NVIDIA H100 PCIe vs NVIDIA H100 SXM5. Specifications

Comparison of basic technical data of graphics cards NVIDIA H100 PCIe and NVIDIA H100 SXM5, chip, information processing units.

• Architecture
  Hopper Hopper
• Execution Units
  7296 8448
• GPU Chip
  GH100 GH100
• Render Output Units
  24 24
• Texture Units
  456 528
• L1 Cache
  - 192 KB
• L2 Cache
  - 50 MB

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

• Memory Bandwidth
  1,280 GB/s 1,920 GB/s
• Memory Bus
  5120 bit 5120 bit
• Memory Clock
  1000 MHz 1500 MHz
• Memory Data Rate
  2 Gbps 3 Gbps
• Memory Size
  80 GB 80 GB
• Memory Type
  HBM2e HBM3

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

• Base Clock
  1065 MHz 1065 MHz
• Boost Clock
  1650 MHz 1780 MHz

Graphics card power limits and recommended system power supply describe different specifications.

• System Power Supply
  750 W (PSU recommendation) 1100 W (PSU recommendation)
• TDP
  350 W 700 W
• Board Number
  - PG520

Connectivity and connections.

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

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

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

Comparison of interfaces and release dates for GPU NVIDIA H100 SXM5 and NVIDIA H100 PCIe.

• Process Technology
  4 nm 4 nm
• Release date
  Q2/2022 Q2/2022

• FP16 (theoretical)
  96.31 TFLOPS 120.3 TFLOPS
• FP32 (theoretical)
  24.08 TFLOPS 30.07 TFLOPS
• FP64 (theoretical)
  12.04 TFLOPS 15.04 TFLOPS
• Pixel Rate
  39.60 GPixel/s 42.72 GPixel/s
• Texture Rate
  752.4 GTexel/s 939.8 GTexel/s