Event ·Multi-source

NVIDIA Shruti Koparkar 谈 AI 工厂能效经济与 CUDA 生态

NVIDIA 的 Shruti Koparkar 在访谈中解释了 AI 基础设施的能效经济，她会优先测试每瓦性能（performance per watt）。她说明了 CUDA 与 CUDA-X 在 AI 工厂软件栈中的实际作用。访谈还提到 Pinterest 用同一套 GPU 集群先后跑推荐系统、语言模型和多模态视觉语言模型。她讲了新模型发布时的负载变化，以及何时保留、改造或替换旧 GPU 的判断标准。

Current read

NVIDIA 的人聊怎么把每瓦电榨出最多算力，还讲了 Pinterest 一套 GPU 换三种模型跑，搞基础设施的可以看看。

4 sources47° AI heatLast updated 2026/11/3 15:27:32

Evidence

4 sources

Primary sourcekimmonismus

2026/11/3 15:27:32View source

Related source 1arXiv cs.AI

2026/11/2 18:19:19View source

Related source 2Fireworks AI

2026/11/3 17:00:45View source

Related source 3PyTorch

2026/11/3 18:00:20View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story