Verified75% confidenceFactExact time
Blackwell's second-generation Transformer Engine adds FP4 precision support, which in theory can improve Transformer model training throughput by 2–4x compared to the H100.
3
Sources
75%
Confidence
Long-term
Relevance
8/8/2026
First Seen
Sources
手握GPU集群却不知做什么？AI项目选题实战指南
Related Claims
UnverifiedB200采用Blackwell架构，引入第二代Transformer Engine和FP4精度支持，推理吞吐量较H100提升最高30倍80% similarUnverifiedFlashAttention-2 has been measured to increase Transformer training speed by 2–4×76% similarUnverifiedNVIDIA H100 的 Transformer Engine 将 FP8 训练和推理的吞吐量相比上一代提升近 4 倍74% similarVerifiedCore innovations of the Blackwell architecture include the second-generation Transformer Engine (supporting FP4 precision), the NVLink Switch system (supporting interconnection of 576 GPUs), 20 petaFLOPS of FP4 compute, and 192 GB HBM3e memory (with bandwidth reaching 8 TB/s).72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/707644API
curl https://kongchang.com/api/v1/knowledge/claims/707644MCP
get_claim(id=707644)