Intel Data Center GPU Max NEXT vs NVIDIA A100 PCIe 80 GB

Comparison of Intel Data Center GPU Max NEXT with 128 GB HBM2e and 20,480 cores vs NVIDIA A100 PCIe 80 GB with 80 GB HBM2e and 6,912 cores.

Loading...

Performance Rating

Intel Data Center GPU Max NEXT outperforms NVIDIA A100 PCIe 80 GB by 142.37% in the overall GPU ARK performance rating

A100 A100
H200 H200
MI325X MI325X

Intel Data Center GPU Max NEXT

62.1

Intel Data Center GPU Max NEXT

62.1
RX 7900 XTX RX 7900 XTX
MI250 MI250
Instinct MI300X Instinct MI300X

NVIDIA A100 PCIe 80 GB

25.6

NVIDIA A100 PCIe 80 GB

25.6

Contents:

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

Memory

Memory Size

128 ГБ
🔥 80 ГБ

Memory Type

HBM2e HBM2e

Memory Bandwidth

🔥 +65% 3.21 TB/s
1.94 TB/s

Memory Bus Width

8,192 бит 5,120 бит

ML Performance

FP16 (Half Precision)

65.54 TFLOPS
🔥 +19% 77.97 TFLOPS

BF16 (Brain Float)

No
🔥 311.84 TFLOPS

TF32 (TensorFloat)

No
🔥 155.92

Compute Power

FP32 (Single Precision)

65.54 TFLOPS
🔥 19.49 TFLOPS

FP64 (Double Precision)

65.54 TFLOPS
🔥 9.746 TFLOPS

CUDA Cores

20,480
🔥 6,912

RT Cores

160 No

Architecture & Compatibility

GPU Architecture

Generation 12.5 Ampere

SM (Streaming Multiprocessor)

No
🔥 108

PCIe Version

PCIe 5.0 x16 PCIe 4.0 x16

ML Software Support

CUDA Version

No 8.0

Clocks & Performance

Base Clock

900
🔥 +18% 1,065

Boost Clock

1,600
🔥 1,410

Memory Clock

1,565
🔥 1,512

Power Consumption

TDP/TGP

800 W
🔥 -62% 300 W

Recommended PSU

1200 W
🔥 -42% 700 W

Power Connector

8-pin EPS 8-pin EPS

Rendering

Texture Units (TMU)

1,280
🔥 432

ROP

160 No

L2 Cache

408 MB
🔥 80 MB

Benchmarks

MLPerf, llama2-70b-99.9 (UINT4)

527.5 tokens/s

MLPerf, llama3.1-8b (UINT4)

1 337 tokens/s

MLPerf, llama3.1-8b-edge (UINT4)

1 652 tokens/s

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

8.21 tokens/s

Additional

Slots

OAM Module Dual-slot

Release Date

No June 28, 2021

Display Outputs

No outputs
No outputs

Renting is cheaper than buying