AMD Radeon Instinct MI250 vs NVIDIA H100 PCIe 80 GB

Comparison of AMD Radeon Instinct MI250 with 128 GB HBM2e and 13,312 cores vs NVIDIA H100 PCIe 80 GB with 80 GB HBM2e and 14,592 cores.

Loading...

Performance Rating

H200 H200
MI325X MI325X
A100 A100

AMD Radeon Instinct MI250

AMD Radeon Instinct MI250

RX 7900 XTX RX 7900 XTX
Instinct MI300X Instinct MI300X
MI250 MI250

NVIDIA H100 PCIe 80 GB

NVIDIA H100 PCIe 80 GB

Contents:

Memory ML Performance Compute Power Architecture & Compatibility ML Software Support Clocks & Performance Power Consumption Rendering Benchmarks Additional

Memory

Memory Size

128 GB 80 GB

Memory Type

HBM2e HBM2e

Memory Bandwidth

3.28 TB/s 2.04 TB/s

Memory Bus Width

8,192 бит 5,120 бит

ML Performance

FP16 (Half Precision)

362.1 TFLOPS 204.9 TFLOPS

BF16 (Brain Float)

No No

TF32 (TensorFloat)

No No

Compute Power

FP32 (Single Precision)

45.26 TFLOPS 51.22 TFLOPS

FP64 (Double Precision)

45.26 TFLOPS 25.61 TFLOPS

CUDA Cores

13,312 14,592

RT Cores

No No

Architecture & Compatibility

GPU Architecture

CDNA 2.0 Hopper

SM (Streaming Multiprocessor)

No 114

PCIe Version

PCIe 4.0 x16 PCIe 5.0 x16

ML Software Support

CUDA Version

No 9.0

Clocks & Performance

Base Clock

1,000 1,095

Boost Clock

1,700 1,755

Memory Clock

1,600 1,593

Power Consumption

TDP/TGP

500 W
🔥 -30% 350 W

Recommended PSU

900 W
🔥 -17% 750 W

Power Connector

2x 8-pin 1x 16-pin

Rendering

Texture Units (TMU)

832 456

ROP

No No

L2 Cache

16 MB 50 MB

Benchmarks

MLPerf, llama2-70b-99.9 (fp8)

1 222 tokens/s

llama.cpp, llama-2-7b-Q4_0

63.9 tokens/s

MLPerf, mixtral-8x7b (fp8)

3 466 tokens/s

Additional

Slots

Dual-slot Dual-slot

Release Date

Nov. 8, 2021 Oct. 13, 2022

Display Outputs

No outputs
No outputs

Renting is cheaper than buying