ChinaTalk· Aqib Zakaria·· 2 小時前AI 評分19
中國的 Groq 在哪裏？
Where is China’s Groq?
AI 導讀
AI 推理市場已不再由單一架構主導，不同晶片各自更適合訓練、預填充或解碼。2025 年底，HBM 曾被視為有效支援模型推理所必需，因解碼受記憶體限制，當時只有 HBM 能提供足夠頻寬；市場其後逐漸成熟。
正文
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.
來源：ChinaTalk · chinatalk.media