GeekRadar / The comparison desk

COMPARE.

GPU

NVIDIA

H100 PCIe 80 GB

Graphics processor

Video memory: 80 GB^*
Memory bus: 5,120 bit^*
Rated memory bandwidth: 2,000 GB/s^*

No matched benchmark results are available for this entry yet.

Processing & platform

Architecture: Hopper^*
Process node: 5 nm^*
GPU base clock: 1,095 MHz^*
GPU boost clock: 1,755 MHz^*

Memory & storage

Video memory: 80 GB^*
Video memory type: Reported value 1: GDDR6X^*
Reported value 2: HBM2e^*
Memory bus: 5,120 bit^*
Rated memory bandwidth: 2,000 GB/s^*

System configuration

Detailed specifications

Specifications 1 · 21 fields

GPU architecture — NVIDIA H100 PCIe — 80 GB HBM2e

Hopper^*

Theoretical peak: BF16 tensor dense per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

756^* TFLOPS

Scope and conditions

Accumulation: FP32^*

Theoretical peak: FP16 tensor dense per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

756^* TFLOPS

Scope and conditions

Accumulation: FP32^*

Theoretical peak: FP32 non-tensor dense per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

51.2^* TFLOPS

Native hardware support: FP4 tensor — NVIDIA H100 PCIe — 80 GB HBM2e

No^*

Theoretical peak: FP64 non-tensor dense per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

25.6^* TFLOPS

Theoretical peak: FP64 tensor dense per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

51.2^* TFLOPS

Native hardware support: FP6 tensor — NVIDIA H100 PCIe — 80 GB HBM2e

No^*

Theoretical peak: FP8 tensor dense per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

1,513^* TFLOPS

Scope and conditions

Accumulation: FP32^*

Theoretical peak: INT8 tensor dense per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

1,513^* TOPS

Theoretical peak: TF32 tensor dense per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

378^* TFLOPS

Physical configuration — NVIDIA H100 PCIe — 80 GB HBM2e

PCIe (FHFL dual-slot)^*

GPU interconnect bandwidth (NVLink 4 (three NVLink bridges, two cards)) — NVIDIA H100 PCIe — 80 GB HBM2e

600^* GB/s

PCIe interface declaration — NVIDIA H100 PCIe — 80 GB HBM2e

PCI Express Gen5 x16^*

Maximum board power — NVIDIA H100 PCIe — 80 GB HBM2e

350^* W

Peak memory bandwidth per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

2,000^* GB/s

Memory interface width per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

5,120^* bit

Memory capacity per GPU — NVIDIA H100 PCIe — 80 GB HBM2e

80^* GB

Memory reported in a documented driver sample — NVIDIA H100 PCIe — 80 GB HBM2e

81,559^* MiB

Scope and conditions

Configuration: Documented driver sample; not nominal VRAM capacity.^*

Memory technology — NVIDIA H100 PCIe — 80 GB HBM2e

HBM2e^*

GPU model — NVIDIA H100 PCIe — 80 GB HBM2e

NVIDIA H100 PCIe^*

Specifications 2 · 8 fields

Product Name: H100 PCIe 80 GB^*
GPU Chip: GH100^*
Released: Mar 22nd, 2022^*
Bus: PCIe 5.0 x16^*
Memory: 80 GB, HBM2e, 5120 bit^*
GPU clock: 1,095^* MHz
Memory clock: 1,593^* MHz
Shaders / TMUs / ROPs: 14592 / 456 / 24^*

Specifications 3 · 52 fields

Brand: NVIDIA^*
Top BUS WIDTH: 5,120^* bit
Graphics Processor GPU Name: GH100^*
Graphics Processor Architecture: Hopper^*
Render Config SM Count: 114^*
Graphics Features CUDA: 9^*
Render Config Tensor Cores: 456^*
Graphics Processor Process Size: 5^* nm
Graphics Processor Transistors: 80,000 million^*
Graphics Processor Die Size: 814^* mm²
Graphics Card Release Date: Oct 2022^*
Graphics Card Generation: Server Hopper (Hxx)^*
Graphics Card Successor: Server Blackwell^*
Graphics Card Production: Active^*
Name: H100 PCIe 80 GB^*
Graphics Card Bus Interface: PCIe 5.0 x16^*
Clock Speeds Memory Clock: 1593 MHz 3.2 Gbps effective^*
Memory Memory Size: 80^* GB
Memory Memory Type: HBM2e^*
Memory Memory Bus: 5,120^* bit
Memory Bandwidth: 2.04 TB/s^*
Render Config TMUs: 456^*
Top GRAPHICS PROCESSOR: GH100^*
Render Config ROPs: 24^*
Theoretical Performance Pixel Rate: 42.12 GPixel/s
Theoretical Performance Texture Rate: 800.3 GTexel/s
Board Design Slot Width: Dual-slot^*
Board Design TDP: 350^* W
Board Design Suggested PSU: 750^* W
Board Design Outputs: No outputs^*
Board Design Board Number: P1010 SKU 200^*
Graphics Features OpenCL: 3^*
source file: nvidia2022.json^*
Graphics Processor Foundry: TSMC^*
Graphics Card Predecessor: Server Ada^*
Board Design Length: 268 mm 10.6 inches^*
Board Design Power Connectors: 1x 16-pin^*
Graphics Processor Density: 98.3M / mm²^*
Board Design Height: 111 mm 4.4 inches^*
Top TMUS: 456^*
Top ROPS: 24^*
Clock Speeds Base Clock: 1,095^* MHz
Clock Speeds Boost Clock: 1,755^* MHz
Top CORES: 14,592^*
Render Config Shading Units: 14,592^*
Render Config L2 Cache: 50^* MB
Top MEMORY SIZE: 80^* GB
Theoretical Performance FP32 (float): 51.22^* TFLOPS
Theoretical Performance FP64 (double): 25.61 TFLOPS (1:2)
Render Config L1 Cache: 256 KB (per SM)^*
Theoretical Performance FP16 (half): 204.9 TFLOPS (4:1)
Top MEMORY TYPE: HBM2e^*

Specifications 4 · 14 fields

manufacturer: NVIDIA^*
rop: 96^*
igp: No^*
bus: PCIe 4.0 x16^*
memType: GDDR6X^*
gpuChip: GA102^*
productName: H100 PCIe 80 GB^*
releaseYear: 2,023^*
memSize: 80^*
memBusWidth: 5,120^*
gpuClock: 1,260^*
memClock: 1,188^*
unifiedShader: 8,960^*
tmu: 280^*

* Provisional data