AMD Radeon Instinct MI325X vs NVIDIA GeForce RTX 3090

Comparison of AMD Radeon Instinct MI325X with 288 GB HBM3e and 19,456 cores vs NVIDIA GeForce RTX 3090 with 24 GB GDDR6X and 10,496 cores.

Loading...

Performance Rating

AMD Radeon Instinct MI325X outperforms NVIDIA GeForce RTX 3090 by 403.27% 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 GeForce RTX 3090

19.9

NVIDIA GeForce RTX 3090

19.9

Expert Comparison

AMD Radeon Instinct MI325X значительно превосходит NVIDIA GeForce RTX 3090 по памяти и пропускной способности: у нее 288 ГБ HBM3e при 10.3 TB/s против 24 ГБ GDDR6X и 936.2 GB/s у RTX 3090. Это делает MI325X более подходящей для вычислений и обработки больших объемов данных. RTX 3090 же сильнее в играх и графических задачах благодаря большему количеству ядер (10496 против 19456) и более высокой базовой и буст-частоте.

Contents:

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

Memory

Memory Size

🔥 +1,100% 288 ГБ
24 ГБ

Memory Type

HBM3e GDDR6X

Memory Bandwidth

🔥 10.3 TB/s
936.2 GB/s

Memory Bus Width

8,192 бит 384 бит

ML Performance

FP16 (Half Precision)

🔥 +1,737% 653.7 TFLOPS
35.58 TFLOPS

BF16 (Brain Float)

No No

TF32 (TensorFloat)

No No

Compute Power

FP32 (Single Precision)

🔥 +130% 81.72 TFLOPS
35.58 TFLOPS

FP64 (Double Precision)

🔥 +14,598% 81.72 TFLOPS
0.556 TFLOPS

CUDA Cores

🔥 +85% 19,456
10,496

RT Cores

No
🔥 82

Architecture & Compatibility

GPU Architecture

CDNA 3.0 Ampere

SM (Streaming Multiprocessor)

No
🔥 82

PCIe Version

PCIe 5.0 x16 PCIe 4.0 x16

ML Software Support

CUDA Version

No 8.6

Clocks & Performance

Base Clock

1,000
🔥 +40% 1,395

Boost Clock

🔥 +24% 2,100
1,695

Memory Clock

🔥 +107% 2,525
1,219

Power Consumption

TDP/TGP

1000 W
🔥 -65% 350 W

Recommended PSU

1400 W
🔥 -46% 750 W

Power Connector

None 1x 12-pin

Rendering

Texture Units (TMU)

🔥 +271% 1,216
328

ROP

No
🔥 82

L2 Cache

16 MB
🔥 6 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 Triple-slot

Release Date

Oct. 15, 2024 Sept. 1, 2020

Display Outputs

No outputs
1x HDMI 2.1
3x DisplayPort 1.4a

Renting is cheaper than buying