Why is GPU utilization not high when using GPU computing?
Why is GPU utilization not high when using GPU computing?
The model has a total of more than 400,000 particles. I used the entry-level GPU and the server's GPU for calculation, and there is almost no difference in the calculation time required. Not high, why is that? Do I need to have special settings in which document?
Comments
I'm so sorry I didn't upload the two photos just now
You do not provide enough information to help you fully.
Try increasing number of particles to 2 and 10 million and report back if the speed up is significant or not - just as a test.
Kind regards
@Asalih3d Thanks for your advice, I am trying as you suggested. In addition, what I want to ask is that my GPU utilization is too low, and the above graph shows only 11%. I want to ask if there are any special settings about GPU before simulation, or about GPU parameters, can achieve the purpose of improving GPU utilization.
We need to know the GPU you are using...
Can you share with us the first part of Run.out?
@Alex Thank you very much for your reply. The attachment above is the first run.out file (the number of particles is more than 400,000, specifically 493221)
Hi, I had a similar problem when accessing the GPU usage through the task manager.
To see the correct usage, open the motherboard application (for me it's MSI center application), and it will show you that the usage for the GPU is above 90%.
Check the attached photos for comparison:
@sam95 Thank you very much, it completely solved my doubt, it turns out that the way I checked the GPU utilization was wrong.
Thanks very much!🤞❤️
Really great Sam, thanks! Was not aware of this, kind of weird behaviour? But now we know at least!
Kind regards
In order to monitor the usage of a GPU you need to monitor CUDA. You will see this graph going up to 90% when in usage.