Fixing external display issues on hybrid NVIDIA laptops (and a nice VRAM bonus)

On some Linux setups, especially with NVIDIA GPUs, it’s common to fall back to running everything on the dGPU. It “just works,” but it comes with downsides, higher power usage, limited flexibility, and in some cases, excessive VRAM usage.

A common symptom that pushes people into this setup is simple: you can’t get both the internal display and an external monitor working at the same time when hybrid graphics is enabled.

This usually isn’t a hardware limitation. It’s a configuration problem.

The root cause

If your system is set to hybrid graphics (via BIOS or a vendor tool like Lenovo Vantage), but Linux still behaves like it’s dGPU-only or won’t use internal/external displays correctly, the likely cause is an old Xorg configuration file forcing NVIDIA as the primary/only GPU.

Typical signs:

• prime-select is set to on-demand
• Both GPUs show up in lspci
• But glxinfo still reports NVIDIA
• External or internal displays don’t behave correctly
• High VRAM usage on the dGPU even at idle

The culprit is usually:

/etc/X11/xorg.conf

Especially if it was generated by nvidia-xconfig.

Why this breaks things

That file explicitly tells Xorg:

• “Use the NVIDIA GPU for everything”

This overrides Ubuntu’s PRIME setup entirely. As a result:

• The iGPU is ignored for display
• Hybrid mode doesn’t actually function
• Display routing becomes limited or incorrect
• VRAM usage stays high because the dGPU handles compositing

The fix

Remove the forced configuration and let Xorg auto-configure.

sudo mv /etc/X11/xorg.conf /etc/X11/xorg.conf.backup
reboot

That’s it.

After reboot, verify

Run:

glxinfo | grep "OpenGL renderer"

You should see your AMD (or Intel) iGPU, not NVIDIA.

Then:

nvidia-smi

You should see:

• much lower VRAM usage
• no heavy Xorg usage on the dGPU

What this fixes

• Internal + external displays working together properly
• Hybrid graphics behaving as intended
• dGPU no longer tied up doing desktop rendering
• Lower idle VRAM usage (often by hundreds of MB)

Side benefit: more VRAM for CUDA

If you run CUDA workloads, this matters.

With the dGPU no longer handling:

• display scanout
• compositing
• desktop rendering

you recover a significant chunk of VRAM that was previously unavailable.

Final note

Avoid using nvidia-xconfig or manually writing full Xorg configs on hybrid systems. Modern setups rely on dynamic configuration, and forcing a single GPU usually causes more problems than it solves.

If hybrid mode is enabled in firmware and prime-select is set to on-demand, the system should work without any custom Xorg configuration.