CUDA Programming - CPU performs faster than GPU

#1

Hi!
I was performing matrix multiplication in CUDA, and while comparing the performance of computations on CPU and GPU for large-sized arrays, turns out that the CPU performs better than the GPU. I was using a tiled matrix multiplication method and then using CUDA Event to monitor time. Uhmm, could you please share any resources to find out the reasons behind this performance change or share your thoughts as to why this could occur…
Thanks !

#2

hey @manmeetkaur0175 ,
A possible reason of this change can be , that Single CPU core works faster than a GPU core.
The faster performance of GPU is due to parallelism in performing tasks , so in your case your task might be getting performed on a single core and hence CPU is working faster then GPU.
There are some python libraries available that help to perform a task on GPU.
You can search about them and can try them , if they help you in your task.

I hope this helped You :slightly_smiling_face:.

#3

Thanks for replying!

#4

I guess your doubt is now resolved.
So i would request you kindly mark this doubt as resolved in your courses’s doubt section and also do provide your valuable feedback .

Thank You and Happy Learning :slightly_smiling_face:.

#5

Sorry sorry I forgot to mark the doubt as resolved. Thanks for clearing it!

#6

No worries.

Thank You and Happy Coding :slightly_smiling_face: .

1 Like