AMD Radeon Instinct MI325X vs NVIDIA GRID M40

Comparison AMD Radeon Instinct MI325X with 288 GB HBM3e and 19,456 cores vs NVIDIA GRID M40 with 8 GB GDDR5 and 384 cores.

Loading...

Performance Rating

AMD Radeon Instinct MI325X outperforms NVIDIA GRID M40 by 8,595.65% in the overall GPU ARK performance rating

A100 A100
H200 H200
MI325X MI325X

AMD Radeon Instinct MI325X

100.0

AMD Radeon Instinct MI325X

100.0
RX 7900 XTX RX 7900 XTX
MI250 MI250
Instinct MI300X Instinct MI300X

NVIDIA GRID M40

1.2

NVIDIA GRID M40

1.2

Expert Comparison

AMD Radeon Instinct MI325X значительно выигрывает по характеристикам над NVIDIA GRID M40. MI325X имеет гораздо больше ядер (19456 против 384), большую память (288 ГБ HBM3e против 8 ГБ GDDR5) и пропускную способность (10.3 TB/s против 83.20 GB/s). Также MI325X предназначена для современных задач, такие как машинное обучение и вычислительная графика, благодаря своей архитектуре CDNA 3.0.

Contents:

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

Memory

Memory Size

🔥 +3,500% 288 ГБ
8 GB

Memory Type

HBM3e GDDR5

Memory Bandwidth

🔥 10.3 TB/s
83.20 GB/s

Memory Bus Width

8,192 бит 128 бит

ML Performance

FP16 (Half Precision)

🔥 653.7 TFLOPS
No

BF16 (Brain Float)

No No

TF32 (TensorFloat)

No No

Compute Power

FP32 (Single Precision)

🔥 +10,201% 81.72 TFLOPS
0.7933 TFLOPS

FP64 (Double Precision)

🔥 +329,416% 81.72 TFLOPS
0.0248 TFLOPS

CUDA Cores

🔥 +4,967% 19,456
384

RT Cores

No No

Architecture & Compatibility

GPU Architecture

CDNA 3.0 Maxwell

SM (Streaming Multiprocessor)

No No

PCIe Version

PCIe 5.0 x16 PCIe 3.0 x16

ML Software Support

CUDA Version

No 5.0

CUDA Toolkit (first supported)

v6

CUDA Toolkit (deprecated)

v12

CUDA Toolkit (support removed)

v13

CUDA Toolkit status

Supported Removed

Clocks & Performance

Base Clock

🔥 1,000
No

Boost Clock

🔥 2,100
No

Memory Clock

🔥 +94% 2,525
1,300

Power Consumption

Recommended PSU

1400 W
🔥 -82% 250 W

Power Connector

None No

TDP/TGP

1000 W
🔥 -95% 50 W

Rendering

Texture Units (TMU)

🔥 +3,700% 1,216
32

ROP

No No

L2 Cache

🔥 +700% 16 MB
2 MB

Benchmarks

MLPerf, llama2-70b-99.9 (Dummy)

3 596 tokens/s

MLPerf, llama2-70b-99.9 (fp8)

1 946 tokens/s

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

22.4 tokens/s

MLPerf, mixtral-8x7b (fp8)

6 975 tokens/s

Additional

Slots

OAM Module Dual-slot

Release Date

Oct. 12, 2024 May 18, 2016

Display Outputs

No outputs
No outputs

Renting is cheaper than buying