AMD Radeon Instinct MI325X vs NVIDIA GeForce RTX 4090 48 GB

Comparison AMD Radeon Instinct MI325X with 288 GB HBM3e and 19,456 cores vs NVIDIA GeForce RTX 4090 48 GB with 48 GB GDDR6X and 16,384 cores.

Loading...

Performance Rating

AMD Radeon Instinct MI325X outperforms NVIDIA GeForce RTX 4090 48 GB by 181.06% 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 4090 48 GB

35.6

NVIDIA GeForce RTX 4090 48 GB

35.6

Expert Comparison

AMD Radeon Instinct MI325X обладает значительно большей памятью (288 ГБ HBM3e против 48 ГБ GDDR6X) и пропускной способностью (10.3 TB/s против 1.01 TB/s), что делает его предпочтительным для вычислений больших данных и AI-задач. В то же время, NVIDIA GeForce RTX 4090 48 GB предлагает более высокую частоту Boost (2520 МГц против 2100 МГц) и больше ядер (16384 против 19456), что делает его лучшим выбором для графики и игр.

Contents:

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

Memory

Memory Size

🔥 +500% 288 ГБ
48 GB

Memory Type

HBM3e GDDR6X

Memory Bandwidth

🔥 +920% 10.3 TB/s
1.01 TB/s

Memory Bus Width

8,192 бит 384 бит

ML Performance

FP16 (Half Precision)

🔥 +692% 653.7 TFLOPS
82.58 TFLOPS

BF16 (Brain Float)

No No

TF32 (TensorFloat)

No No

Compute Power

FP32 (Single Precision)

81.72 TFLOPS
🔥 +1% 82.58 TFLOPS

FP64 (Double Precision)

🔥 81.72 TFLOPS
No

CUDA Cores

🔥 +19% 19,456
16,384

RT Cores

No
🔥 128

Architecture & Compatibility

GPU Architecture

CDNA 3.0 Ada Lovelace

SM (Streaming Multiprocessor)

No
🔥 128

PCIe Version

PCIe 5.0 x16 PCIe 4.0 x16

ML Software Support

CUDA Version

No 8.9

CUDA Toolkit (first supported)

v11

CUDA Toolkit status

Supported Supported

Clocks & Performance

Base Clock

1,000
🔥 +124% 2,235

Boost Clock

2,100
🔥 +20% 2,520

Memory Clock

🔥 +92% 2,525
1,313

Power Consumption

Recommended PSU

1400 W
🔥 -39% 850 W

Power Connector

None 1x 16-pin

TDP/TGP

1000 W
🔥 -55% 450 W

Rendering

Texture Units (TMU)

🔥 +138% 1,216
512

ROP

No
🔥 128

L2 Cache

16 MB
🔥 +350% 72 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 Sept. 20, 2022

Display Outputs

No outputs
HDMI
DisplayPort

Rent on Intelion Cloud

RTX 4090 (48GB)

Available for rent on Intelion Cloud

from 81.85 RUB/hour Rent
Pay-as-you-go, per-second billing
Prices include 22% VAT
EDI & closing documents
Server ready in 3 minutes
Russian-speaking support
Card, SBP or invoice