My understanding is that mainstream gamers using high end GPUs tend to have to limit their resolutions to 1080p for best gameplay refresh rates.
Also deferred rendering and similar display pipelines are inherently bandwidth intensive so a 4x or 8x Jump in resolution cannot be achieved with the 10-20% GPU improvements we tend to get every couple of years.
There are Forward+ rendering pipelines and Ray Tracing* hardware that could allow rendering techniques to cope with the rise in resolution but are they enough?
So let’s discuss the coming UHD Apocalypse and what techniques, technologies and ideas might solve the Display GAP?
Actually when we have powerful enough GPUs to drive these displays will our CPUs become the bottleneck?
*RTX in theory could simplify the workload of your GPU so that it does not have to do more work for shadows, indirect lighting, reflections and ambient occlusion it just has dedicated hardware to ray trace and denoise the results simplifying the process and speeding it up via hardware.
Native is king, I would never run my games at anthing but 1440p which is my monitors native res. To be honest the fillrate on my 1080 TI has no problem outputting 1440p at ultra settings. And the 2080 TI is probably even better at it.
Probably by:
A) more math in the vertex shaders instead of pixel shaders.
B) texture space shading? nvidia added that to turing as well.
C) rendering at lower res and upscaling
D) doing fancy culling in compute shaders to minimize overdraw
Well in theory your 1080 Ti has a fill rate of 139 Giga Pixels a Second*, so should be able to run a 4k display at 16,758 hz, so fill rate is not the limiting factor here…
*Assuming gigapixels are a billion/1,000,000,000 pixels.
Youtuber tests 1080ti at 5k and 8k resolutions and even though the cards fill rate should manage this it massively underperforms < 60 fps at 8k. https://www.youtube.com/watch?v=Vj_2iGNu5JE
Of course there is the flip side of the need for higher resolution displays where we surpass a point where the human eye is able to distinguish between super high resolutions.
In summary this is very dependent on the distance between the eye and the screen and as we use gaming/computer monitors about an arms length away we should be good up to about 8k at 16k we are going to need larger desks or better eyes.
What about layered scene construction, just like in parallax scrolling, skyboxes, imposters* once something is far enough away from the player it’s image does not need to be refreshed until the player moves enough for it’s representation to be rotated enough to change it visually.
Or until it moves/rotates enough relative to the player.
*Imposter - Using pre/rendered sprites to represent far away objects in a 3D scene.
Post process effects are also resolution bound. I dont think 8k or even 4k is needed for desktop sized monitors. I have a 120" projector with a 4k panel. sure, on that size I do notice a difference from 1080p. But there are other factors that are more important like great contrast.
In VR we can benefit from 8k per eye though with foveated rendering. So the center of vision is rendered at 1:1 pixelratio. And then you lower the resolution further out,
It’s worth noting that foveated rendering could also be apply to personal screen, as long as the tracking is faster than the refresh rate and eyes perceptions.
Given the fact that 4K displays have been available for eight billion years in the PC space and have next to no marketshare still, this entire thread relies on pretty pointless speculation.
Optimizing on the CPU (lets assume Unity and DOTS have this aspect covered).
Maximise Batch Size
Reduce The Cost of Vertex Transfer
Optimise Vertex Processing
Speeding Up Fragment Shading
Reducing Texture Bandwidth
Optimising Frame-Buffer Bandwidth
The main take away point here is that GPUs are complex with stages to the rendering process and therefore a simple measure like fill rate is no longer a good metric for performance in modern games at higher resolutions.
It looks like the first 4k monitor was launched around May 2013 (source) for a whopping $5,000 compared to today’s prices of around $400 for the same size 4k monitor. So six years and a 12.5x times drop in price so if you assume a halfving of the price every 2 years then 4k monitors should start hitting the $200 sweet spot very soon (a quick search and you will find 4k displays just over $200).
These sweet spot price points tend to be when very large changes in marketshare happen. So expect 2019-2021 to be transitional years for 4k displays.
The question is are there rendering pipelines that can scale better than current deferred rendering?
I can definitely see the use case for large 4K/8K/16K monitors for 2D workstation tasks. It is basically just a more elegant solution than using multiple monitors in a workstation. However, I don’t see any compelling reason to switch from 1080p or 1440p to 4K/8K/16K for a gaming monitor. For playing modern 3D games on a PC, I personally like 1440p @ 144 Hz the best. I personally like 27" monitors for gaming. Anything physically larger is a bit too wide for my tastes for PC desktop gaming. I see zero need for 4K/8K/16K ultra large monitors for PC gaming.
That’s cool the question is what do your consumers want if another game runs at 4k @ 120hz and yours does not which one will they choose to show off their new 4k display?
Given the lack of practical content for these higher-end displays, I would say that they will have very little influence over the market. 4K isn’t even considered standard yet. The market is still adapting to this change, and struggling, to a fair degree. For the moment, the only real “need” for such high-resolution displays is for very, VERY large displays. If it isn’t 65+ inches, it benefits marginally from 4K, and not at all from 8K. Worrying about running a game at such resolutions is something most of us won’t actually have to concern ourselves with for some time.
it’s not just resolution x framerate, add bit depth to that (hdr) I haven’t check but I think we are going from 8bit to 12bit per color, ie lotta data.
