localhost Multicore Support / select second GPU on local maschine

ParaView Support
1

Hello!
I have two very simple questions, but I can not find a solution. Here in the forum I could not find to it, on the Internet there is apparently also no one who uses my configuration.
OS: Windows 10
Pareaview 5.10 RC1
PC: Workstation , Intel, 6 cores, 12 threads.
First GPU is on-board Intel. second GPU is nvidia Quadro R2000

Question 1:
On my workstation I create and calculate my simulations with OpenFOAM. No HPC is used! Therefore I want to use my PC power on my workstation for post processing with paraview.
Until Paraview version 5.9 you could activate the local MPI under Edit/Settings/General the area under Multicore Support. Since V 5.10 this is no longer available.
Solution 1: Install version 5.9 in parallel and work with it. But bugfixes and new functions from 5.10 are missing.
Solution 2: Start the pvserver on the wokstation via the console and connect paraview locally. Is this the way? Or do I understand something wrong?

Question 2:
In my workstation is an Intel GPU on-board. And a nvidia Quadro P2000. The GUI of paraview always takes the Intel GPU, but it is rather weak. I am not talking about nvidia OptiX for raytracing or IndeX for volume rendering. I just want to use the GPU power of my nvidia for normal rendering iun the GUI, streamlines, slices, … How do I get radiance in client mode on my workstation to use the nvidia GPU and not the intel GPU? At the moment, evaluating simulations for large cases in the GUI is very tedious. And not every user has a HPC in the background. :slight_smile:

thx, eike

2

ouch. Question 2 has just been clarified. When I work directly on my PC, the nvidia GPU is also used for the GUI. But not when accessing the PC remotely, then apparently the GUI is processed via the intel GPU.
Remains the problem, how do I get my GUI to use the nvidia via remote protocol in the home office. :smiley:

3

This option was buggy and unoptimized.

Start the pvserver on the wokstation via the console and connect paraview locally. Is this the way? Or do I understand something wrong?

This is the way to go

but not when accessing the PC remotely, then apparently the GUI is processed via the intel GPU.

Remote desktop is generally not the right solution, why not running the pvserver on the PC and connecting to it with ParaView ?

4

Hello!
Thanks for the quick reply!
Remote desktop is necessary because my workstation is in the office. And because of the pandemic the virtual way from home is via VPN and remote desktop to the office.
Since I sit at home in front of a Linux, I would have to try to find the PC there via VPN. Are two networks, could be tricky :wink: Let’s see.
many greetings, eike

5

Then you need to make you remote desktop solution able to user your nvidia GPU driver, not related to ParaView though.

6

Hi CDF-ZWP, Hi Mathieu,

As far as I understand, it is possible to do this by explicit writing the display number when running pvserver remotely. E.g. by command:

./mpiexec -np 6 ./pvserver -display :0.0

where :0.0 is the number of proper display for NVidia GPU.
Am I right?

Best Regards,
Pavel