雙 DGX Spark (華碩 GX10) MiniMax M2.7 效能測試結果

hi all I have dual 3090 and 8 x mi50 32gb and I was tired of heat and loudness of these machines. So inspired by this post and others on nvidia forum I've purchased dual Asus GX10 (dgx spark) and I'm so happy. Each GX10 consumes about 100W during inference. Time to first token is quite high but for me it's a win Without a hassle I can run https://huggingface.co/cyankiwi/MiniMax-M2.7-AWQ-4bit/ I've used open code and hermes agent, no errors, just going - I love it! Here are my results using