STAR WARS: Episode I Racer - Horrible performance with NVIDIA #48

Open
opened 2025-11-12 14:02:14 +00:00 by dCo3lh0 · 4 comments

I have a hybrid setup in my linux machine a integrated Ryzen 7 5700G and a NVIDIA RTX 2060, Star Wars Racer behaves really well in Ryzen, goes 60 fps and maintains there, but on my RTX 2060 which is much stronger its 25 fps and simply behaves badly with Nvidia, why is it that this happens? Its just this game.

I have a hybrid setup in my linux machine a integrated Ryzen 7 5700G and a NVIDIA RTX 2060, Star Wars Racer behaves really well in Ryzen, goes 60 fps and maintains there, but on my RTX 2060 which is much stronger its 25 fps and simply behaves badly with Nvidia, why is it that this happens? Its just this game.

what's up brother. Vulkan is not handled well with nvidia cards apparently. This seems to have worked for many. try these solutions. let me know any of this fixes that for you.


1. Run Citron using PRIME offloading

On hybrid systems Citron will NOT use NVIDIA correctly unless you launch it like this:

prime-run citron

If you don’t use prime-run, Citron often:

loads wrong Vulkan ICD

fails to initialize NV pipeline cache

falls back to integrated GPU paths

or mixes Mesa/Nvidia layers → massive slowdown

This ONE step fixes the issue for 70% of users.

2. Set NVIDIA as the ONLY Vulkan ICD for Citron

Check your Vulkan ICD list:

ls /usr/share/vulkan/icd.d

You want ONLY your NVIDIA JSON active.
If you see BOTH AMD and NVIDIA ICDs, Citron may grab the wrong one.

To force NVIDIA:

sudo mv /usr/share/vulkan/icd.d/amd_icd64.json /usr/share/vulkan/icd.d/amd_icd64.json.bak

(You can restore it later.)

3. Install the newest NVIDIA Vulkan drivers

RTX 2060 has HUGE performance differences depending on driver branch.

Use 550+ (not 470, 510, or 525).

4. Force Async Shader Building ON in Citron

In Citron:

Async Shaders: ON

Pipeline Cache: ON

Accurate GPU Emulation: OFF

VSync: OFF

Fast Memory: ON

These settings massively reduce NVIDIA stutter.

5. Delete all existing shader/pipeline caches

Old Yuzu/Ryujinx-style pipeline caches break Citron on NVIDIA.

Delete:

rm -r ~/.local/share/citron/shader_cache
rm -r ~/.cache/citron

  1. Use the Flatpak version instead

Flatpak packages its own Vulkan loader which removes hybrid-GPU ICD conflicts.

Most people with hybrid laptops/PCs get better NVIDIA performance this way.

what's up brother. Vulkan is not handled well with nvidia cards apparently. This seems to have worked for many. try these solutions. let me know any of this fixes that for you. ------------------------------------------------------------------------------------------------------------------ **1. Run Citron using PRIME offloading** On hybrid systems Citron will NOT use NVIDIA correctly unless you launch it like this: prime-run citron If you don’t use prime-run, Citron often: loads wrong Vulkan ICD fails to initialize NV pipeline cache falls back to integrated GPU paths or mixes Mesa/Nvidia layers → massive slowdown This ONE step fixes the issue for 70% of users. **2. Set NVIDIA as the ONLY Vulkan ICD for Citron** Check your Vulkan ICD list: ls /usr/share/vulkan/icd.d You want ONLY your NVIDIA JSON active. If you see BOTH AMD and NVIDIA ICDs, Citron may grab the wrong one. To force NVIDIA: sudo mv /usr/share/vulkan/icd.d/amd_icd64.json /usr/share/vulkan/icd.d/amd_icd64.json.bak (You can restore it later.) **3. Install the newest NVIDIA Vulkan drivers** RTX 2060 has HUGE performance differences depending on driver branch. Use 550+ (not 470, 510, or 525). **4. Force Async Shader Building ON in Citron** In Citron: Async Shaders: ON Pipeline Cache: ON Accurate GPU Emulation: OFF VSync: OFF Fast Memory: ON These settings massively reduce NVIDIA stutter. **5. Delete all existing shader/pipeline caches** Old Yuzu/Ryujinx-style pipeline caches break Citron on NVIDIA. Delete: rm -r ~/.local/share/citron/shader_cache rm -r ~/.cache/citron 6. Use the Flatpak version instead Flatpak packages its own Vulkan loader which removes hybrid-GPU ICD conflicts. Most people with hybrid laptops/PCs get better NVIDIA performance this way.
Author

i can't setup this icd thing because if i do it, i will probably break my system since i have a hybrid GPU desktop setup, I'm connected through the motherboard GPU inputs, my GPU doesn't have a single cable connected. i think my best option would be using the Flatpak to test if its really this issue, but my issue is, which flatpak? i didn't find anywhere, not in flathub, not here, can you help me?

i can't setup this icd thing because if i do it, i will probably break my system since i have a hybrid GPU desktop setup, I'm connected through the motherboard GPU inputs, my GPU doesn't have a single cable connected. i think my best option would be using the Flatpak to test if its really this issue, but my issue is, which flatpak? i didn't find anywhere, not in flathub, not here, can you help me?
Author

found the github repo, building here, since it doesn't have a pre-compiled here.

found the github repo, building here, since it doesn't have a pre-compiled here.

I am not super familiar with that situation / OS unfortunately, hope you get it locked in / resolved.

I am not super familiar with that situation / OS unfortunately, hope you get it locked in / resolved.
Sign in to join this conversation.
No labels
No milestone
No project
No assignees
2 participants
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
Citron/Emulator#48
No description provided.