# GPU selection with Multiple GPUs on Server

**URL:** <https://forum.beamng.tech/t/gpu-selection-with-multiple-gpus-on-server/167>\
**Category:** General\
**Tags:** beamng-tech\
**Created:** [25 April 2025 18:31 UTC](https://forum.beamng.tech/t/gpu-selection-with-multiple-gpus-on-server/167 "2025-04-25T18:31:32Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![raymondsong00](https://forum.beamng.tech/user_avatar/forum.beamng.tech/raymondsong00/32/70_2.png) [@raymondsong00](https://forum.beamng.tech/u/raymondsong00)\
**Post date:** [25 April 2025 18:31 UTC](https://forum.beamng.tech/t/gpu-selection-with-multiple-gpus-on-server/167/1 "2025-04-25T18:31:32Z")

</div>

Hi,

I am trying to run BeamNG linux on a server that has multiple Nvidia GPUs.

Currently, the simulator launches in headless mode and defaults to the first GPU.

I would like to spin up multiple instances of Beam using all GPUs available.

Is there an option to select which GPU the simulator launches?

Raymond

---

<div class="post-metadata">

**Author:** ![aivora](https://forum.beamng.tech/user_avatar/forum.beamng.tech/aivora/32/13_2.png) [@aivora](https://forum.beamng.tech/u/aivora)\
**Post date:** [28 April 2025 21:35 UTC](https://forum.beamng.tech/t/gpu-selection-with-multiple-gpus-on-server/167/2 "2025-04-28T21:35:25Z")

</div>

Hi,  
we don’t expose an option to select a GPU at the moment, but you should be able to use a Docker container (we provide a template, see here: [Headless Mode](https://documentation.beamng.com/beamng_tech/headless_mode/#docker)) and set the `NVIDIA_VISIBLE_DEVICES` environment variable appropriately for each instance of BeamNG: [Specialized Configurations with Docker — NVIDIA Container Toolkit](https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/latest/docker-specialized.html).
