NVIDIA H100 CNX videocard review

H100 CNX videocard released by NVIDIA; release date: 22 Mar 2022. The videocard is based on Hopper microarchitecture codenamed GH100.

Core clock speed - 690 MHz. Boost clock speed - 1845 MHz. Texture fill rate - 841.3 GTexel/s. Pipelines - 14592. Manufacturing process technology - 4 nm. Transistors count - 80000 million. Power consumption (TDP) - 350 Watt.

Memory type: HBM2e. Maximum RAM amount - 80 GB. Memory bus width - 5120 bit. Memory clock speed - 1593 MHz, 3.2 Gbps effective. Memory bandwidth - 2,039 GB/s.

Specifications (specs)

Essentials
Architecture	Hopper
Code name	GH100
Launch date	22 Mar 2022
Place in performance rating	not rated
Technical info
Boost clock speed	1845 MHz
Core clock speed	690 MHz
Manufacturing process technology	4 nm
Pipelines	14592
Pixel fill rate	44.28 GPixel/s
Texture fill rate	841.3 GTexel/s
Thermal Design Power (TDP)	350 Watt
Transistor count	80000 million

Video outputs and ports
Display Connectors	No outputs
Compatibility, dimensions and requirements
Form factor	Dual-slot
Interface	PCIe 5.0 x16
Length	267 mm, 10.5 inches
Recommended system power (PSU)	750 Watt
Supplementary power connectors	8-pin EPS
Width	111 mm, 4.4 inches
API support
OpenCL	3.0
Memory
Maximum RAM amount	80 GB
Memory bandwidth	2,039 GB/s
Memory bus width	5120 bit
Memory clock speed	1593 MHz, 3.2 Gbps effective
Memory type	HBM2e