# Using Tesla K40c for rendering locally.

**URL:** https://discourse.paraview.org/t/using-tesla-k40c-for-rendering-locally/7207
**Category:** ParaView Support
**Created:** [May 11, 2021, 1:18pm UTC](https://discourse.paraview.org/t/using-tesla-k40c-for-rendering-locally/7207 "2021-05-11T13:18:06Z")
**Posts on this page:** 6
**Page:** 1

<div class="post-metadata">

### Author: ![ChrisX](https://discourse.paraview.org/letter_avatar_proxy/v4/letter/c/65b543/32.png) [@ChrisX](https://discourse.paraview.org/u/ChrisX)
#### Post date: [May 11, 2021, 1:18pm UTC](https://discourse.paraview.org/t/using-tesla-k40c-for-rendering-locally/7207/1 "2021-05-11T13:18:06Z")

</div>

Hello,

I make animations from calculations with ParaView and save the images as \*\*.png files.

My computer has 2 graphics cards:  
-Nvidia Quadro K620 for the screen.  
-Nvidia Tesla K40c for cuda calculations without screen.

For cuda calculations, the Tesla K40c card is always used automatically. ParaView, on the other hand, always uses the Quadro K620.  
This is much slower and makes the screen very sluggish.

So far I have been using a script via Pvserver as a workaround:

mpirun -n 1 /home/chris/Paraview/ParaView-5.9.1-RC1-egl-MPI-Linux-Python3.8-64bit/bin/pvserver --egl-device-index=0 --use-offscreen-rendering --disable-xdisplay-test --server-port=11111 &  
mpirun -n 1 /home/chris/Paraview/ParaView-5.9.1-RC1-egl-MPI-Linux-Python3.8-64bit/bin/pvserver --egl-device-index=0 --use-offscreen-rendering --disable-xdisplay-test --server-port=11112 & …

This way I can start several servers and open several windows of ParaView and run the animations in parallel.

This works with many windows but brings a considerable overhead.

Is there a direct way to run the calculations of ParaView exclusively on the Tesla K40c card (without starting Pserver)?

Greetings

Christian

---

<div class="post-metadata">

### Author: ![mwestphal](https://discourse.paraview.org/user_avatar/discourse.paraview.org/mwestphal/32/17_2.png) [@mwestphal](https://discourse.paraview.org/u/mwestphal)
#### Post date: [May 11, 2021, 1:27pm UTC](https://discourse.paraview.org/t/using-tesla-k40c-for-rendering-locally/7207/2 "2021-05-11T13:27:37Z")

</div>

Does your tesla card has a DISPLAY associated with it ?

---

<div class="post-metadata">

### Author: ![ChrisX](https://discourse.paraview.org/letter_avatar_proxy/v4/letter/c/65b543/32.png) [@ChrisX](https://discourse.paraview.org/u/ChrisX)
#### Post date: [May 12, 2021, 4:28am UTC](https://discourse.paraview.org/t/using-tesla-k40c-for-rendering-locally/7207/3 "2021-05-12T04:28:55Z")

</div>

No, Tesla has no output for a screen, its for calculation only. Nvidia x-server settings show only screen 0.

Christian

---

<div class="post-metadata">

### Author: ![mwestphal](https://discourse.paraview.org/user_avatar/discourse.paraview.org/mwestphal/32/17_2.png) [@mwestphal](https://discourse.paraview.org/u/mwestphal)
#### Post date: [May 14, 2021, 7:21am UTC](https://discourse.paraview.org/t/using-tesla-k40c-for-rendering-locally/7207/4 "2021-05-14T07:21:48Z")

</div>

You should be able to use EGL then: [ParaView: ParaView and Offscreen Rendering](https://kitware.github.io/paraview-docs/latest/cxx/Offscreen.html)

---

<div class="post-metadata">

### Author: ![ChrisX](https://discourse.paraview.org/letter_avatar_proxy/v4/letter/c/65b543/32.png) [@ChrisX](https://discourse.paraview.org/u/ChrisX)
#### Post date: [May 15, 2021, 10:24am UTC](https://discourse.paraview.org/t/using-tesla-k40c-for-rendering-locally/7207/5 "2021-05-15T10:24:35Z")

</div>

Thanks for the answer,

unfortunately the link doesn’t help me since I want to render locally. I found another similar link:

> **[Off-screen Rendering through the Native Platform Interface (EGL) - Kitware Blog](https://blog.kitware.com/off-screen-rendering-through-the-native-platform-interface-egl/)**
>
> Motivation ParaView can run on a supercomputer with thousands of nodes to provide visualization and analysis of very large datasets. In this configuration, the same version of ... Read More

I can’t do anything with this link either. As I wrote in the first post I can use the workaround via pvserver and use egl.  
But this is exactly what I want to avoid because of the overhead.

The compiled binaries with “egl” do not include a “paraview” client. The versions with “paraview” client do not contain egl.

Therefore I need the exact procedure to start (or configure) the “paraview” client so that the paraview client can choose the graphics card to be used locally for rendering.

TESLA = 0  
QUADRO = 1

Greetings

Christian

---

<div class="post-metadata">

### Author: ![mwestphal](https://discourse.paraview.org/user_avatar/discourse.paraview.org/mwestphal/32/17_2.png) [@mwestphal](https://discourse.paraview.org/u/mwestphal)
#### Post date: [May 15, 2021, 10:28am UTC](https://discourse.paraview.org/t/using-tesla-k40c-for-rendering-locally/7207/6 "2021-05-15T10:28:04Z")

</div>

You cannot use a graphic card without a display and not use pvserver.

You can use pvserver locally.
