Codex Desktop Linux: severe lag and high CPU/GPU usage on Hyprland with hybrid graphics
Nobody has claimed this yet.
- Dominant language
- Rust
- Stars
- 125k
- Forks
- 19.5k
- PR merge metrics
- PR metrics pending
Description
What version of the Codex App are you using (From “About Codex” dialog)?
Powered by Codex & OWL Version 26.803.81509 Released Aug 11, 2026
What subscription do you have?
Plus
What platform is your computer?
Linux 7.1.8-200.fc44.x86_64 x86_64 unknown
What issue are you seeing?
Codex Desktop is severely laggy immediately after launch and causes unusually high CPU/GPU power use on my Linux laptop.
On a fresh launch, I measured:
- Codex processes: ~134% combined CPU and ~2.5 GiB RAM
- Codex GPU process: ~40% NVIDIA GPU utilization
- NVIDIA RTX 3050 Mobile: ~28 W power draw
- Laptop temperature reached about 90°C before troubleshooting
- The UI is visibly laggy while typing and using the app
The issue persists after closing Telegram and relaunching Codex.
Attempted mitigations:
- Fully quit Codex and relaunch with
DRI_PRIME=0- Codex still used the NVIDIA GPU.
- Launch with
--disable-gpu- The app failed/crashed and printed a JavaScript stack trace.
Hardware:
- Intel Iris Xe integrated GPU
- NVIDIA GeForce RTX 3050 Mobile discrete GPU
- Hyprland session
What steps can reproduce the bug?
- Start a Hyprland Linux session on a hybrid Intel/NVIDIA laptop.
- Install and launch the official Codex Desktop Linux RPM.
- Open Codex and wait about one minute.
- Observe severe UI lag.
- Monitor processes with
top/htopand GPU activity withnvidia-smi.
Expected: Codex should remain responsive and use reasonable CPU/GPU power while idle or during normal use.
Actual: Codex uses roughly 134% CPU, ~2.5 GiB RAM, and substantial NVIDIA GPU power; the laptop becomes hot and the UI is laggy.
What is the expected behavior?
Codex should remain responsive during normal use and should not sustain high CPU/GPU utilization or heat the laptop to around 90°C when no intensive task is running.
Additional information
I installed the official RPM from https://openai.com/codex/
System observed:
- NVIDIA GPU temperature: ~61–62°C during testing
- CPU package temperature varied up to ~83°C during testing
- Hyprland also used the NVIDIA GPU, but Codex was independently using substantial CPU and GPU resources.
Potentially related reports, but not exact duplicates:
- openai/codex#37638: similar ~40% NVIDIA GPU use attributed to the Codex Desktop UI on a hybrid-GPU laptop, but on Windows.
- openai/codex#13749: Codex visible-window/compositor GPU issue on Windows, but different display stack and symptoms.
- #10885 / openai/codex#10432: Codex Desktop CPU/GPU power and heat reports on macOS.
This report is specifically for the official Linux RPM on Hyprland with Intel Iris Xe + NVIDIA RTX 3050 Mobile. It includes severe UI lag, sustained CPU/GPU usage, and high laptop temperature.
I can provide sanitized logs or more measurements if needed.
User feedback: Linux app consumes high CPU/GPU while idle
- Source: https://x.com/BadGatewayCo/status/2090838316237164983
- Source: https://x.com/BadGatewayCo/status/2090840817120190619
- Source: https://x.com/BadGatewayCo/status/2090844564097527974
- Type: external
- User reports:
- “Soon as it launches its just eating CPU while doing nothing.”
- “The Codex Electron UI is consuming roughly 123% CPU: its main process is ~58% and one renderer is ~65%; GPU is another ~10%.”
- “GPU utilization: 34–35%” and “Power draw: ~19 W.”
- “chatgpt --ozone-platform=wayland is a little better, but still 10-15% of gpu while totally.”
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
Reproduce on the official Linux RPM under Hyprland with Intel Iris Xe and RTX 3050 hybrid graphics, using top/htop and nvidia-smi as described. Compare normal launch with DRI_PRIME=0 and --disable-gpu, recording the UI lag, CPU/GPU usage, power and temperatures. Done means the idle app remains responsive without sustained excessive resource use, with the relevant logs or measurements captured.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- electron, linux
- Domain
- desktop, operating-systems, performance
- Issue type
- Bug
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Quiet
- Clarity
- Mostly clear
- Newbie friendliness
- 48/100