AITOPAI热报
PapersMulti-source

斯坦福研究：NVIDIA B200 每单位能源产出为 AMD MI350X 的 9 倍以上

rohanpaul_ai@rohanpaul_ai
Post· 5 sources
View the post
Why we picked it

斯坦福拿同一个 Qwen 模型在 4 块芯片上跑了 220 个真实任务，B200 的能源效率是 MI350X 的 9 倍多，买卡前值得看看这组数据。

斯坦福研究者在 220 个真实专业任务上，用同一模型 Qwen3.6-27B 在 4 种单芯片上运行，测量每单位能源产出的有用工作量。结果显示 NVIDIA B200 相比 AMD MI350X 每美元价值超过 4 倍，每单位能源节省的专家时间超过 9 倍。研究说明在模型不变的情况下，芯片本身就能决定电力预算带来的经济回报。

Original · rohanpaul_ai

Stanford's new research: The NVIDIA B200 delivered over 4x the dollar value and over 9x the expert time saved per unit of energy compared with the AMD MI350X.

i.e. chip alone, with the model unchanged, can decide how much economic return you get from your power budget.

Here, Stanford researchers ran the same AI model, Qwen3.6-27B, on 4 different single AI chips across 220 real professional tasks and measured how much useful work each produced for the energy it used.

#NVIDIA B200#AMD MI350X#Qwen3.6-27B#斯坦福#能耗基准

Keep reading

Recent stories with the same tag or section
PapersBuildersSelected01:24
VFlash Parallel Vision Model Speeds Robot Inference 5.1x

The counterintuitive result is that skipping autoregressive decoding improves accuracy while speeding it up, since parallel output avoids compounding coordinate errors. Robotics and edge-AI developers working on Jetson-class hardware can use the open-source release to cut inference latency for manipulation and trajectory tasks. The parallel-decoding idea may also transfer to other fixed-structure prediction problems beyond robotics.

#VFlash#Physical AI#robotics#Jetson Thor
Jim Fan@DrJimFanSource