DeepSeek V4 Pro 1.6T · GPU 对比

DeepSeek V4 Pro 1.6T — B300 vs GB300 NVL72

B300NVIDIA Blackwell)与 GB300 NVL72NVIDIA Blackwell)在 DeepSeek V4 Pro 1.6T 上的正面 AI 推理基准测试对比。涵盖各类 LLM 工作负载的延迟、吞吐量与成本。使用下方图表控件切换序列、精度和指标——交互方式与主推理图表相同。

B300 / GB300 NVL72 在 DeepSeek V4 Pro 1.6T 上以 72 tok/s/user 运行:6234 / 9384 tok/s/GPU,$0.10 / $0.08 每百万 token。GB300 NVL72 每 token 成本低 31%;GB300 NVL72 每 GPU 吞吐量高出 51%。

在 15–244 tok/s/user 交互性区间的中部,即 DeepSeek V4 Pro 1.6T 上以 130 tok/s/user 运行时:B300 达到 1365 tok/s/GPU($0.48/百万 token),GB300 NVL72 达到 3474($0.22/百万)。GB300 NVL72 每 token 成本低 124%;GB300 NVL72 每 GPU 吞吐量高出 154%。

以 187 tok/s/user 为目标在 DeepSeek V4 Pro 1.6T 上运行时,B300 产出 518 tok/s/GPU(每百万 token $1.24),GB300 NVL72 产出 502($1.44)。B300 每 token 成本低 16%;B300 每 GPU 吞吐量高出 3%。 (数据反映此 URL 的默认 8k/1k · fp4 选择——如果您在控件中更改序列、精度或模型,下方表格和图表会自动更新。)

查看每美元性能对比 →

Interpolated from real benchmark data. Edit target interactivity values below to compare at different operating points.
Metric
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Interactivity (tok/s/user)
Throughput (tok/s/gpu)
B300:6234.0GB300 NVL72:9384.4
B300:1365.5GB300 NVL72:3473.9
B300:518.0GB300 NVL72:501.8
Cost ($/M tok)
B300:$0.103GB300 NVL72:$0.078
B300:$0.485GB300 NVL72:$0.216
B300:$1.242GB300 NVL72:$1.436
tok/s/MW
B300:3281057GB300 NVL72:4426587
B300:718681GB300 NVL72:1638613
B300:272625GB300 NVL72:236721
Concurrency
B300:~315GB300 NVL72:~1163
B300:~36GB300 NVL72:~288
B300:~1GB300 NVL72:~13

推理性能

不同模型、硬件配置和服务参数下的推理性能指标。

厂商:
聚合模式:
投机解码: