Home GPU Comparison NVIDIA H100 CNX vs NVIDIA RTX A6000

NVIDIA H100 CNX vs NVIDIA RTX A6000

AI GPU We compared a GPU: 80GB VRAM H100 CNX and a Professional market GPU: 48GB VRAM RTX A6000 to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.

Main Differences

NVIDIA H100 CNX 's Advantages
Released 2 years and 5 months late
Boost Clock has increased by 2% (1845MHz vs 1800MHz)
More VRAM (80GB vs 48GB)
Larger VRAM bandwidth (2039GB/s vs 768.0GB/s)
3840 additional rendering cores
NVIDIA RTX A6000 's Advantages
Lower TDP (300W vs 350W)

Score

Benchmark

FP32 (float)
H100 CNX +39%
53.84 TFLOPS
RTX A6000
38.71 TFLOPS
VS

Graphics Card

Mar 2023
Release Date
Oct 2020
Tesla Hopper
Generation
Quadro Ampere
AI GPU
Type
Professional
PCIe 5.0 x16
Bus Interface
PCIe 4.0 x16

Clock Speeds

690 MHz
Base Clock
1410 MHz
1845 MHz
Boost Clock
1800 MHz
1593 MHz
Memory Clock
2000 MHz

Memory

80GB
Memory Size
48GB
HBM2e
Memory Type
GDDR6
5120bit
Memory Bus
384bit
2039GB/s
Bandwidth
768.0GB/s

Render Config

114
SM Count
84
-
Compute Units
-
14592
Shading Units
10752
456
TMUs
336
24
ROPs
112
456
Tensor Cores
336
-
RT Cores
84
256 KB (per SM)
L1 Cache
128 KB (per SM)
50 MB
L2 Cache
6 MB

Theoretical Performance

44.28 GPixel/s
Pixel Rate
201.6 GPixel/s
841.3 GTexel/s
Texture Rate
604.8 GTexel/s
215.4 TFLOPS
FP16 (half)
38.71 TFLOPS
53.84 TFLOPS
FP32 (float)
38.71 TFLOPS
26.92 TFLOPS
FP64 (double)
604.8 GFLOPS

Graphics Processor

GH100
GPU Name
GA102
-
GPU Variant
-
Hopper
Architecture
Ampere
TSMC
Foundry
Samsung
4 nm
Process Size
8 nm
80 billion
Transistors
28.3 billion
814 mm²
Die Size
628 mm²

Board Design

350W
TDP
300W
750 W
Suggested PSU
700 W
No outputs
Outputs
4x DisplayPort 1.4a
8-pin EPS
Power Connectors
8-pin EPS

Graphics Features

N/A
DirectX
12 Ultimate (12_2)
N/A
OpenGL
4.6
3.0
OpenCL
3.0
N/A
Vulkan
1.3
9.0
CUDA
8.6
N/A
Shader Model
6.7

Related GPU Comparisons

© 2024 - TopCPU.net   Contact Us Privacy Policy