Dual NVIDIA GPU Configuration Guide
RandR Multi-Monitor Setup with Different Display Specifications
Overview
This guide provides complete instructions for configuring dual discrete NVIDIA GPUs on X.org, where each GPU independently drives a display with different resolutions, refresh rates, or physical specifications. This configuration uses the modern RandR (Resize and Rotate) extension rather than legacy technologies like BaseMosaic or Xinerama.
Use Case
This configuration is appropriate when:
- You have two discrete NVIDIA GPUs installed
- Each GPU is connected to one physical display
- Displays have different resolutions or refresh rates
- You need mouse movement and window dragging between displays
- Neither GPU is used for GPGPU or compute workloads
This setup provides a unified desktop experience where each monitor runs at its native resolution and the mouse can move freely between displays.
System Requirements
Hardware Setup Checklist
- Both NVIDIA GPUs properly seated in PCIe slots
- Each GPU connected to exactly one display via HDMI/DP/DVI
- Power connectors attached to both GPUs
- BIOS/UEFI settings configured (see below)
BIOS/UEFI Configuration
- Secure Boot: Disabled (or MOK keys enrolled for NVIDIA driver)
- Above 4G Decoding: Enabled (if available)
- Primary Display Adapter: Set to PCIe
- Integrated Graphics: Disabled (if present)
NVIDIA Driver Installation
Remove Existing Drivers
Remove any existing NVIDIA drivers and the nouveau driver:
# For Debian/Ubuntu
sudo apt remove --purge '*nvidia*'
# For Fedora/RHEL
sudo dnf remove '*nvidia*'
# For Arch
sudo pacman -R nvidia nvidia-utils
Blacklist Nouveau Driver
Create a blacklist configuration file:
echo "blacklist nouveau" | sudo tee /etc/modprobe.d/blacklist-nouveau.conf
echo "options nouveau modeset=0" | sudo tee -a /etc/modprobe.d/blacklist-nouveau.conf
Update initramfs:
# Debian/Ubuntu
sudo update-initramfs -u
# Fedora/RHEL/Arch
sudo dracut --force
Install NVIDIA 580 Proprietary Driver
Method 1: Distribution Repository (Recommended)
# Debian/Ubuntu
sudo apt update
sudo apt install nvidia-driver-580 nvidia-settings
# Fedora (RPM Fusion)
sudo dnf install akmod-nvidia xorg-x11-drv-nvidia-cuda
Method 2: NVIDIA .run Installer
# Download from nvidia.com
chmod +x NVIDIA-Linux-x86_64-580.xx.xx.run
sudo ./NVIDIA-Linux-x86_64-580.xx.xx.run
- Reboot the system after driver installation
Verify Driver Installation
After rebooting, verify the driver loaded successfully:
# Check kernel modules
lsmod | grep nvidia
# Verify both GPUs detected
nvidia-smi
Expected output: nvidia-smi should show two GPUs with their model names and driver version 580.xx.
X.org Configuration
Identify GPU PCI Bus IDs
Find the PCI bus ID for each GPU:
lspci | grep -i 'vga\|3d\|display'
Example output:
01:00.0 VGA compatible controller: NVIDIA Corporation ...
02:00.0 VGA compatible controller: NVIDIA Corporation ...
Convert to X.org format: 01:00.0 becomes PCI:1:0:0 (remove leading zeros from bus number).
Create /etc/X11/xorg.conf
Backup any existing configuration:
sudo cp /etc/X11/xorg.conf /etc/X11/xorg.conf.backup 2>/dev/null || true
Create the configuration file with the following content:
sudo nano /etc/X11/xorg.conf
Complete xorg.conf Configuration
Replace the PCI:1:0:0 and PCI:2:0:0 values with your actual bus IDs:
Section "ServerLayout"
Identifier "Layout0"
Screen 0 "Screen0" 0 0
Option "Xinerama" "0"
EndSection
Section "Device"
Identifier "Device0"
Driver "nvidia"
VendorName "NVIDIA Corporation"
BusID "PCI:1:0:0" # Replace with your first GPU
Screen 0
EndSection
Section "Device"
Identifier "Device1"
Driver "nvidia"
VendorName "NVIDIA Corporation"
BusID "PCI:2:0:0" # Replace with your second GPU
Screen 0
EndSection
Section "Monitor"
Identifier "Monitor0"
EndSection
Section "Monitor"
Identifier "Monitor1"
EndSection
Section "Screen"
Identifier "Screen0"
Device "Device0"
Device "Device1"
Monitor "Monitor0"
Monitor "Monitor1"
DefaultDepth 24
Option "AllowEmptyInitialConfiguration" "True"
Option "UseDisplayDevice" "none"
SubSection "Display"
Depth 24
EndSubSection
EndSection
Configuration Notes
- AllowEmptyInitialConfiguration: Allows X to start without predetermined resolution
- UseDisplayDevice "none": Disables static configuration, enabling RandR control
- Xinerama "0": Disabled (incompatible with modern compositing and RandR)
- One Screen section with two Device entries: Unified desktop across both GPUs
Display Configuration with RandR
Test X.org Configuration
Before restarting the display manager, test the configuration:
sudo X -config /etc/X11/xorg.conf -retro
Press Ctrl+Alt+Backspace to exit. Check for errors:
grep EE /var/log/Xorg.0.log
Restart Display Manager
# GDM (GNOME)
sudo systemctl restart gdm
# SDDM (KDE)
sudo systemctl restart sddm
# LightDM (XFCE/others)
sudo systemctl restart lightdm
Configure Displays with xrandr
After logging in, list available displays:
xrandr
Example output:
Screen 0: minimum 8 x 8, current 4480 x 1440, maximum 16384 x 16384
DP-0 connected primary 2560x1440+0+0 (normal left inverted right x axis y axis)
2560x1440 144.00*+ 59.95
DP-2 connected 1920x1080+2560+0 (normal left inverted right x axis y axis)
1920x1080 60.00*+
Set Display Layout and Resolutions
Configure each display with its native resolution and position:
# Set first display (left monitor)
xrandr --output DP-0 --mode 2560x1440 --rate 144 --primary --pos 0x0
# Set second display (right monitor)
xrandr --output DP-2 --mode 1920x1080 --rate 60 --right-of DP-0
Alternative: Configure both in one command:
xrandr --output DP-0 --mode 2560x1440 --rate 144 --primary --pos 0x0 \
--output DP-2 --mode 1920x1080 --rate 60 --pos 2560x0
Position Options
Making Display Settings Permanent
Display configuration needs to be applied on every login. Choose one method:
Method 1: Desktop Environment Settings (Recommended)
Use your desktop environment's display configuration tool:
- GNOME: Settings → Displays → Arrange displays and click Apply
- KDE Plasma: System Settings → Display Configuration
- XFCE: Settings → Display
These tools save settings to ~/.config/monitors.xml or similar configuration files.
