Free tools Windows power users keep installed
One-click scans. No signup required.
Yes—on a compatible hybrid-graphics computer, an AI application can run on the discrete GPU (dGPU) while a display connected to the integrated GPU (iGPU) continues to use the iGPU’s display pipeline. The display cable does not select the GPU that computes AI: configure the application to use the dGPU, then verify that it actually does. The display’s physical wiring and any hardware mux determine which GPU handles scanout.
Separate AI processing from display output
There are two different GPU choices involved:
- Compute or rendering: which GPU runs the AI workload or renders an application. An operating-system preference or launch option can influence this choice.
- Display scanout: which GPU’s display pipeline sends the image to a panel or connector. This depends on the computer’s electrical design and, where present, its display mux.
NVIDIA’s Optimus Developer Guide describes a supported arrangement in which the NVIDIA GPU renders frames and transfers them to the integrated GPU’s display pipeline when the display is connected to the IGP: “If the driver decides to run the application on the NVIDIA GPU, the final rendered frames are copied to the IGP’s display pipeline for scanout.” NVIDIA also says CUDA applications can discover and create a context on the NVIDIA GPU even when it is not the primary display device.
This is not a universal way to reroute a port. A GPU preference can select an application’s compute or rendering device; it cannot change which GPU a particular HDMI, DisplayPort, USB-C port, or internal panel is physically wired to.
Choose the dGPU for the AI application
Windows: set an app-level graphics preference
- Open Settings > System > Display > Graphics Settings. On some Windows versions or computers, the page or labels may differ.
- Add the desktop application or select it from the app list, then open Options.
- Choose the high-performance option if it identifies the dGPU you want to use, and save the choice.
- Run the workload and check Task Manager > Performance. The GPU Engine column can help show which GPU an application is using.
These menu steps are documented in Microsoft’s Surface Book 3 graphics-preference guidance; treat them as a Windows example, since labels and availability can vary by PC and Windows version. A graphics preference is not a guarantee that every application will use a particular adapter. Microsoft defines DXGI_GPU_PREFERENCE_HIGH_PERFORMANCE as a preference for the highest-performing GPU, such as a dGPU or eGPU, and DXGI_GPU_PREFERENCE_MINIMUM_POWER as a preference for the minimum-powered GPU, such as an iGPU, in its DXGI GPU preference documentation.
#1 Best Overall
- AI Performance: 767 AI TOPS
- OC mode: 2632 MHz (OC mode)/ 2602 MHz (Default mode)
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Axial-tech fan design features a smaller fan hub that facilitates longer blades and a barrier ring that increases downward air pressure
- A 2.5-slot design maximizes compatibility and cooling efficiency for superior performance in small chassis
Linux: launch with NVIDIA PRIME offload
On a Linux system using NVIDIA PRIME, start by listing the available GPUs and their indices:
switcherooctl list
Then launch the target application on the selected GPU, replacing <index> with the appropriate index and <command> with the application command:
Rank #2
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Powered by GeForce RTX 5070 Ti
- Integrated with 16GB GDDR7 256bit memory interface
- PCIe 5.0
- WINDFORCE cooling system
switcherooctl launch -g <index> <command>
NVIDIA documents this method in its PRIME Render Offload guide. Desktop integration may also offer a “launch on discrete GPU” option. For applications or launchers that need explicit environment variables, use the ones appropriate to the graphics API: NVIDIA documents __NV_PRIME_RENDER_OFFLOAD=1 for render offload, __GLX_VENDOR_LIBRARY_NAME=nvidia for an OpenGL context, and __VK_LAYER_NV_optimus=NVIDIA_only for Vulkan device selection. They are not one universal recipe for every application.
