Unverified50% confidenceFactExact time
SM90 (Hopper) introduced the Tensor Memory Accelerator (TMA) to provide hardware-level asynchronous multidimensional tensor movement, requiring strict data layout alignment.
1
Sources
50%
Confidence
Long-term
Relevance
10/17/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedTensor芯片从架构设计之初就将TPU（张量处理单元）作为核心模块嵌入芯片中，围绕机器学习和AI推理任务优化，而非追求跑分极致性能76% similarVerifiedTPU (Tensor Processing Unit) is an AI-specific chip independently developed by Google since 2015.71% similarUnverifiedCompared with the previous generation Ampere's mma instruction, WGMMA supports larger matrix tiles and higher throughput, and can overlap in a pipeline with asynchronous copy mechanisms such as TMA (Tensor Memory Accelerator).71% similarUnverifiedGoogle的TPU采用脉动阵列架构，NVIDIA的Tensor Core在GPU的SM内嵌入专用矩阵运算单元，支持从FP64到FP4的多种精度格式70% similarUnverifiedA kernel hand-tuned for the NVIDIA Ampere (A100) architecture may no longer be optimal on Hopper (H100), because the new architecture introduced new memory access instructions such as the Tensor Memory Accelerator (TMA).69% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1050004API
curl https://kongchang.com/api/v1/knowledge/claims/1050004MCP
get_claim(id=1050004)