NVIDIA H100 SXM5 gegen NVIDIA GeForce GTX 1630. Specifications

Vergleich der grundlegenden technischen Daten von Grafikkarten NVIDIA H100 SXM5 und NVIDIA GeForce GTX 1630, Chip, Informationsverarbeitungseinheiten.

• Architektur
  Hopper Turing
• Execution Units
  8448 512
• GPU Chip
  GH100 TU117
• L1 Cache
  192 KB 64 KB
• L2 Cache
  50 MB 1024 KB
• Render Output Einheiten
  24 16
• Textur Einheiten
  528 32
• CUDA Cores
  - 512

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

• Memory Bandwidth
  1,920 GB/s 96 GB/s
• Memory Bus
  5120 bit 64 bit
• Memory Clock
  1500 MHz 1500 MHz
• Memory Data Rate
  3 Gbps 12 Gbps
• Speichergröße
  80 GB 4 GB
• Speicher Typ
  HBM3 GDDR6

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

• Grundfrequenz
  1065 MHz 1740 MHz
• Maximale Frequenz
  1780 MHz 1785 MHz

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

• Board Number
  PG520 PG174 SKU 0
• System Power Supply
  1100 W (PSU recommendation) 300 W (Minimum system power requirement)
• TDP
  700 W 75 W
• Graphics Card Power
  - 75 W
• Maximum GPU Temperature
  - 90-94 °C

Konnektivität und Verbindungen.

• Max. Anzeigen
  1 -
• PCIe Interface
  PCIe 5.0 x16 CIe 3.0 x8
• Power Connectors
  8-pin EPS 6-pin
• DisplayPort Ports
  - 1 × DisplayPort 1.4a
• DVI-Ports
  - 1 × DVI
• HDMI-Ports
  - 1 × HDMI 2.0
• Other Display Outputs (unparsed)
  - HDMI 2.0b; DL-DVI-D; DP 1.4a

Technische Daten, die in Computerspielen voll ausgeschöpft werden.

• CUDA
  9.0 7.5
• DirectX
  Nein 12.1
• Max. Auflösung
  3840x2160@60Hz 2560 × 1440
• OpenCL
  3.0 3.0
• OpenGL
  Nein 4.6
• Shader Model
  Nein 6.6
• Vulkan
  Nein 1.3

Vergleich von Schnittstellen und Freigabeterminen für GPU NVIDIA GeForce GTX 1630 und NVIDIA H100 SXM5.

• Process Technology
  4 nm 12 nm
• Datum der Veröffentlichung
  Q2/2022 Q2/2022
• Manufacturer specification scope
  - NVIDIA Founders Edition/reference design. Clock specifications apply to gaming use. AIB shipping specifications and complete-system power needs may differ.
• Multi-monitor support
  - Ja

• FP16 (theoretical)
  120.3 TFLOPS 3.656 TFLOPS
• FP32 (theoretical)
  30.07 TFLOPS 1.828 TFLOPS
• FP64 (theoretical)
  15.04 TFLOPS 0.05712 TFLOPS
• Pixel Rate
  42.72 GPixel/s 28.56 GPixel/s
• Texture Rate
  939.8 GTexel/s 57.12 GTexel/s