Regarding GPU memory and Heating issue

[Vijay R]

Hello Guys,

I am currently running MuMax3 simulations using an NVIDIA RTX 2000 Ada GPU with 16 GB memory. Before this, I mainly used Google Colab for my simulations.

However, I noticed that the simulation runtime on my local GPU is almost the same as what I experienced on Google Colab. In addition, while running simulations, my GPU temperature reaches around 90°C, even though the system is kept in an air-conditioned room. Typically, my simulations take around 3–4 hours to complete.

I have a couple of questions:

1. Is it normal/safe for the GPU temperature to reach ~90°C while running MuMax3 simulations continuously for several hours?

2. Is it possible to add another GPU(higher variant lets say RTX 4000 series) along with the existing one to improve the simulation speed? If yes, does MuMax3 support utilizing multiple GPUs effectively?

I would appreciate any suggestions or insights from the community.

Thank you!

[Антон Луценко]

1. It is better to try lowering the temperatures. 

2. Mumax can only use one GPU per simulation, so you won't be able to improve the speed by adding more GPUs. 

[Vijay R]

Thanks for your quick response and kind suggestion.

So, if I understood correctly, you are suggesting that instead of using an additional higher-variant GPU together with the existing one, it would be better to use the higher-variant GPU alone. Right?

Also, could you please give me some suggestions on how to reduce the GPU temperature while running long MuMax3 simulations?

Once again, thank you so much for your valuable feedback.

regards 

Vijay 

[Антон Луценко]

Well, having two GPUs would mean that you can run two simulations in parallel. So if you are planning to run multiple simulations (e.g. parameter sweeps), then you would benefit from having two GPUs rather than one. If you ONLY want to run a single simulation at any time, then having a better GPU would mean faster simulation. If you can access computational cluster, that would probably bring the biggest boost to your simulation speed. 

While this forum is probably not the best place to discuss thermal management for a PC, there's a lot of information online. In general, you can either try to manage the airflow (even just rearranging the internals and fans helps a lot; some had significant success with 3D-printed air ducts to guide the air streams), or to switch to liquid cooling (though that would require more investment and in general more intrusive and demanding). 

[Vijay R]

Completely understood.

Thank you so much for your help and assistance.

[Felipe Garcia]

Hi,

Adding to what has already been said. 90 degrees Celsius during simulations is not so bad. It is better to have in the 80s degrees Celsius. Going for example beyond that will reduce the life of the GPU. Normally, they become obsolete before that happens, so that is not a really big issue. There is a second consequence. If it enters in the range of 90s degrees, then it may have throttling, which means that it reduces speed in order to reduce heat. This is why, in an old laptop, when the fans become dirty, the fans enter quickly and the laptop slows down. Obviously, one does not want that. It does not depend only on the external temperature. It depends also on the thermal dissipation of your GPU and the airflow in your server box.

What I mean is that if after adding the two GPUs to your server, the temperature increases and reaches for example 100 ºC, then I would look for another box or at least remove the least powerful one. Definitely, I would check the temperature in any case after adding the two GPUs.

Best regards,

Felipe

--
You received this message because you are subscribed to the Google Groups "mumax2" group.
To unsubscribe from this group and stop receiving emails from it, send an email to mumax2+un...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/mumax2/e86a2793-e117-43ff-9c59-bb3e24e62ab8n%40googlegroups.com.

[Vijay R]

  Thanks a lot for your suggestion. I really appreciate your help and guidance.