US The short answer
An MXM GPU can run in a desktop, but not on its own. A desktop has PCIe slots, not an MXM connector, so the module needs an MXM to PCIe adapter: a carrier board with an MXM 3.x connector on one side and a PCIe x16 edge on the other. With the module screwed down and a cooler fitted, the assembly behaves much like a graphics card.
Embedded GPU makers list these carrier boards for their own modules, and third-party adapters exist for laptop modules. Whether the result works well depends less on the adapter than on the module, its firmware and how it is cooled. If the format is new to you, start with what an MXM GPU is and how an MXM slot works.
What an MXM to PCIe adapter does
The adapter stands in for the carrier board the module would normally sit on. It has four jobs:
-
//
Routes the PCIe lanes. It connects the slot’s lanes to the MXM connector. The link runs at the lower of the slot’s and the module’s PCIe generation, over the lanes the module uses, which is often eight.
-
//
Supplies power. A PCIe x16 slot provides up to 75 W. The adapter converts that to the module’s rails, and higher-power modules need an auxiliary PCIe power connector from the power supply.
-
//
Carries the display outputs. DisplayPort or HDMI connectors on the bracket take the module’s display signals. A headless module has no outputs, so the screens stay on the desktop’s own graphics.
-
//
Holds the cooler. The module has no fan of its own, so the adapter or the module maker provides a heat sink and fan sized for the module’s power.
Embedded modules vs laptop modules
This is where most desktop MXM projects succeed or fail. Embedded MXM modules, such as NVIDIA’s RTX Embedded range, are sold as standalone products with a standard video BIOS and published drivers. The MXM GPU modules we supply, for example, run on 64-bit Windows 10 and 11 and Linux, with CUDA, OpenCL, DirectX and Vulkan support, so they behave the same on a bench as in the finished system.
Modules taken from laptops are different. Many were built for one laptop model, with a video BIOS, power settings and display wiring that assume that laptop’s motherboard. On an adapter they may not start, may need modified firmware or may run only with particular drivers. Some later laptop GPU boards were never MXM at all, but proprietary designs of a similar shape. Treat laptop modules as an enthusiast project rather than a supported build.
Power and cooling on the bench
Size the adapter and cooler for the module’s full rating. The modules in our range run from 35 W for the headless RTX A500 Embedded to 115 W for the discrete builds of the RTX A4500, 3500 Ada and 5000 Ada Embedded. A 60 W Type A module may run from slot power alone, depending on the adapter; a 115 W Type B+ module needs auxiliary power and a cooler designed for its layout.
Match the cooler to the setting too. The fan coolers offered for embedded modules are rated for ambient temperatures such as 0 °C to 40 °C or 0 °C to 55 °C, and a desktop case is usually well inside that. A heat spreader version, meant to pass its heat into a rugged chassis, will overheat on an open bench unless it is bolted to a heat sink. Our guide to conduction cooling explains that heat path.
When a desktop MXM setup makes sense
The common reason is development. A team building an embedded or rugged system can put the same MXM module in a desktop or workstation, write and tune the software, then move it to the target hardware with the GPU, drivers and model behaving the same way. It also suits test benches that check modules before they are fitted, and labs comparing several modules side by side.
It rarely makes sense as a cheaper graphics upgrade. A PCIe graphics card with the same class of GPU is simpler to buy, cool and support, and usually runs faster because it is allowed more power. For a factory or cabinet computer with room for a card, our industrial GPU computers take PCIe graphics cards directly, and our workstation GPUs cover the cards themselves.
Going the other way
A desktop graphics card cannot be turned into an MXM module, so if the target system has an MXM slot, the GPU has to be an MXM module from the start. When planning a rugged build, choose the module that will ship in it, develop on that same module, in an adapter if needed, then fit it in one of our MXM GPU computers.
Quick answers on MXM to PCIe
MXM in a desktop: common questions
Is an MXM to PCIe adapter the same as an eGPU?
No. An eGPU is a graphics card in an external box connected over Thunderbolt or OCuLink. An MXM to PCIe adapter sits inside the desktop in a normal slot.
Will an MXM GPU perform the same in a desktop?
Close to it, if the adapter gives it the same PCIe link, power and cooling it is rated for. Heat is the usual reason for lower speed.
Can I use a laptop MXM GPU in a desktop?
Sometimes. It often needs a matching video BIOS and particular drivers, and is not a supported setup.
Do I need an x16 slot?
Usually. Many modules use eight lanes or fewer, but most adapters have a full-length x16 edge.
We fit MXM modules in our systems with the cooling, drivers and CUDA software set up as standard, and supply the same modules for development desktops.