It’s finally here, GDC17 is underway!
After a minor mishap with some ligaments, my expo-walking, booth-reviewing GDC plans have had to be slightly readjusted, but if nothing else, immobility means the opportunity to sit in on all the awesome talks my heart desires.
If you’ve read any of my other blogs you probably know I’m something of a VR enthusiast. My last blog dedicated to the future of the field considered what needs to happen next to take the industry truly mainstream. It’s probably therefore not surprising that first up this morning I joined Patrick Walker of Eedar for his data insights into the evolution of the VR industry so far, and what he sees for the market in 2017. ‘VR was a victim of its own hype in 2016’. An interesting concept, particularly when followed up with the idea that people have been comparing VR to 3D TV and predicting it will have the same lacklustre fate. What was great (but not surprising) to see, is that this simply isn’t the case. Something as simple as Google search data shows that the public interest in 3D TV never even approached the still-growing interest in VR. Whilst there are still barriers to mainstream adoption, as Patrick described, accessibility and immersion, we’ve already seen great leaps in user uptake, and a 1.2 billion dollar market so far. Mobile VR was the stealth winner in 2016, ‘sneaking under the radar’ and surprising experts with the number of users it drew in.
What’s even better is that most people who tried VR in any form, loved it. They were happy with the experience and they were prepared to pay for it. Whilst there is of course an enormous range of price points, in both headsets and content, just about all of them deliver on user expectations. Whilst people aren’t playing for hours at a time as they would with traditional games, they are coming back again and again and telling their friends. As content and quality continues to improve people are sure to play for longer and this in turn encourages developers to keep pushing the boundaries of the content.
Patrick also talked about what could happen over the next 18 months or so to change the game again. Cost reduction is an obvious factor, only 3% of people are prepared to pay for the top end $700+devices, 34% the $400 range, whereas more than 85% of people would buy something along the lines of a Samsung Gear VR. This is key to the continued impact of the mobile form factor on VR adoption. As the next generation of headsets hits the market with more powerful and performant GPUS (whether in a smartphone-powered system or a standalone) mobile could actually take even more market from console. Growing headset quality also enables growth in content quality and the potential value in AAA games being released simultaneously across platforms instantly ramps up the industry’s potential. This was supported by the next talk I attended, where ARM’s very own Sam Martin joined Alon Orbach and Juan Wee of Samsung Electronics to talk about their view of mobile as the future of VR.
Juan set the scene by discussing the rapid evolution of how we use our phones from calls to texts to pictures to games. Now, with VR, we’re pushing mobile devices to the very limits of what they can do. Perhaps surprisingly, only 44% of VR content consumption is gaming. Video has taken a much bigger role than we might have thought, with more than 10 million hours consumed in 2016. So why might this be? The rapid growth of the technology is one factor, mobile platforms weren’t originally designed for VR but have done an incredible job of adapting. Other user experience features we’d expect from conventional gaming aren’t yet 100% possible in VR, at least not without risking user comfort. Interaction is going to be another important factor to mass adoption, current levels of movement in, and interaction with, a virtual environment are effectively limited to rotation, and simple inputs from buttons within the headset or with basic handsets. To be able to broaden these options and improve immersion we need to be able to add complexity without increasing latency beyond the broadly accepted 20ms where movement is imperceptible. We’ve talked before about developing latency reduction technologies like low persistence displays, front buffer rendering and timewarp; and this ‘just-in-time’ rendering approach to account for last minute view changes is now a VR staple. However, this is a rotation only correction and can’t account for other aspects like hand movement. With this in mind, Oculus have implemented a technology called Spacewarp which can synthesise a third frame from two previous ones with impressive accuracy. This means changes other than rotation can be taken into account for a more immersive, realistic experience.
We’ve also spoken before about resolution, so is this still relevant? Absolutely; but it’s not all about latency. Once you’ve achieved that crucial <20ms, there are diminishing returns. As Sam said, imperceptible is an absolute, you can’t make something more imperceptible. Resolution gains therefore allow flicker reduction, an issue still present in most VR from time to time. This could also mean LCD displays could reach the level of VR applicability that only low persistence OLED displays are currently capable. OLED displays are significantly more expensive than LCD so this seemingly small gain could actually make a huge difference to cost barrier of widespread VR adoption.
Another major improvement is the emergence of the Khronos Group’s Vulkan API. As Alon said, Vulkan is simply much better designed for modern GPU architectures, and ideal for VR. Not only does it give developers explicit control but it’s also much more friendly to multithreading than OpenGl ES. This means bandwidth-heavy CPU drawcalls are no longer a bottleneck and up to 30% of power can be saved. From a developer perspective this means you can choose to use the extra power for more impressive graphics, or pass on 10-12% more gameplay time to players.
So what’s next? As Sam said, ‘a long list of things before we reach VR Nirvana’ but we’re making progress. Performance isn’t limited by GPU size but by thermal limits and we have therefore have to learn to do more with less – doing more with more is easy. There are hardware optimizations to be made. As I’ve mentioned before, mobile VR isn’t necessarily powered by a slotted-in smartphone. Standalone headsets broaden the possibilities and headsets attached to a phone that doubles as a controller still more so. This not only improves the ergonomics of the head mounted display but also allows greater opportunity for heat dissipation. Not only that, but the interaction that could be enabled by a smartphone-as-controller option takes us right back to one of the key improvements needed for mainstream adoption.
Foveated rendering, too, is progressing. The ability to reduce the rendered quality of areas of the environment that our eyes cant detect gives us the chance for the 2x improvement needed to facilitate 4k displays, another thing we talked about earlier. It’s easy to see how these improvements, both incremental and monumental, are adding up to a massive increase in quality, usability, affordability, and therefore value.
Whilst we’re not quite at Sam’s nirVRana, one thing is for sure, it’s going to be an exciting couple of years for virtual reality.