AMD Radeon Instinct MI325X vs NVIDIA RTX PRO 6000 Blackwell Server

Comparison of AMD Radeon Instinct MI325X with 288 GB HBM3e and 19,456 cores vs NVIDIA RTX PRO 6000 Blackwell Server with 96 GB GDDR7 and 24,064 cores.

Loading...

Performance Rating

AMD Radeon Instinct MI325X outperforms NVIDIA RTX PRO 6000 Blackwell Server by 74.0% 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 RTX PRO 6000 Blackwell Server

57.5

NVIDIA RTX PRO 6000 Blackwell Server

57.5

Expert Comparison

AMD Radeon Instinct MI325X имеет значительно большую память (288 ГБ HBM3e против 96 ГБ GDDR7) и пропускную способность (10.3 TB/s против 1.79 TB/s), что делает его более эффективным для вычислений и больших объемов данных. Однако NVIDIA RTX PRO 6000 Blackwell Server предлагает более высокую производительность FP32 (126.0 TFLOPS против 81.72 TFLOPS) и меньший TDP (600 W против 1000 W), что делает его более энергоэффективным для серверных задач.

Contents:

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

Memory

Memory Size

🔥 +200% 288 ГБ
96 ГБ

Memory Type

HBM3e GDDR7

Memory Bandwidth

🔥 +475% 10.3 TB/s
1.79 TB/s

Memory Bus Width

8,192 бит 512 бит

ML Performance

FP16 (Half Precision)

🔥 +419% 653.7 TFLOPS
126.0 TFLOPS

BF16 (Brain Float)

No No

TF32 (TensorFloat)

No No

Compute Power

FP32 (Single Precision)

81.72 TFLOPS
🔥 +54% 126.0 TFLOPS

FP64 (Double Precision)

🔥 +4,052% 81.72 TFLOPS
1.968 TFLOPS

CUDA Cores

19,456
🔥 +24% 24,064

RT Cores

No
🔥 188

Architecture & Compatibility

GPU Architecture

CDNA 3.0 Blackwell 2.0

SM (Streaming Multiprocessor)

No
🔥 188

PCIe Version

PCIe 5.0 x16 PCIe 5.0 x16

ML Software Support

CUDA Version

No 12.0

Clocks & Performance

Base Clock

1,000
🔥 +59% 1,590

Boost Clock

2,100
🔥 +25% 2,617

Memory Clock

🔥 +44% 2,525
1,750

Power Consumption

TDP/TGP

1000 W
🔥 -40% 600 W

Recommended PSU

1400 W
🔥 -29% 1000 W

Power Connector

None 1x 16-pin

Rendering

Texture Units (TMU)

🔥 +62% 1,216
752

ROP

No
🔥 188

L2 Cache

🔥 16 MB
128 MB

Benchmarks

MLPerf, llama2-70b-99.9 (Dummy)

3 596 tokens/s

MLPerf, llama2-70b-99.9 (fp4)

3 250 tokens/s

MLPerf, llama2-70b-99.9 (fp8)

1 946 tokens/s

MLPerf, llama3.1-8b (fp4)

5 758 tokens/s

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

22.4 tokens/s

Geekbench AI, FP16

53 322 points

Geekbench AI, INT8

28 264 points

Geekbench AI, FP32

37 299 points

MLPerf, mixtral-8x7b (fp8)

🔥 +85% 6 975 tokens/s
3 767 tokens/s

Additional

Slots

OAM Module Dual-slot

Release Date

Oct. 15, 2024 March 18, 2025

Display Outputs

No outputs
4x DisplayPort 2.1b

Renting is cheaper than buying