mariyasha/cuda_speedtest

A notebook testing CPU speed vs GPU speed with Pytorch and CUDA

CUDA_speedtest

A notebook testing CPU speed vs GPU speed with Pytorch and CUDA


This notebook was uploaded to support a video tutorial on YouTube:
https://youtu.be/r9IqwpMR9TE

The hardware I've tested:
CPU : Intel's i9-12900K 12th-Gen Alder Lake Processor (+5200mhz DDR5 memory)
GPU : Nvidia GeForce RTX 3090

A special thank you to Robert from Nvidia for helping me with this notebook!

cuda_speedtest's Issues

loading thumbnail image

YouTube comment

Hi Mariya.
I apologise for contacting you this way, but I have been unable to add my comment to your video. I've tried several times in Firefox, and also in Chrome. The comment shows as posted, but when I refresh the page it's gone again!

Your videos are so very informative and clearly delivered. Thank you.

My results. What a massive speed difference. I can see why cryptocurrency calculations use GPU!
Running in PyCharm on laptop.
Intel Core i7 7700K Kaby Lake-S
Memory 64Gb DDR4 SDRAM 1197.2MHz
nVidia GeForce GTX1080 8Gb GDDR5X SDRAM
CUDA Version 11.3
NVIDIA CUDA Cores - 2560
CPU - 66% utilization, clock at 3095MHz
GPU - 4.6% utilization, 1898MHz

Using cuda device
********* CPU SPEED *********
37.804046630859375
Verify device: cpu
********* GPU SPEED *********
1.8232934474945068
Verify device: cuda:0
********* GPU SPEED *********
1.4021873474121094
Verify device: cuda:0
********* GPU SPEED *********
1.3883006572723389
Verify device: cuda:0