Good virtual reality, the type that is at least somewhat convincing, is a resource intensive thing to create. Microsoft Research has created what they’re calling ‘Flashback’ which seeks to make decent virtual reality a possibility on lower-end phones and PC’s, arguably the majority of the consumer market.

Microsoft Research Flashback

Flashback is method of rendering only what we see, similar to tile-based deferred rendering

The problem is generally that the CPU and GPU horsepower just isn’t quite enough to provide an actual visceral and engaging enough experience. But Flashback is segmenting the entire scene into mega frames. These mega-frames are pre-rendered parts of what you see that are stored in a cache, in VRAM essentially. ThoseĀ assets, or mega-frames, are compressed in memory so that you can store a good portion of whatever the VR experience is. The technology can use any number of storage devices, putting things on the RAM, the GPU CRAM or even on the Flash storage of the device. Which medium it’s on does have an effect on performance, of course, as those mega-frames get read back while you interact with the environment.

Microsoft and the researchers behind this tech say that there is an 8-times better framerate and that it’s 97-times more energy efficient with a whopping 15-times reduction in latency. That’s a massive increase in performance just by caching the various elements. Furthermore, it uses something akin to a tile-based deferred rendering technique, like that used by Imagination Technology, to only actually bring out the parts of the object that you can see. If you can’t see it, it doesn’t get brought out from the cache. This saves plenty in resources and clearly has quite the influence on performance.

In the research paper, they took a rather low-end HP Pavilion Mini and paired it with an Oculus Rifr DK2 to great effect. Check out the demo below. It’s actually rather impressive, considering. The implications for the technology are very wide, and deep. This could make VR a mainstream reality as opposed to being limited to only those with higher-end specs. I could also conceivably increase the quality of mobile VR by a few orders of magnitude.