NVIDIA H100 PCIe — Specifications

Specifications Benchmarks Synonyms

Specifications

Technical data

GPU

Architecture
Hopper
Execution Units
7296
GPU Chip
GH100
Render Output Units
24
Texture Units
456

Memory

Memory Bandwidth
1,280 GB/s
Memory Bus
5120 bit
Memory Clock
1000 MHz
Memory Data Rate
2 Gbps
Memory Size
80 GB
Memory Type
HBM2e

Clock Speeds

Base Clock
1065 MHz
Boost Clock
1650 MHz

Thermal Design

System Power Supply
750 W (PSU recommendation)
TDP
350 W

Connectivity

Max. Displays
1
PCIe Interface
PCIe 5.0 x16
Power Connectors
8-pin EPS

Featureset

CUDA
9.0
DirectX
No
OpenCL
3.0
OpenGL
No
Shader Model
No
Vulkan
No

Additional data

Process Technology
4 nm
Release date
Q2/2022

Theoretical Performance

FP16 (theoretical)
96.31 TFLOPS
FP32 (theoretical)
24.08 TFLOPS
FP64 (theoretical)
12.04 TFLOPS
Pixel Rate
39.60 GPixel/s
Texture Rate
752.4 GTexel/s

Positions in all rankings

Common positions NVIDIA H100 PCIe GPU in popular benchmarks, for comparison with other models.

No data

Benchmarks

Performance tests GPU

No benchmark results are available yet. This model is listed with specifications only.

These are saved results from the earlier catalog. Measurement dates and test configurations were not recorded. Models with no comparable results are omitted.

Recent models: test results with identified primary sources — Game frame rates

Comparable benchmark results are required to suggest performance alternatives.

Base Clock is the base GPU frequency stated by the manufacturer for this model.

Boost Clock describes automatic frequency increases. Actual clock speed varies with workload, temperature and power limits; it is not a guarantee of manual overclocking.

Rate the video card
NVIDIA H100 PCIe