ChinaTalk·· 3 天前AI 评分31
中国的 Groq 在哪里？
AI 导读
随着 AI 市场成熟，推理解码作为内存受限操作不再仅依赖 HBM，多种硅片分别适用于训练、prefill 或 decode 阶段。当前不同硬件架构在 AI 训练与推理各环节展现出各自的优势。
正文 · 原文
In late 2025, HBM was all the rage. You need HBM for AI inference; inference decode is a memory-constrained operation, and only HBM was able to reach bandwidth speeds to serve models effectively.
But the market has matured since then, and there is no longer one architecture to rule them all. Different silicon works better for training, prefill, or decode…
Subscribe for free to continue reading
Enter your email and get every post from ChinaTalk.
来源：ChinaTalk · chinatalk.media