特斯拉工程師爆料:Nvidia H100 故障率 9%?!根本不配上機器人! #特斯拉 #Nvidia #H100 #AI #AI晶片 #機器人 #Optimus #馬斯克 #護城河
大鱼聊电动·2025-11-11 11:43

Chip Design Philosophy - Tesla's senior engineer emphasized that designing chips for robots is 100 times more difficult than designing for data centers [1] - Nvidia's H100 chip is likened to being in a "hot tub," suitable for controlled environments, while Tesla requires chips for "battlefield" conditions [1] - The industry average Annual Failure Rate (AFR) for H100 chips is 9% [1] Application and Reliability - Data centers can use hot-swapping to replace failed chips, but robots require chips that can withstand outdoor conditions and cannot fail [1] - Tesla needs chips that can operate reliably in harsh environments, unlike the H100 which is designed for data centers [1] Strategic Implications - Tesla is developing its own chips because Nvidia's products are not suitable for its robotics applications [1] - Tesla's chip development is considered a "deadly weapon" by Musk [1]

特斯拉工程師爆料:Nvidia H100 故障率 9%?!根本不配上機器人! #特斯拉 #Nvidia #H100 #AI #AI晶片 #機器人 #Optimus #馬斯克 #護城河 - Reportify