Method 2: Autostart Script
Create an xrandr script to run on login:
nano ~/.xprofile
Add the following content (adjust for your displays):
#!/bin/bash
xrandr --output DP-0 --mode 2560x1440 --rate 144 --primary --pos 0x0 \
--output DP-2 --mode 1920x1080 --rate 60 --pos 2560x0
Make it executable:
chmod +x ~/.xprofile
Method 3: Static X.org Configuration (Not Recommended)
You can add static resolution settings to xorg.conf, but this is less flexible and harder to modify. The RandR approach (Method 1 or 2) is preferred.
Verification and Testing
Verification Checklist
- Both displays are active and showing content
- Each display running at its native resolution
- Mouse pointer moves freely between displays
- Windows can be dragged between displays
- Each GPU driving its assigned display (verify in nvidia-smi)
Verification Commands
Check GPU status and driver version:
nvidia-smi
Verify display configuration:
xrandr --listproviders
xrandr --verbose
Check screen count (should be 1):
xdpyinfo | grep "number of screens"
Test OpenGL rendering on each display:
glxinfo | grep 'OpenGL renderer'
glxgears # Move window between displays
Troubleshooting
X Server Fails to Start
- Check X.org logs: grep EE /var/log/Xorg.0.log
- Verify BusID matches lspci output (remove leading zeros)
- Test configuration: sudo X -config /etc/X11/xorg.conf -retro
- Check kernel module loading: dmesg | grep nvidia
One Display Not Working
- Verify cable connections and monitor power
- Check display detection: xrandr
- Try swapping displays between GPUs to isolate hardware issue
- Verify both GPUs visible: nvidia-smi
Mouse Cannot Move Between Displays
- Verify Xinerama is disabled: grep Xinerama /etc/X11/xorg.conf
- Check screen count: xdpyinfo | grep screens (should be 1)
- Verify unified desktop: xrandr (should show relative positions)
Displays Running at Wrong Resolution
- List available modes: xrandr
- Manually set resolution: xrandr --output DP-0 --mode 2560x1440
- Save settings in desktop environment or .xprofile
Screen Tearing
Enable ForceCompositionPipeline for each display:
nvidia-settings --assign CurrentMetaMode="DP-0: nvidia-auto-select +0+0 {ForceCompositionPipeline=On}, DP-2: nvidia-auto-select +2560+0 {ForceCompositionPipeline=On}"
Or add to xorg.conf Device section:
Option "metamodes" "DP-0: nvidia-auto-select +0+0 {ForceCompositionPipeline=On}, DP-2: nvidia-auto-select +2560+0 {ForceCompositionPipeline=On}"
Display Configuration Not Persistent
- Verify .xprofile is being executed: add echo statement
- Check desktop environment display settings are saved
- Ensure xrandr commands complete without errors
Optional Optimizations
Power Management
Set adaptive power mode:
nvidia-settings -a GPUPowerMizerMode=1
Add to .xprofile to make permanent.
Per-Monitor DPI Scaling
For displays with different pixel densities:
xrandr --output DP-0 --scale 1x1 --dpi 109
xrandr --output DP-2 --scale 1x1 --dpi 92
Note: Perfect per-monitor DPI scaling on X11 is challenging. Wayland handles this better if you're willing to switch display servers.
Disable Unused Features
Add to Device sections in xorg.conf:
Option "AllowGPUCUDA" "0"
Since you're not using GPGPU/compute, this can reduce overhead.
Monitor GPU Temperature
watch -n 1 nvidia-smi
Additional Resources
- NVIDIA Driver Documentation: https://download.nvidia.com/XFree86/Linux-x86_64/
- X.org RandR Documentation: https://www.x.org/wiki/Projects/XRandR/
- Arch Wiki NVIDIA: https://wiki.archlinux.org/title/NVIDIA
- NVIDIA Forums: https://forums.developer.nvidia.com/
Summary
This configuration provides a unified desktop across two NVIDIA GPUs, where:
- Each GPU independently drives one display
- Each display runs at its native resolution and refresh rate
- Mouse and windows move seamlessly between displays
- Modern RandR provides flexible, dynamic display management
- No deprecated technologies like Xinerama or BaseMosaic required
This setup is optimal for dual-GPU workstations with heterogeneous displays and provides the best balance of flexibility, performance, and user experience.
No comments:
Post a Comment