Skip to main content

MXM GPU computing on board a digital train

TypeCase study
Published18 Oct 2026
AuthorAndrew WhitehouseManaging Director
Reading time8 min

A digital train gathers data from cameras and sensors throughout the vehicle and acts on it, whether that is watching the track ahead, monitoring saloons or recording evidence-grade video. Doing that analysis on the train, rather than sending every stream to the ground, needs a rugged AI computer with a discrete GPU that can live in a rolling stock equipment bay.

Here we look at how the Vecow IVX-1000 series is put together for that role. Both builds are fanless railway computers designed for EN 50155 rolling stock duty, with an NVIDIA MXM GPU module fitted by request alongside up to a 24-core Intel Core i9-13900E. As standard, we fit the GPU module, memory and drives, load the operating system and analytics stack, and set the BIOS and ignition behaviour before the system ships.

What the job demands · 6 checksOn-board GPU computing

What a digital train asks of on-board computing

A train is a hard place for a GPU: the power is noisy, the vibration never stops and there is no room for a server rack.

  1. 01

    GPU power for many cameras

    Running detection on several video streams at once needs more than integrated graphics can offer.

  2. 02

    Rolling stock power

    Train supplies vary widely between fleets and carry surges, so the input range and isolation matter.

  3. 03

    Constant vibration and shock

    Every component, including the GPU, has to stay seated and connected for the life of the vehicle.

  4. 04

    Camera power and data on one cable

    PoE+ on locking connectors keeps the wiring down to one cable per camera along the carriage.

  5. 05

    Storage for recorded video

    Footage kept as evidence needs several drives with RAID, ideally removable from the front.

  6. 06

    Rail EMC compliance

    Equipment on rolling stock is expected to meet the EMC requirements of EN 50155 and EN 50121-3-2.

Demands

Why an MXM GPU suits a train

MXM is a standard for compact GPU modules that mount flat on the computer’s board and are screwed in place, rather than standing in a slot like a desktop graphics card. That lets a discrete GPU sit inside a low, sealed chassis built for vehicles, with no tall card to support. For the difference between a module like this and the graphics built into the processor, see our iGPU vs dGPU FAQ.

How it is put together · 5Build

How the system is put together

Five parts of the IVX-1000 build decide how well it handles life on the train.

Compute and GPU

Intel 12th or 13th Gen Core processors on the R680E chipset, up to the 24-core Core i9-13900E at 65 W, handle the system work. An NVIDIA Quadro MXM module is fitted by request for GPU video analytics.

Cameras and network

Eight X-coded M12 2.5GbE ports carry the cameras and on-board network, four of them with PoE+. A further 2.5GbE port supports TSN, and a GbE port with Intel AMT allows out-of-band management from the depot.

Recorded video

The IVX-1000 ICY has six front-access 2.5″ drive trays with software RAID 0, 1, 5 or 10 and an optional UPS. The IVX-1000 has two trays with RAID 0 or 1. Both add an M.2 PCIe x4 boot drive.

Train power

A 16 V to 160 V DC input with 4 kV isolation and 0.5 kV surge protection takes the train supply, and 16-mode software ignition control manages start-up and shutdown.

Rail ratings

The chassis is fanless, with IEC 61373 shock and vibration and EN 50155 and EN 50121-3-2 EMC listings. It wall-mounts, with rack kits as an option. Ratings like these are what separate a rugged industrial PC from an office machine.

Computers

MXM module or PCIe card

Feature IVX-1000 ICY IVX-1000 ECX-3200MX PEG
GPU NVIDIA Quadro MXM module, by request NVIDIA Quadro MXM module, by request PCIe x16 slot for an NVIDIA or AMD card, up to 200 W
Processor Intel 12th/13th Gen Core, up to i9-13900E Intel 12th/13th Gen Core, up to i9-13900E Intel 12th–14th Gen Core and Core 200S, up to 24 cores
Camera LAN 8× 2.5GbE M12 X-coded (4× PoE+) 8× 2.5GbE M12 X-coded (4× PoE+) 4× 2.5GbE PoE+ M12 X-coded
Other LAN 1× 2.5GbE (TSN), 1× GbE 1× 2.5GbE (TSN), 1× GbE 2× GbE
Drive trays 6× 2.5″, RAID 0/1/5/10 2× 2.5″, RAID 0/1 2× 2.5″, RAID 0/1
Power input 16–160 V DC, 4 kV isolation 16–160 V DC, 4 kV isolation 12–50 V DC
Cooling Fanless Fanless System fan
Operating temperature −25 to 75 °C (35 W CPU, no GPU) −25 to 75 °C (35 W CPU, no GPU) −20 to 45 °C
Shock and vibration IEC 61373 IEC 61373 MIL-STD-810G
Rail EMC EN 50155, EN 50121-3-2 EN 50155, EN 50121-3-2 EN 50155

