Report an Error

NVIDIA GB200

NVIDIA's GB200 GPU uses the Blackwell architecture and is made using a 5 nm production process at TSMC. GB200 does not support DirectX. For GPU compute applications, OpenCL version 3.0 and CUDA 12.0 can be used. It features 18432 shading units, 576 texture mapping units and 24 ROPs. Also included are 576 tensor cores which help improve the speed of machine learning applications.
Further reading: Blackwell Architecture Whitepaper

Graphics Processor

Released
Unknown
GPU Name
GB200
Codename
NV190
Architecture
Blackwell
Foundry
TSMC
Process Type
4NP FinFET
Process Size
5 nm
Transistors
104,000 million
Die Size
unknown

Graphics Features

DirectX
N/A
OpenGL
N/A
OpenCL
3.0
Vulkan
N/A
CUDA
12.0
Shader Model
N/A
Tensor Cores
5th Gen
PureVideo HD
VP12
VDPAU
Feature Set L

Render Config

Shading Units
18432
TMUs
576
ROPs
24
Tensor Cores
576
L2 Cache
61440 KB
FP16 Units
73,728
FP64 Units
9216
INT32 Units
9216
SFUs
2304
TPCs
72
GPCs
8
Tex L1 Cache
64 KB per SM

All Blackwell GPUs

NVIDIA GPU Architecture History

GB200 GPU Notes

Tensor Cores: 5th Gen
NVENC: No Support
NVDEC: No Support
PureVideo HD: VP12
VDPAU: Feature Set L

Most specs unknown
Placeholder
Apr 4th, 2025 10:18 EDT change timezone

Latest GPU Drivers

New Forum Posts

Popular Reviews

Controversial News Posts