Unverified50% confidenceFactExact time
NVIDIA Ampere 架构引入了异步内存拷贝和稀疏矩阵加速单元，Hopper 架构引入了 Transformer Engine 专用硬件
1
Sources
50%
Confidence
Long-term
Relevance
7/10/2026
First Seen
Sources
Related Claims
VerifiedNVIDIA Hopper架构（H100/H200）首次加入FP8 Tensor Core，并引入Transformer Engine自动管理精度转换75% similarUnverifiedNVIDIA GPU从Turing架构开始引入INT8张量核心，Ampere架构进一步增强了整数运算能力75% similarVerifiedNVIDIA从Volta架构开始引入Tensor Core，这是一种专门用于矩阵乘加运算的硬件单元74% similarUnverifiedNVIDIA在2020年后推出A100、H100系列GPU，其张量核心的矩阵乘法加速设计契合Transformer workload72% similarUnverifiedH100 GPU基于Hopper架构，其Transformer引擎能在FP8和FP16精度间动态切换，将吞吐量提升至A100的约2-3倍71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/425086API
curl https://kongchang.com/api/v1/knowledge/claims/425086MCP
get_claim(id=425086)