AMD Radeon Instinct MI325X vs NVIDIA B100

Comparison AMD Radeon Instinct MI325X with 288 GB HBM3e and 19,456 cores vs NVIDIA B100 with 96 GB HBM3e and 16,896 cores.

Loading...

Performance Rating

AMD Radeon Instinct MI325X outperforms NVIDIA B100 by 135.63% 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 B100

42.4

NVIDIA B100

42.4

Expert Comparison

AMD Radeon Instinct MI325X имеет значительное преимущество в пропускной способности памяти (10.3 TB/s против 4.10 TB/s) и FP32 вычислениях (81.72 TFLOPS против 32.95 TFLOPS). Это делает его более подходящим для задач машинного обучения и больших вычислений. В то же время, NVIDIA B100 предлагает больше ядер (16896 против 19456) и меньшую потребляемую мощность (1000 Вт), что может быть важным фактором при ограниченных ресурсах.

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 GB ×2 (192 ГБ)

Memory Type

HBM3e HBM3e

Memory Bandwidth

🔥 +151% 10.3 TB/s
4.10 TB/s ×2 (8.2 TB/s)

Memory Bus Width

8,192 бит 4,096 бит ×2 (8192 бит)

ML Performance

FP16 (Half Precision)

🔥 +396% 653.7 TFLOPS
131.8 TFLOPS

BF16 (Brain Float)

No No

TF32 (TensorFloat)

No No

Compute Power

FP32 (Single Precision)

🔥 +148% 81.72 TFLOPS
32.95 TFLOPS

FP64 (Double Precision)

🔥 +396% 81.72 TFLOPS
16.47 TFLOPS

CUDA Cores

🔥 +15% 19,456
16,896 ×2 (33792)

RT Cores

No No

Architecture & Compatibility

GPU Architecture

CDNA 3.0 Blackwell

SM (Streaming Multiprocessor)

No 132

PCIe Version

PCIe 5.0 x16 PCIe 5.0 x16

ML Software Support

CUDA Version

No 10.1

CUDA Toolkit (first supported)

v12

CUDA Toolkit status

Supported Supported

Clocks & Performance

Base Clock

🔥 +43% 1,000
700

Boost Clock

🔥 +115% 2,100
975

Memory Clock

🔥 +26% 2,525
2,000

Power Consumption

Recommended PSU

1400 W 1400 W

Power Connector

None No

TDP/TGP

1000 W 1000 W

Rendering

Texture Units (TMU)

🔥 +130% 1,216
528 ×2 (1056)

ROP

No No

L2 Cache

🔥 16 MB
50 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
🔥 SXM Module

Release Date

Oct. 12, 2024 No

Display Outputs

No outputs
No outputs

Renting is cheaper than buying