放棄核心數堆疊！輝達 Vera 轉向單晶片設計，目標打造代理式 AI 低延遲新標準
TechNews (TW)zh-TW
輝達（NVIDIA）Vera 團隊近日表示，過去業界追求以晶粒化設計（chiplet design）堆疊核心數 […]
This is a short summary published by AI Global Wire. The full article is owned and hosted by TechNews (TW) — open it there to read it in full.
Read the full story at TechNews (TW)Related AI news
- Anthropic says it will commit 5% of its shares to 'public benefit fund'BBC Technology · October 19, 2026
- AI 加持 晶圓切割機廠 DISCO 傳業績優、營益創新TechNews (TW) · October 19, 2026
- Australia’s Nextdc flags earnings risk over power delayTech in Asia · October 19, 2026
- DeepSeek’s leaner AI model hits Samsung, SK hynix sharesTech in Asia · October 19, 2026
- China's growth accelerates as AI export boom outweighs domestic weaknessBBC Technology · October 19, 2026
- iOS 27「電話」App 重大更新，三大實用功能與 AI 新亮點一次掌握TechNews (TW) · October 19, 2026