The next Game Developers Conference is going to be the most interesting in terms of virtual reality. AMD has already confirmed that it will give a talk on how to minimize latency to get the most out of VR viewers.
As we read on the GDC website, this talk will provide a detailed explanation about various mechanisms that graphics engine developers can use to reduce both real and perceived latency, as well as micro-stops in the graphics of virtual reality applications on Moderna GPUs such as those based on AMD GCN architecture. Examples of real engines of AAA titles will be used and concrete figures will be given before and after applying the optimizations.
The talk will give developers the tools and the level of understanding necessary to fully exploit the Moderna GPUs architectures to take their engines to a new level: extremely low latency, absence of micro-stops and smooth graphics at high refresh rates (90-120Hz and beyond).
A few months ago it was nVidia who made its intentions regarding VR clear with the presentation of Maxwell, although many of the announced features (such as the special SLI mode for virtual reality in which each GPU renders an eye) still show no signs of life. It is clear that AMD does not want to lag behind in what will be the great workhorse of the coming years. The next AMD GPU, which we all assume will be called the R9-390X, will be the first to introduce the stacked memory, which could offer some numbers that pull backwards, since there is talk of a bandwidth around 640 Gb / s (triple that of the GTX 980 for example). There are many who think that another leap similar to the one that occurred with the first 256-bit GPU, the Radeon 9700 Pro, could be experienced, one of the biggest jumps so far, and this is something that virtual reality is clamoring for.