在行业标准推理引擎vLLM上的测试显示，NVIDIA GB300 NVL72的实测端到端性能已达GB200 NVL72的2.7倍。尽管其纸面参数仅显示NVFP4算力提升约1.5倍、HBM容量增加1.5倍且带宽相同，但在大多数服务商实际运行的中段负载区间，凭借全栈优化的复合增益，GB300实现了远超理论算力提升的性能飞跃。此次测试基于NVIDIA、Inferact和CoreWeave为开源项目提供的临时GB300系统完成，结果印证了端到端实测性能才是衡量硬件效能的黄金标准，而非单纯的纸面理论算力。
纸面 FP4 算力只多 50% 的 GB300，实际推理却快了 2.7 倍，全栈优化的复合增益比参数表好看太多，做推理服务的该重新算算 TCO 了。
MINECRAFT STEVE ALERT: GB300 ultra NVL72 is already 2.7x faster 🚀 than GB200 NVL72 on one of the industry standard inference engine known as @vllm_project. On paper, GB300 only has ~1.5x faster NVFP4 FLOP & 1.5x more HBM capacity & same HBM BW than GB200 but due to the full stack optimization with compounding gains, in the middle of the curve where most providers serve at, GB300 is up to 2.7x faster. End to End performance is the gold standard of performance, not on paper theoretical flops.
Thanks to the 10x engineers at NVIDIA & @inferact & @coreweave for this temporary gb300 for open source projects!
来源：SemiAnalysis · x.com