NVIDIA H200 SXM 141 GB vs Palit GTX 670

Comparison of NVIDIA H200 SXM 141 GB with 141 GB HBM3e and 16,896 cores vs Palit GTX 670 with 2 GB GDDR5 and 1,344 cores.

Loading...

Performance Rating

H200 H200
MI325X MI325X
A100 A100

NVIDIA H200 SXM 141 GB

NVIDIA H200 SXM 141 GB

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

Palit GTX 670

Palit GTX 670

Contents:

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

Memory

Memory Size

141 GB 2 GB

Memory Type

HBM3e GDDR5

Memory Bandwidth

4.89 TB/s 192.3 GB/s

Memory Bus Width

6,144 бит 256 бит

ML Performance

FP16 (Half Precision)

267.6 TFLOPS No

BF16 (Brain Float)

No No

TF32 (TensorFloat)

No No

Compute Power

FP32 (Single Precision)

66.91 TFLOPS 2.634 TFLOPS

FP64 (Double Precision)

33.45 TFLOPS 0.1098 TFLOPS

CUDA Cores

16,896 1,344

RT Cores

No No

Architecture & Compatibility

GPU Architecture

Hopper Kepler

SM (Streaming Multiprocessor)

132 No

PCIe Version

PCIe 5.0 x16 PCIe 3.0 x16

ML Software Support

CUDA Version

🔥 9.0
3.0

Clocks & Performance

Base Clock

1,500 915

Boost Clock

1,980 980

Memory Clock

1,593 1,502

Power Consumption

TDP/TGP

700 W
🔥 -76% 170 W

Recommended PSU

1100 W
🔥 -59% 450 W

Power Connector

8-pin EPS 2x 6-pin

Rendering

Texture Units (TMU)

528 112

ROP

No No

L2 Cache

50 MB 512 KB

Benchmarks

MLPerf, llama2-70b-99.9 (UNSET)

3 534 tokens/s

MLPerf, llama2-70b-99.9 (fp16)

3 553 tokens/s

MLPerf, llama2-70b-99.9 (fp8)

2 444 tokens/s

MLPerf, llama3.1-405b (fp16)

40.8 tokens/s

MLPerf, llama3.1-405b (fp8)

25.3 tokens/s

MLPerf, llama3.1-8b (fp8)

5 161 tokens/s

MLPerf, deepseek-r1 (fp8)

1 113 tokens/s

MLPerf, mixtral-8x7b (fp8)

7 132 tokens/s

Additional

Slots

SXM Module Dual-slot

Release Date

Nov. 18, 2024 May 10, 2012

Display Outputs

No outputs
2x DVI
1x HDMI 1.4a
1x DisplayPort 1.2

Renting is cheaper than buying