27B-NVFP4 baseline has a prefill of about 13,000 tok/s. The baseline tok/s is 72 tok/s, but at mtp7, it's 157 tok/s, and w/ dflash7 that goes up to 215 tok/s. On mtp-bench, DFlash2 gets a hair under 300 tok/s w/ the code_python prompt.
simonw · · focus · HN ↗
lhl · · focus · HN ↗
This is with llama.cpp. You can of course use vLLM/SGLang well on these cards and they're even faster. On vLLM w/ NVIDIA/Qwen3.8-27B-NVFP4 baseline has a prefill of about 13,000 tok/s. The baseline tok/s is 72 tok/s, but at mtp7, it's 157 tok/s, and w/ dflash7 that goes up to 215 tok/s. On mtp-bench, DFlash2 gets a hair under 300 tok/s w/ the code_python prompt.