Unverified50% confidenceFactExact time
NVIDIA began natively supporting FP8 computation starting with the Hopper architecture (H100).
1
Sources
50%
Confidence
Long-term
Relevance
9/2/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedNative FP8 hardware acceleration was first implemented in the fourth-generation Tensor Cores of the NVIDIA Hopper architecture (H100, 2022), with a peak throughput of 3958 TFLOPS.80% similarUnverifiedThe NVIDIA H100 is a data center GPU based on the Hopper architecture, with its SXM5 version having a TDP as high as 700W.79% similarUnverifiedNVIDIA's H100 GPU natively supports the FP8 format, with throughput approximately 2× that of FP16.78% similarVerifiedNVIDIA's MIG technology enables hardware-level resource isolation on high-end GPUs such as the A100/H100.76% similarVerifiedThe NVIDIA Hopper architecture (H100/H200) introduced FP8 Tensor Cores for the first time and added the Transformer Engine to automatically manage precision conversion.75% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/843034API
curl https://kongchang.com/api/v1/knowledge/claims/843034MCP
get_claim(id=843034)