NVIDIA GeForce RTX 4070 SUPER vs NVIDIA H100 SXM5. Specifications

Comparison of basic technical data of graphics cards NVIDIA GeForce RTX 4070 SUPER and NVIDIA H100 SXM5, chip, information processing units.

• CUDA Cores
  7168 -
• Shader
  7168 -
• Architecture
  Ada Lovelace -
• Based on
  NVIDIA GeForce RTX 4070 SUPER -
• L2 Cache
  - 50 MB
• L1 Cache
  - 192 KB
• Texture Units
  - 528
• Render Output Units
  - 24
• Execution units
  - 8448
• GPU Chip
  - GH100

Video memory capacity, type, bus width and bandwidth describe different properties. These values alone do not determine graphics performance.

• Memory Bus
  192 bit -
• Memory Type
  GDDR6X HBM3
• Memory Size
  12 GB 80 GB
• Memory Interface
  - 5120 bit
• Memory Bandwith
  - 1,920 GB/s
• Memory Speed
  - 3 Gbps
• Memory Clock
  - 1.5 GHz

Published GPU clock frequencies describe different operating modes. Clock speed alone does not determine graphics performance.

• Boost Clock
  2.48 GHz 1.78 GHz
• Base Clock
  1.98 GHz 1.065 GHz

Connectors, the number of thermal watts emitted in normal mode and at overclocking.

• Maximum GPU Temperature
  90 C -
• Required System Power
  650 W -
• TDP
  220 W 700 W
• Board Number
  - PG520
• Suggested PSU
  - 1100 W
• Power Connectors
  - 8-pin EPS

Connectivity and connections.

• PCIe
  4.0 -

Technical data that is used to its full potential in computer games.

• CUDA
  - 9.0
• Shader Model
  - No
• Vulkan
  - No
• OpenCL
  - 3.0
• OpenGL
  - No
• DirectX
  - No
• Max. resolution
  - 3840x2160@60Hz
• Max. Displays
  - 1

Built-in support for video and image compression standards.

• AV1
  Decode / Encode -

Comparison of interfaces and release dates for GPU NVIDIA H100 SXM5 and NVIDIA GeForce RTX 4070 SUPER.

• Specification source
  https://www.nvidia.com/en-us/geforce/graphics-cards/compare/ -
• Verified on
  2026-09-30 -
• Official source
  https://www.nvidia.com/en-us/geforce/graphics-cards/40-series/rtx-4070-family/ -
• Release date
  - Q2/2022
• Structure size
  - 4 nm
• GPU Interface
  - PCIe 5.0 x16
• Architecture
  - Hopper

• FP64 (double)
  - 15.04 TFLOPS
• FP32 (float)
  - 30.07 TFLOPS
• FP16 (half)
  - 120.3 TFLOPS
• Texture Rate
  - 939.8 GTexel/s
• Pixel Rate
  - 42.72 GPixel/s