# NVIDIA B200

- 产品 ID: `NV-B200`
- 类目: 计算加速 / 数据中心 GPU
- 品牌: NVIDIA (`BR-NVIDIA`)
- 系列: 数据中心加速卡 (`SER-NV-DC`)
- 层级: 品牌 → 系列 → 型号（series.brand_id 必须等于 brand_id）
- 可信级: verified
- 来源: https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/
- 更新时间: 2026-11-11 20:20:00

## 参数（带口径与来源）

| 字段 | 值 | 口径 | 可信级 | 来源 | 时间 |
|---|---|---|---|---|---|
| 显存容量 | 192 GB | Table 2 行「Max HBM capacity」，Blackwell 列；HBM3E | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| 显存代际 | HBM3e | Table 2 行「Max HBM capacity」 | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| 显存带宽 | 8 TB/s | Table 2 行「Max HBM bandwidth」，Blackwell 列，页面原文 8 TB/s | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| 整卡功耗 | 1200 W | Table 2 行「Max power (TGP)」，Blackwell 列，为 Blackwell 全系上限（非 HGX B200 SXM 单卡典型值） | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| PCIe 代际 | Gen5 | Table 1 行「PCIe Interface」，Blackwell GPU 列 128 (Gen 5)，单位 GB/s 双向 | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| 互联形态 | NVLink 5 | 第五代 NVLink | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| NVLink 带宽 | 1800 GB/s | Table 2 行「NVLink bandwidth」，Blackwell 列，页面原文 1,800 GB/s | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| NVFP4 算力 | 20 PFLOPS | sparse（行「NVFP4 dense | sparse performance」，稀疏值为右侧 20）；页面另印同字段其它口径：dense（行「NVFP4 dense | sparse performance」，密集值为左侧 1 = 10PFLOPS | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| FP8 算力 | 10 PFLOPS | sparse（行「FP8 dense | sparse performance」，稀疏值 10）；页面另印同字段其它口径：dense（行「FP8 dense | sparse performance」，密集值 5） = 5PFLOPS | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| 晶体管数 | 208 十亿 | Table 2 行「Transistors」，Blackwell 列 | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| 制程节点 | TSMC 4NP | Table 2 行「Manufacturing process」，Blackwell 列 | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| C2C 互联带宽 | 900 GB/s | NVLink-C2C（CPU-GPU），Table 1 Blackwell GPU 列 | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |
| NVLink 最大 GPU 数 | 576 个 | NVLink 域上限（Blackwell 与 Blackwell Ultra 同） | verified | https://developer.nvidia.com/blog/inside-nvidia-blackwell-ultra-the-chip-powering-the-ai-factory-era/ | 2026-11-11T04:59:22 |

## 认证报价快照（24h 内 / 未过期 / 认证商家）

暂无可对智能体下发的认证报价。

## 官方视频（外链登记，不转载）

- [NVIDIA Blackwell 官方解析视频（AI Factory 世代，官方博客内嵌）](https://www.youtube.com/watch?v=eMVlMONCqxg)（版权：NVIDIA 官方发布 · 仅登记外链与元数据，不转载）

> 信息完善度：资料完整（等级 L2；衡量本站资料覆盖度，不代表产品性能，也不参与参数排序）

> 报价仅供参考，以最终合同为准。机器侧遵循字段白名单；商家与 AI 生成文本不进入此视图。

机器出口：JSON-LD https://www.1p2p3p.com/p/NV-B200.jsonld ｜ OpenAPI https://www.1p2p3p.com/openapi.json ｜ MCP https://www.1p2p3p.com/mcp ｜ llms.txt https://www.1p2p3p.com/llms.txt