NVIDIA B100 对比 NVIDIA B300 — GPU 比较 (Apr 2026)
NVIDIA B100 (192GB HBM3e, 1,750 TFLOPS FP16, Blackwell) vs NVIDIA B300 (288GB HBM3e, 2,250 TFLOPS FP16, Blackwell Ultra)
|
NVIDIA B100
192GB HBM3e · Blackwell
|
NVIDIA B300
288GB HBM3e · Blackwell Ultra
|
||
|---|---|---|---|
| 规格 | |||
| 制造商 | NVIDIA | NVIDIA | |
| 架构 | Blackwell | Blackwell Ultra | |
| VRAM | 192 GB HBM3e | 288 GB HBM3e | |
| 带宽 | 8,000 GB/s | 8,000 GB/s | |
| FP16 (Tensor) | 1750.0 TFLOPS | 2250.0 TFLOPS | |
| FP32 | 60.0 TFLOPS | 75.0 TFLOPS | |
| TDP | 700W | 1400W | |
| 发布年份 | 2024 | 2025 | |
| 细分市场 | Data center | Data center | |
| 适用场景 | AI training large-scale inference | Frontier AI training largest model workloads | |
| 云端价格 | |||
| 最便宜的按需 | — | — | |
| 最便宜的竞价 | — | — | |
| 供应商 | 0 | 1 | |
NVIDIA B100
NVIDIA B300