Skip to main content

NVIDIA B200 (Blackwell)

Product Overview

The NVIDIA B200 is the flagship GPU of the Blackwell architecture, with 18 PFLOPS FP4 (sparse) per GPU — the most powerful AI training and inference GPU as of 2025. Like the B100, it uses a dual-die design, but TDP climbs to 1,000 W, requiring redesigned servers and liquid cooling. NVIDIA also offers the GB200 Grace Blackwell Superchip (2× B200 + Grace CPU) and the NVL72 rack (72 B200 GPUs + 36 Grace CPUs).

Core Specifications

ParameterValue
ArchitectureBlackwell GB100 (dual-die)
Process NodeTSMC 4NP
Transistor Count208 billion
Memory192 GB HBM3e
Memory Bandwidth8 TB/s
FP4 Tensor Core18 PFLOPS (sparse)
FP8 Tensor Core9 PFLOPS (sparse)
FP16 Tensor Core4.5 PFLOPS (sparse)
FP6440 TFLOPS
NVLink1.8 TB/s (5th Gen)
TDP1,000 W (up to 1,200 W)
PCIeGen 5

Vendor Information

ParameterValue
ManufacturerNVIDIA Corporation
Official Websitehttps://www.nvidia.com
Product Pagehttps://www.nvidia.com/en-us/data-center/blackwell/
ReleaseGTC 2024, mass production 2025
Cloud Price~$5.87/hr (180GB variant)

Key B200 Systems

SystemDescription
HGX B2008× B200 server baseboard
GB200Grace CPU + 2× B200 Superchip
NVL7272× B200 + 36× Grace, 1.4 ExaFLOPS FP4 per rack

Software & Drivers

Key Features

  • 2nd Gen Transformer Engine: FP4 / FP6 automatic precision switching
  • Mixture of Experts (MoE) Optimization: Hardware-accelerated MoE model routing
  • Confidential Computing: Hardware-level encrypted computation
  • RAS Engine: Server-grade reliability

Use Cases

  • Trillion-parameter LLM training
  • Large-scale AI inference (trillion-parameter single-GPU inference)
  • MoE models (Mixtral, Llama 4, etc.)
  • Multimodal large models
  • Scientific computing (HPC)