Jump to content

PC-Powered AR

From VR & AR Wiki

PC-Powered AR is augmented reality in which the headset does little of the heavy computing itself and instead relies on a separate, more powerful computer to do the work. Sensor data from the headset is sent to the PC, the PC's processor and graphics card render the scene, and the finished frames are sent back to the headset to display. The connection can be a cable or, in some designs, a short-range wireless link. On this wiki it is treated as one of the main device subtypes of AR, alongside Standalone AR and Phone-Powered AR. It is also called PC-tethered AR or tethered mixed reality.

The reason to do this is graphics power. A desktop or workstation GPU can render far more detail than the mobile-class chip that fits inside a headset, which matters most for photorealistic content, high resolutions, and demanding professional applications. The cost is mobility: you are tied to the host machine and, usually, to a cable. This is the same fidelity-versus-freedom trade-off that separates PC-Powered VR from Standalone VR, applied to see-through and pass-through AR hardware.

What counts as PC-powered

The dividing line, as with the other AR subtypes, is where the compute lives. A PC-powered device pushes the rendering (and often the tracking) onto an external machine and treats the headset largely as a high-quality display with sensors. The two things it is defined against keep the compute closer to the user.

Standalone AR does everything on hardware it carries, with an onboard processor and battery and no external machine in the loop.

Phone-Powered AR leans on a smartphone for both processing and, often, power; in its most common form the augmented scene appears on the phone's own screen rather than through worn optics.

PC-powered AR moves the work to a desktop, laptop, or workstation. That buys desktop-class graphics at the price of a tether.

Subtype Where the compute lives Power source Tether
PC-Powered AR An external computer (desktop, laptop, or workstation) Wall power, through the PC Cable, or short-range wireless, to the PC
Standalone AR Onboard, inside the headset or glasses Onboard battery None
Phone-Powered AR A connected smartphone The phone (and sometimes a cable to it) The phone, by cable or wireless

How it works

There are two ways the work gets split, and they sit at different points on a spectrum from "the headset is just a screen" to "the headset does the sensing but the PC does the rendering."

In the first arrangement the headset is close to a pure display with cameras and trackers attached. It sends its camera and sensor streams to the PC, the PC runs the application and renders the left and right eye images, and the rendered video comes back over the cable to drive the headset's panels. This is how dedicated PC-tethered mixed reality headsets such as the Varjo XR-3 and Varjo XR-4 are built: they connect to a workstation over DisplayPort plus USB and depend on a high-end NVIDIA graphics card to produce their image.[1][2]

In the second arrangement the headset is itself a capable Standalone AR computer, but it can switch into a mode where a PC takes over the heavy rendering. Microsoft's Holographic Remoting for the Microsoft HoloLens is the clearest example. Microsoft's documentation describes the case where you "want the resources of a PC to power your app instead of relying on the HoloLens on-board resources": the app actually runs on the PC so it can use the PC's more powerful hardware, which Microsoft says "can be especially helpful if your app has high-resolution assets or models and you don't want the frame rate to suffer."[3] In that mode, inputs from the headset (gaze, gesture, voice, and spatial mapping) are sent to the PC, and "the PC renders content in a virtual immersive view and sends the rendered frames to the HoloLens" over standard Wi-Fi.[3] The same headset is standalone on its own and PC-powered when remoting; the category describes how it is being used at a given moment, not a fixed property of the hardware.

Either way the round trip has to be fast. The headset's pose can change between the moment its sensor data leaves and the moment a frame returns, so latency is the central engineering problem. Microsoft's Holographic Remoting Player reports both frame rate and the average time for a frame to travel from the PC to the headset, because that latency is what determines whether the image feels locked to the world.[4] Wired connections such as DisplayPort avoid the variable delay of wireless and can carry far higher resolutions, which is why the highest-fidelity AR and mixed reality headsets stay tethered by cable.

Representative devices

The most fully PC-powered AR hardware is the high-end professional mixed reality headset, where image quality is the entire point and the workstation is part of the product.

