NVIDIA H100 CNX Professional Graphics Card

Product status: Official | Last Update: 2022-03-22 | Report Error

Graphics Processor

GH100

Cores

NVIDIA H100 CNX

Manufacturer

NVIDIA

Original Series

Data Center Hopper

Launch Date

March 22nd, 2022

GPU Model

GH100

Architecture

Hopper

Fabrication Process

5 nm (TSMC N4)

Die Size

814 mm²

Transistors Count

80B

Transistors Density

98.3M ^TRAN/_mm²

CUDAs

14592

Tensor Cores

456

114

GPCs

Base Clock

TBC MHz

Boost Clock

1775 MHz

Memory Clock

1560 MHz

Effective Memory Clock

3120 Mbps

Memory Size

81920 MB

Memory Type

HBM2e

Memory Bus Width

5120-bit

Memory Bandwidth

1,996.8 GB/s

Interface

PCI-Express 5.0 x16

Power Connectors

1× 12-pin

TDP/TBP

350 W

Recommended PSU

700 W

DirectX

Vulkan

OpenGL

OpenCL

Peak FP32

51.8 TFLOPS

FP32 Perf. per Watt

148 GFLOPS/W

FP32 Perf. per mm²

63.6 GFLOPS/mm²

Pictures

Front

Data Center Hopper Series4

Model	Cores	Boost Clock	Memory Clock	Memory Config.
NVIDIA H100 PCIe 120GB	TBC	TBC	TBC	TBC
NVIDIA H100 SXM 80GB	16896	1775 MHz	4.8 Gbps	80 GB 5120b
NVIDIA H100 CNX	14592	1775 MHz	3.1 Gbps	80 GB 5120b
NVIDIA H100 PCIe 80GB	14592	1775 MHz	3.1 Gbps	80 GB 5120b

GH100 GPU Cards4

Model	Cores	Boost Clock	Memory Clock	Memory Config.
NVIDIA H100 PCIe 120GB	TBC	TBC	TBC	TBC
NVIDIA H100 SXM 80GB	16896	1775 MHz	4.8 GB/s	80 GB 5120b
NVIDIA H100 CNX	14592	1775 MHz	3.1 GB/s	80 GB 5120b
NVIDIA H100 PCIe 80GB	14592	1775 MHz	3.1 GB/s	80 GB 5120b