Raster graphics future

jusomdude

At what point do you think will it be pointless to increase the polygon/texture resolution? Honestly I think we're already there. Only when zoomed in to extremes is texture filtering and polygons even noticeable. I think it's about time developers start focusing more on raytracing or something. Pretty much the only area that makes a huge difference in realistic rendering right now is lighting quality which raytracing has the leg up over raster graphics.

Deivos

'Ish

While you can say we've been on the curve where the return in detail and visual quality compared to the amount of polygons and resolution scaling is on the decline, there is still quite a lot of benefit you generally see to the increases of such still in environmental detail, and back-end aspects that use polygons but not necessarily for visual rendering such as collision meshes and destruction models still have quote a lot that they can benefit from a continued progression of rendering capacity on at least the polygon side of things.

As for textures, that actually is highly dependent on the content as if you're talking about a small object with a 4k texture then yeah it's gonna look good even when it's close to the monitor since most monitors at their largest standard resolution are 4k. Lower resolutions that are still common such as 1024 and 2048 are good at a distance, but inevitably blurry up close. And then again when you look to large objects and environmental rendering you are butting heads with limited texture scales stretched across large areas that make even high resolution textures have some issues.

While it may not be quite the dramatic change that it used to be, there is still a decent amount to be gained from improving raster technology. Both in the sense of simply increasing the visual fidelity, and also for back-end use of things for improving on the game mechanics themselves.

That said, there's already quite the shift done towards improving other visual elements of games such as post-processing shaders and volumetric fog/light. Things like raytracing is something that's done in tandem with raster polygon rendering, as raytracing is using projection to map the shapes of objects in order to create dynamic light/shadow effects. An increase in detail on the poly models also causes an increase in the visual detail of the raytracing (depending on how detailed the projection/stencling method is).

Where all of this stuff could benefit most greatly is finding more effective/efficient ways of data storage and retrieval and the ability to do more calls at once to a large memory space in order to handle a great deal of content scaling up and down quickly.

Just throwing more memory and more processing power at an engine helps, but the engine needs to be capable of efficiently managing the hardware's usage so that it's not throttling it's own performance by under-utilizing some assets and/or overloading the wrong parts (such as can be the case with rigs utilizing a lot of multi-core processors still).

In short, everything still needs lots of improvement, but I feel like there's more that has to be done on the baseline performance of engines that is too often neglected, which would see some of the highest benefit to both developers and players in the long run for not just visual quality and performance, but also gameplay.

Cleffy

Not anytime soon. The amount of calculations needed are far higher in Ray tracing than Rasterizing. Right now we can achieve fairly realistic effects in games using PBR techniques and more detailed textures. This can be seen in UE4 titles. If we took a game from 2004 and used a ray-tracing lighting approach, it would consume a similar amount of resources. The effects would come no where close to what we can do today with raster. Sure the shadows may be more accurate, but no one is going to notice that slight shade difference on a couple pixels. More advanced ray-tracing techniques like Mental Ray or Cycles Rendering still take minutes to calculate a comparable image instead of the 1/60th a second it would need to.