The Varjo XR-3, which Varjo called "the first true mixed reality headset," is a video pass-through headset with a 115-degree horizontal field of view, over 70 pixels per degree at the center of the image, dual 12-megapixel pass-through cameras running at 90 Hz, and 200 Hz eye tracking. It connects to the host with two DisplayPort and two USB-A connections and renders on a high-end PC.[2] Its successor, the Varjo XR-4, shipped from December 2023 with 3840 x 3744 pixels per eye, a 120-degree by 105-degree field of view, dual 20-megapixel pass-through cameras, and 200 Hz eye tracking, and is aimed at training and simulation for enterprise and government users.[5] Both depend on a tethered Windows PC with a high-end NVIDIA GPU; UploadVR notes the XR-4 "requires a powerful PC with a high-end graphics card which it must be tethered to."[5] Varjo's own system requirements for the XR-4 list NVIDIA desktop cards such as the GeForce RTX 5090 or RTX Pro 6000 Blackwell, a DisplayPort 1.4 connection, and a USB-C link, and state plainly that AMD graphics cards are not supported.[1] Varjo discontinued the XR-3 (along with the VR-3 and the Varjo Aero) and ended software support for it on 1 January 2026, concentrating on the XR-4 line.[6]

Earlier consumer-facing AR hardware was PC-powered out of necessity, because mobile chips of the time could not drive a wide field of view. The Meta 2 development kit, announced in 2016 and shipped to developers from late 2016, was an optical see-through AR headset with a 90-degree diagonal field of view and a 2560 x 1440 display that connected to a Windows PC by a single nine-foot cable carrying video, data, and power, using an HDMI 1.4b or DisplayPort video link.[7]

The Microsoft HoloLens and HoloLens 2 sit in the boundary case described above. Both are self-contained Standalone AR computers, but Holographic Remoting lets a developer run an application on a PC and stream the rendered result to the headset, which makes the headset PC-powered for that session. Microsoft positions this for content too heavy for the device's on-board chip to render at full frame rate.[3]

Device Year See-through type Compute / rendering Connection Notes
Varjo XR-3 2020 Video pass-through Tethered Windows PC, high-end NVIDIA GPU[2] 2x DisplayPort + 2x USB-A[2] 115-degree FOV; 70+ ppd; dual 12 MP pass-through at 90 Hz; discontinued, support ended 2026[2][6]
Varjo XR-4 2023 Video pass-through Tethered Windows PC (e.g. RTX 5090 / RTX Pro 6000)[1] DisplayPort 1.4 + USB-C[1] 3840 x 3744 per eye; 120 x 105-degree FOV; dual 20 MP pass-through; for training and simulation[5]
Meta 2 (dev kit) 2016 Optical see-through Tethered Windows PC[7] Single nine-foot cable (video, data, power); HDMI 1.4b / DisplayPort[7] 90-degree diagonal FOV; 2560 x 1440 display[7]
Microsoft HoloLens / HoloLens 2 (remoting) 2016 / 2019 Optical see-through Standalone, or PC-rendered via Holographic Remoting[3] Wi-Fi (or USB)[3] PC-powered only when remoting; otherwise standalone[3]

Where it is used

PC-powered AR is mostly a professional and enterprise tool, for the same reason as PC-Powered VR: it exists where image quality and simulation accuracy outweigh the inconvenience of a cable. The video pass-through Varjo headsets are sold for training and simulation across air, land, and sea, and for design and engineering review, with named customers in the automotive and defense industries.[5] These are seated or stationary uses at a workstation, where being tethered costs little. The same logic supports remoting on the HoloLens: an enterprise can author a complex 3D model that the headset cannot render on its own and stream it from a PC or server.[3]

The category overlaps with Standalone AR at its edges, because the most capable AR headsets can run either way. As mobile silicon improves, more rendering moves onto the device, and remoting tends to be reserved for the heaviest content rather than used by default. Cloud and edge rendering extend the same idea further, replacing the local PC with a remote server, though that introduces network latency over and above the round-trip delay already inherent in offloaded rendering.

See also

References