Check which GPU owns the display
Find the routing for the exact display you use—internal panel or external monitor, and the specific port—by checking the computer manufacturer’s documentation or graphics control panel. NVIDIA’s Optimus documentation notes that display heads may be connected to either the IGP or the discrete GPU.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Clear out junk files and repair common Windows errors3Fix the driver behind crashes, sound loss and screen glitchesRank #3
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Powered by GeForce RTX 5060
- Integrated with 8GB GDDR7 128bit memory interface
- PCIe 5.0
- WINDFORCE cooling system
- Display connected to an iGPU-controlled head: on a supported Optimus system, the dGPU can render and transfer frames to the iGPU pipeline for scanout.
- Display connected to a dGPU-controlled head: selecting the dGPU for AI does not reroute that connector through the iGPU.
- System with a hardware mux: the mux may switch a panel’s connection between GPUs, if the computer implements and supports that feature.
Do not assume that a BIOS option, NVIDIA Control Panel setting, Windows graphics preference, adapter, or dock can change a connector’s wiring. Routing is specific to the machine and port.
Understand Windows Advanced Display Switching
Microsoft documents Advanced Display Switching (ADS) for Windows 11 version 24H2 update 2025.01D (WDDM 3.2) and later. ADS is optional: it requires coordinated support from Windows, the platform’s ACPI and mux firmware, and both GPU drivers. Windows version alone does not establish that a particular computer supports it.
Rank #4
- Powered by the NVIDIA Blackwell architecture and DLSS 4. System Requirements: Minimum 850W PSU with 16-pin 12V-2x6 (12VHPWR) connector required. Verify before purchasing.
- Military-grade components deliver rock-solid power and longer lifespan for ultimate durability. Compatibility: 348mm (13.7") length, 3.6 slots, 4.3 lbs. Confirm case clearance and slot spacing. GPU bracket included.
- Protective PCB coating helps protect against short circuits caused by moisture, dust, or debris
- 3.6-slot design with massive fin array optimized for airflow from three Axial-tech fans
- Phase-change GPU thermal pad helps ensure optimal thermal performance and longevity, outlasting traditional thermal paste for graphics cards under heavy loads
The initial ADS implementation switches the internal laptop panel; it does not switch external connectors. Microsoft’s ADS documentation describes the feature and its requirements. Check the computer manufacturer’s specifications or support information for mux and graphics-mode options on the specific system.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Verify the AI workload, not just the display
A graphics-renderer check can confirm which GPU handles a graphics application, but it does not prove that an AI process selected CUDA. Check the device reported by the AI framework or application while the workload is running. On Windows, Task Manager’s GPU Engine column provides an additional check; on Linux, confirm CUDA device use in the framework itself.
For Linux graphics-rendering applications, NVIDIA’s PRIME guide demonstrates checking the OpenGL renderer with glxinfo -B and checking runtime power state under /sys/bus/pci/devices/.../power/runtime_status. Those checks are useful for graphics offload, but an OpenGL result alone does not establish which device a separate AI process uses.
Troubleshoot in this order
- Confirm both adapters and drivers are present. On Linux, NVIDIA’s guide demonstrates
lspcito identify Intel and NVIDIA adapters. On either operating system, check that both GPUs appear and their drivers are loaded. - Select the dGPU for the AI app. Use the Windows graphics preference or the Linux desktop/PRIME launch method that applies to your system.
- Run the workload and inspect its device report. Use the AI framework’s own device information; on Windows, also inspect Task Manager’s GPU Engine column.
- If the image is on the wrong output, investigate routing separately. Check the wiring of that panel or port and any supported mux controls. Changing the AI process’s GPU does not itself change display routing.
- For an internal laptop panel, confirm mux support. Check the manufacturer’s graphics-mode documentation and whether the platform supports ADS; do not infer support from the Windows version alone.
Check AI software requirements separately
Using an NVIDIA GPU for CUDA work requires that the application support the relevant device and software stack. The cited Optimus documentation establishes that CUDA can use an NVIDIA GPU that is not the display device; it does not establish GPU model, VRAM, or framework requirements for a particular AI application. Check that application’s requirements before choosing hardware or diagnosing compatibility. A CUDA-capable NVIDIA graphics card may be necessary for some workloads, but it will not reroute a display connector physically wired to the dGPU.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
