Home GPU Comparison AMD Radeon Instinct MI8 vs NVIDIA H100 PCIe

AMD Radeon Instinct MI8 vs NVIDIA H100 PCIe

AI GPU We compared a Professional market GPU: 4GB VRAM Radeon Instinct MI8 and a GPU: 80GB VRAM H100 PCIe to see which GPU has better performance in key specifications, benchmark tests, power consumption, etc.

Main Differences

AMD Radeon Instinct MI8 's Advantages
Lower TDP (175W vs 350W)
NVIDIA H100 PCIe 's Advantages
Released 5 years and 3 months late
Boost Clock1755MHz
More VRAM (80GB vs 4GB)
Larger VRAM bandwidth (2039GB/s vs 512.0GB/s)
10496 additional rendering cores

Score

Benchmark

FP32 (float)
Radeon Instinct MI8
8.192 TFLOPS
H100 PCIe +525%
51.22 TFLOPS
VS

Graphics Card

Dec 2016
Release Date
Mar 2022
Radeon Instinct
Generation
Tesla Hopper
Professional
Type
AI GPU
PCIe 3.0 x16
Bus Interface
PCIe 5.0 x16

Clock Speeds

-
Base Clock
1095 MHz
-
Boost Clock
1755 MHz
500 MHz
Memory Clock
1593 MHz

Memory

4GB
Memory Size
80GB
HBM
Memory Type
HBM2e
4096bit
Memory Bus
5120bit
512.0GB/s
Bandwidth
2039GB/s

Render Config

-
SM Count
114
64
Compute Units
-
4096
Shading Units
14592
256
TMUs
456
64
ROPs
24
-
Tensor Cores
456
-
RT Cores
-
16 KB (per CU)
L1 Cache
256 KB (per SM)
2 MB
L2 Cache
50 MB

Theoretical Performance

64.00 GPixel/s
Pixel Rate
42.12 GPixel/s
256.0 GTexel/s
Texture Rate
800.3 GTexel/s
8.192 TFLOPS
FP16 (half)
204.9 TFLOPS
8.192 TFLOPS
FP32 (float)
51.22 TFLOPS
512.0 GFLOPS
FP64 (double)
25.61 TFLOPS

Graphics Processor

Fiji
GPU Name
GH100
Fiji XT CA (215-0862120)
GPU Variant
-
GCN 3.0
Architecture
Hopper
TSMC
Foundry
TSMC
28 nm
Process Size
4 nm
8.9 billion
Transistors
80 billion
596 mm²
Die Size
814 mm²

Board Design

175W
TDP
350W
450 W
Suggested PSU
750 W
No outputs
Outputs
No outputs
1x 8-pin
Power Connectors
1x 16-pin

Graphics Features

12 (12_0)
DirectX
N/A
4.6
OpenGL
N/A
2.1
OpenCL
3.0
1.2.170
Vulkan
N/A
-
CUDA
9.0
6.5
Shader Model
N/A

Related GPU Comparisons

© 2024 - TopCPU.net   Contact Us Privacy Policy