A few weeks ago, at Qualcomm’s Tech Summit in Hawaii, the company announced its first dedicated platform for augmented reality glasses. While Qualcomm has long been an enabler of chip consolidation in VR and AR headsets with its XR1 and XR2 platforms, Qualcomm designed those platforms for standalone use cases. The XR1 then slowly shifted to fill a gap in the market for lightweight AR headsets to help OEMs consolidate their bill of materials with fewer chips.
We also saw Qualcomm do this with Lenovo in the ThinkReality A3 smart glasses, which take advantage of the Snapdragon XR1 to simplify the design and optimize performance while connecting to a Snapdragon-based smartphone or a Windows PC for compute. This is the beginning of Qualcomm’s journey of delivering distributed processing for AR headsets, starting with cabled headsets like the ThinkReality A3 and moving onward to wireless.
The Snapdragon AR Gen 1
The new Snapdragon AR2 Gen 1 platform leans further into Qualcomm’s distributed processing strategy of having an AR viewer attached wirelessly to a smartphone and supercharging it. The XR1 platform consolidated components and optimized simpler AR headset designs to minimize the need for more custom silicon and microcontrollers and drivers. By contrast, Qualcomm designed the new Snapdragon AR2 platform to localize compute and reduce power consumption while still boosting performance and doing it all wirelessly. Qualcomm accomplishes this by splitting the AR2 architecture into three chips: display, perception and connectivity.
The AR processor
At the heart of the AR2 Gen 1 platform is the AR processor, which enables much of the headset’s onboard processing and computing capabilities. It also takes advantage of the latest in chip foundry technologies using a 4nm process node, ensuring the best possible power and performance while taking advantage of Qualcomm’s “one technology” roadmap, which brings together the latest CPU, GPU, ISP, AI, security, sensing hub, and fixed function visual processing with the Engine for Visual Analytics (EVA). Many of these different core technologies can be found in other Qualcomm XR products, but not with a specific focus on pushing the limit on performance and power for AR.
The AR co-processor
While the world continues to move towards more-consolidated chipsets, this approach does not work so well for a wearable, especially one that has many displays and sensors, plus tight requirements for weight, power, size, and thermal qualities. Qualcomm has addressed these concerns by creating a co-processor whose primary goal is to reduce the routing and complexity of AR headset designs and simplify how the different parts of the headset work together.
This AR co-processor has its own dedicated CPU, memory, AI accelerator, computer vision engine, sensor, camera aggregation, and security. This means that it helps with eye-tracking, display, and perception performance while also helping to distribute heat away from the headset’s sides, where one usually finds most of the processing.
The FastConnect 7800
The third chip in the AR2 Gen 1 platform is Qualcomm’s FastConnect 7800, Qualcomm’s first Wi-Fi 7 chip designed with XR in mind. While the first device with the FastConnect 7800 is Xiaomi’s 13 Pro smartphone, Qualcomm designed the FastConnect for XR applications with some impressive optimizations. The FastConnect 7800 not only handles the Wi-Fi connectivity, but also implements world-class Bluetooth connectivity for other wearable devices like smartwatches or earbuds for audio.
Because power is so crucial for the AR2 Gen 1 platform, it should be no surprise that the FastConnect 7800 improves power consumption by 40% compared to the previous generation, the FastConnect 6900. The FastConnect 7800 reduces power and latency to as low as 2ms of sustained latency, which is critical for connecting to a smartphone or a PC for processing and 5G connectivity. In addition, the FastConnect 7800 has a peak throughput of 5.8 Gbps, which means that there are no limits to what kind of quality images can be displayed on the headset, even with help from a smartphone or PC. Beyond that, the FastConnect 7800 comes with the second version of the FastConnect XR Software Suite, further optimizing for better XR experiences with prioritization of latency-critical traffic, interference mitigation with preferential channel access, and XR purpose-built power modes.
Benefits of the complete platform
Qualcomm claims that, when combined, the whole Snapdragon AR2 platform consumes less than 1 watt of power, which represents a total reduction of 50% compared to the XR2, along with 2.5X better AI performance and a reduction of 40% in terms of PCB size. The 45% reduction in wires for the headset design also helps reduce product complexity and costs as well as size. The AR2 processor is a tiny 10mm x 10mm package, making it extremely easy to fit anywhere in the headset. Meanwhile, the co-processor is only 6mm x 4mm, enabling it to fit comfortably within the nose bridge. Qualcomm also claims to have achieved less than 9ms of motion-to-photon latency on the AR2 Gen 1, which should deliver a seamless wireless AR experience without any judder.
Partnerships and Spaces-ready phones
The Snapdragon AR2 Gen 1 does not exist in a vacuum, and when you consider that it needs to connect to a phone or a PC for high-quality 3D rendering and 5G connectivity, it makes sense that Qualcomm needs to have a robust ecosystem around it. To that end, Qualcomm has already partnered with headset makers such as Lenovo, LG, Niantic, Nreal, Oppo, Pico, Qonoq, Rokid, Sharp, TCL, Vuzix, and Xiaomi for the AR2 Gen 1 platform.
In addition to partnering with the leading headset manufacturers of the world, Qualcomm has also deployed a Snapdragon Spaces-ready program, enabling any AR2 Gen 1 headset to pair seamlessly with any Snapdragon 8 Gen smartphone from Motorola, Nubia, OnePlus, Oppo, Redmagic, Sharp, Xiaomi, or ZTE. This ecosystem approach makes Qualcomm’s distributed processing vision for AR headsets work with a broader base of devices. While it won’t work with competitors’ chipsets, I do hope that one day there will be standards like OpenXR (which Qualcomm uses in Snapdragon Spaces) that enable such functionality.
At the Tech Summit, Qualcomm also touted a partnership with Microsoft in helping to define the AR2 Gen 1 platform; however, Microsoft wasn’t mentioned in any of the conference materials as a headset or smartphone partner, given that it’s Microsoft’s Surface Duo mobile devices that run on Snapdragon chipsets. But Microsoft did announce at CES 2022 that it would partner with Qualcomm to use Qualcomm’s chipsets in its next-generation lightweight AR headsets.
In addition to Microsoft, Qualcomm also brought Adobe on stage to talk about how Adobe is not only bringing its 3-D and immersive experiences to Snapdragon Spaces, including support for the USD format, which is becoming an industry standard for collaboration in spatial computing. However, I do not believe that it will take the place of formats like glTF, which are inherently more performant and designed for the Web.
Qualcomm has given us a vision for the future of AR with the Snapdragon AR2 platform, including its work with Niantic in building its next-generation reference device for Lightship apps that leverage Snapdragon Spaces. Qualcomm also teased that the AR2 Gen 1 sits alongside the XR2 in its product stack in terms of features and performance, but above the XR1 and whatever AR1 chipset Qualcomm might introduce in the future for extremely lightweight low-cost devices.
Overall, I think Qualcomm has done an excellent job of clearly showing its vision for augmented reality and positioning itself and its partners well for the future of XR. I believe that the ball is now in the OEMs’ court to deliver designs that take advantage of this architecture and for Qualcomm to continue to ramp up its Snapdragon Spaces program to get developers interested in developing for upcoming headsets like the ones Niantic has shown. The AR journey will not be easy, nor will it come quickly, but there are definitely some very important milestones being reached with the AR2 Gen 1, which is very promising for the future.