For equipment on the train itself, the IVX-1000 builds are the closer fit: the wide isolated input and IEC 61373 rating are written for rolling stock. The ECX-3000 PEG series takes a larger PCIe graphics card, which suits a cooler, controlled space such as a depot or trackside cabinet.

What it gives · 4 benefitsDigital train

What this setup gives the operator

With the GPU on board, the train analyses its own cameras and sends results rather than raw video.

  1. Outcome 01

    Analytics where the cameras are

    Detection runs beside the cameras, so decisions do not wait on a link to the ground.

  2. Outcome 02

    Fewer boxes in the equipment bay

    GPU, PoE+ camera ports, RAID storage and train power input share one fanless chassis.

  3. Outcome 03

    Evidence kept safe

    Front-access drives with RAID 5 or 10 and an optional UPS on the ICY build protect recorded footage.

  4. Outcome 04

    Managed from the depot

    Intel AMT on the management port lets engineers reach the computer even when the operating system is not responding.

For another view of computing in rail, see how rugged railway computers support track inspection, and our explainer on video analytics.

06FAQs

On-board GPU computing FAQs

  1. What Is EN 50121-3-2?

    EN 50121-3-2 is the European standard for electromagnetic compatibility of electrical and electronic equipment fitted to railway rolling stock. It limits the interference a device may emit and sets how well it must withstand interference from the train, so a computer neither disrupts nor is disrupted by signalling and traction systems.

  2. What Is IEC 61373?

    IEC 61373 is the international standard for shock and vibration testing of equipment fitted to railway rolling stock. It sets test levels by where the equipment is mounted, from inside the car body to the bogie and the axle, so a computer can be qualified for the vibration it will actually see on a train.

  3. What Is an MXM GPU?

    An MXM GPU is a graphics processor on a Mobile PCI Express Module, a compact standard board that mounts flat on a carrier instead of standing in a PCIe slot. It lets laptops, embedded systems and rugged computers add discrete NVIDIA, AMD or Intel graphics in a small space, often cooled through the chassis rather than by its own fan.

  4. iGPU vs dGPU: What Is the Difference?

    An iGPU (integrated GPU) is built into the processor and shares the system memory, so it adds little power, heat or cost. A dGPU (discrete GPU) is a separate graphics processor with its own memory, offering far more performance but needing more power, cooling and space. For edge AI, iGPUs suit lighter inference and dGPUs suit heavy workloads.

  5. What Is Video Analytics?

    Video analytics is software that analyses video automatically to detect, classify and track people, vehicles and events, so operators do not have to watch every screen. Modern systems use AI models and often run at the edge, on a computer beside the cameras, to react in real time and avoid sending every stream to the cloud.

  6. What Is an NVR?

    An NVR (network video recorder) is a recorder that receives video streams from IP cameras over a network, stores them, and lets users watch live and recorded footage locally or remotely. PoE NVRs also power the cameras. An AI NVR adds a GPU or AI accelerator so it can analyse the video, detecting people, objects and events as it records.

  7. What Is Condition Monitoring?

    Condition monitoring is measuring a machine’s health continuously or at regular intervals, using parameters such as vibration, temperature, oil quality, sound and electrical current, to detect changes that point to a developing fault. It turns breakdowns into planned repairs and is the data foundation that predictive maintenance is built on.

// Next Step

Adding GPU processing to rolling stock?

Tell us the cameras, storage and analytics each carriage needs, and we'll specify the IVX-1000 build, GPU module and image to match.

Phone · UK +44(0)1785 558 300