Adaptive Refresh
Informative interview on www.sweclockers.com. Seems that soon variable frame rate rendering will be common place in the industry for PCs (both AMD and NVIDIA). Good reason for an update of a GPU and monitor. For the majority of engines which are not built around consistent frame time, this is going to be a serious quality of service improvement.
Will be interesting to see if variable frame rate can ever find some application in VR. After the Oculus presentation at UNC I had a chance to chat for a minute with Michael Abrash. Michael's Power Graphics Programming book was my goto guide for VGA docs when I was a kid. Great to meet someone who was instrumental in the process I took to learn graphics programming. After that talk, sounds like there are possibly some currently unsolved challenges getting VR and low persistent displays to mix well with adaptive refresh rates.
Low Persistence
Low persistence displays are here now, and showing up more in the future. For more information, Blur Busters is a great resource, and this PCMonitors.info article is also a great place to start. The TFT Central tech article on the BenQ XL2720Z display is an interesting read. One advantage of this display is that turning on the low persistence mode does not have that bad of an effect on tonal curves on output. Also note that low persistence on 60Hz is produced by double strobing a frame which will produce movie theater style judder (actually 2 frames instead of 3 frames). Using 120Hz or 144Hz is effectively required.
How These Technologies Will Effect Gaming
There is an opportunity to specialize and leverage each technology in different ways. Gamers who like high resolution (2560x1600 and beyond) and devs who like more general engine technology are likely to favor adaptive refresh. Seems like adaptive refresh is more friendly to the traditional AAA industry. I have a feeling that the way I tuned the temporal AA in UE4 for instance actually leverages the typical LCD scan-and-hold induced blur to mask the blur introduced by delaying the contrast change reduces the temporal aliasing. So much easier to push fast graphics R&D when consistent frame time is not a design requirement.
Gamers/devs who like high frame rates, ultra-low latency, and not motion blur are going to need to trade resolution (or visual quality, or adapt content) in order to get the required high frame rate in comparison with the adaptive refresh case. Ideally, engine tech will be radically different and built around as fixed cost as possible for rendering. Technologies involving draws which adapt to current time before v-sync are going to be very important in insuring consistent frame rates. Post processing will be mostly removed. Antialiasing becomes more important as sub-pixel edge perception is more important due to lower resolution. Temporal noise becomes an even more important tool, as high refresh rates can make very high amounts of "film grain" acceptable, and this visually pleasing noise can mask a large number of rendering problems. This allows the mind to fill in the gaps producing a more "real" image.
Oculus and low persistence enables a revival of games which return to first principles of graphics: the route Pixar followed. Get framerate solved, then aliasing, then animation, only then use what is left for shading features. Strip out anything which would distract from the perception of ultimate quality in a given visual style. For example if specular aliasing and reflections can not be solved without artifacts, toss them, and focus on something else. An attempted carbon copy of physical shading and materials from the real world are not likely to pass this test. This is the realm for stylized art. This is an opportunity for an explosion of radically different independent works which pursue the purity of expression.
Informative interview on www.sweclockers.com. Seems that soon variable frame rate rendering will be common place in the industry for PCs (both AMD and NVIDIA). Good reason for an update of a GPU and monitor. For the majority of engines which are not built around consistent frame time, this is going to be a serious quality of service improvement.
Will be interesting to see if variable frame rate can ever find some application in VR. After the Oculus presentation at UNC I had a chance to chat for a minute with Michael Abrash. Michael's Power Graphics Programming book was my goto guide for VGA docs when I was a kid. Great to meet someone who was instrumental in the process I took to learn graphics programming. After that talk, sounds like there are possibly some currently unsolved challenges getting VR and low persistent displays to mix well with adaptive refresh rates.
Low Persistence
Low persistence displays are here now, and showing up more in the future. For more information, Blur Busters is a great resource, and this PCMonitors.info article is also a great place to start. The TFT Central tech article on the BenQ XL2720Z display is an interesting read. One advantage of this display is that turning on the low persistence mode does not have that bad of an effect on tonal curves on output. Also note that low persistence on 60Hz is produced by double strobing a frame which will produce movie theater style judder (actually 2 frames instead of 3 frames). Using 120Hz or 144Hz is effectively required.
How These Technologies Will Effect Gaming
There is an opportunity to specialize and leverage each technology in different ways. Gamers who like high resolution (2560x1600 and beyond) and devs who like more general engine technology are likely to favor adaptive refresh. Seems like adaptive refresh is more friendly to the traditional AAA industry. I have a feeling that the way I tuned the temporal AA in UE4 for instance actually leverages the typical LCD scan-and-hold induced blur to mask the blur introduced by delaying the contrast change reduces the temporal aliasing. So much easier to push fast graphics R&D when consistent frame time is not a design requirement.
Gamers/devs who like high frame rates, ultra-low latency, and not motion blur are going to need to trade resolution (or visual quality, or adapt content) in order to get the required high frame rate in comparison with the adaptive refresh case. Ideally, engine tech will be radically different and built around as fixed cost as possible for rendering. Technologies involving draws which adapt to current time before v-sync are going to be very important in insuring consistent frame rates. Post processing will be mostly removed. Antialiasing becomes more important as sub-pixel edge perception is more important due to lower resolution. Temporal noise becomes an even more important tool, as high refresh rates can make very high amounts of "film grain" acceptable, and this visually pleasing noise can mask a large number of rendering problems. This allows the mind to fill in the gaps producing a more "real" image.
Oculus and low persistence enables a revival of games which return to first principles of graphics: the route Pixar followed. Get framerate solved, then aliasing, then animation, only then use what is left for shading features. Strip out anything which would distract from the perception of ultimate quality in a given visual style. For example if specular aliasing and reflections can not be solved without artifacts, toss them, and focus on something else. An attempted carbon copy of physical shading and materials from the real world are not likely to pass this test. This is the realm for stylized art. This is an opportunity for an explosion of radically different independent works which pursue the purity of expression.