SemiAnalysis + Asianometry on Intel, Gemini 2.5, and Chinese Robots
ChinaTalk
00:00
Evaluating GPU Performance: H20 vs H100
This chapter focuses on a technical analysis of NVIDIA's H20 and H100 GPU models, highlighting their performance metrics such as memory bandwidth and FLOPS. It also discusses the challenges faced by hyperscalers with legacy systems and the implications for GPU efficiency in real-world workloads.
Transcript
Play full episode
