Nvidia B200 (Blackwell)
B200 is Nvidia’s first Blackwell data-center GPU, with 208 billion transistors on TSMC 4NP.
- Vendor
- Nvidia
- Architecture
- Blackwell
- Memory
- Up to 192 GB HBM3E (180 GB per GPU in HGX B200)
- Memory bandwidth (TB/s)
- 8 TB/s
- FP8 (TFLOPS)
- 5,000 TFLOPS
- FP4 (PFLOPS)
- 10 PFLOPS
- Power (W)
- 1,200 W
- Form factor
- SXM module (HGX B200 / DGX B200)
- Status
- Shipping
The B200 joins two reticle-limited dies with a 10 TB/s link. Nvidia’s Blackwell GPU comparison table lists up to 192 GB of HBM3E, 8 TB/s of memory bandwidth, 5 PFLOPS of dense FP8, 10 PFLOPS of dense NVFP4 and up to 1,200 W. In the HGX B200 board Nvidia lists 180 GB of HBM3e per GPU, up to 8 TB/s and up to 1 kW per GPU; a DGX B200 totals 1,440 GB and 72 PFLOPS FP8 (sparse) across eight GPUs.
- Memory (GB)
- 192 GB
- Precision note (sparse or dense)
- Dense figures from Nvidia's Blackwell GPU table: FP8 5 PFLOPS dense / 10 sparse; NVFP4 10 PFLOPS dense / 20 sparse. HGX B200 8-GPU board: FP8 72 PFLOPS sparse, FP4 144 sparse / 72 dense.
- Vendor source
- developer.nvidia.com
- Checked
Questions
How much memory does the Nvidia B200 have?
Nvidia’s Blackwell comparison table lists up to 192 GB of HBM3E; in the HGX B200 board each GPU has 180 GB of HBM3e.
What is the B200’s FP8 performance?
Nvidia lists 5 PFLOPS of dense FP8 and 10 PFLOPS with sparsity for the B200.
How much power does the B200 use?
Nvidia’s Blackwell table gives up to 1,200 W; in the HGX B200 board Nvidia lists up to 1 kW per GPU.
What process is the B200 built on?
The B200 is built on TSMC 4NP and has 208 billion transistors across two dies joined by a 10 TB/s link.
Source
Every figure is taken from the vendor’s own datasheet, product page, technical documentation or announcement, linked on each chip with the date it was last checked. How we pick and check sources: editorial standards.