4-to-1 Scaling: the Key to Smooth FPS on Ballooning Resolutions

[Ferzerp]

I trust a monitor to that less than I trust any other computer component to do what I want. Outside of business notebooks, it has been 5+ years since I have seen a display not do blurry scaling. I'm sure some are out there, but if it were done by the GPU, to a native res framebuffer, then the monitor could be entirely removed as a variable, which would be a good thing.

Not merely that, but since the whole point would be lower res videos and video games, even a substantial GPU performance hit would still result in both better FPS/quality, compared to full rendering of 4x or higher the number of pixels. As I'm sure you can guess, I'm more interested in the quality part.

I don't have a camera of high enough quality to capture it, but it appears that my U3011 handles it fine.

 

[Cerb]

I don't have a camera of high enough quality to capture it, but it appears that my U3011 handles it fine.

I'll trust you on that...and engineers thinking about that sort thing might happen more often with $1200 monitors than most of our $200-400 ones. In NVidia or AMD handled it, though, it would be pretty much a non-issue.

 

[know of fence]

If you want to know how well your GPU would do, enable 2xSSAA at 1080p, Super Sampling Anti-Alasing effectively renders the scene at 4k then downsamples that to achieve the AA effect.

Great point to mention. FSAA, SSAA, MSAA all seem to be pretty much the same thing, with MSAA doing partial/local super sampling. Clearly these methods come from the exactly opposite approach, which basically says that AA becomes obsolete as soon as resolutions get high enough. Those high resolution work because MSAA is applied rather early to just the geometry it seems.
http://www.youtube.com/watch?v=3EKECcdKXbs&t=3m31s

It will be a reasonable way for a low end graphics card to achieve acceptable quality if 4k means we will get pixel densities 4x todays screens. But if the pixel density drops it wont look good.

It should be noted that Pixel Per Inch is linear, and only has to double. going from around ~90 to ~180 PPI. But what you are saying is my intention exactly, we shouldn't be afraid to upgrade monitors for better = denser screens, fearing that higher resolution may decrease the experience, due to having to run native. To have iphone like density a 4K screen, would have to be just 13.5" though .

I'd really like to know how 720p looks on a 120 PPI, 27", 2K screen.
Calc. -> http://members.ping.de/~sven/dpi.html

 

[PrincessFrosty]

Great point to mention. FSAA, SSAA, MSAA all seem to be pretty much the same thing, with MSAA doing partial/local super sampling. Clearly these methods come from the exactly opposite approach, which basically says that AA becomes obsolete as soon as resolutions get high enough. Those high resolution work because MSAA is applied rather early to just the geometry it seems.

This depends, pixel density is a function of both physical area and number of pixels, increasing the resolution is only half the story, what happens to the physical size/area will decide the overall PPI and that will decide the need for AA or not.

 

[know of fence]

Edit to add: I would also add that framerate doesn't drop linearly with resolution, believe it or not. Just because you're on triplescreen does not mean you get 33% of the fps, more like 40-60% depending on the game.

You are not wrong about non linear, also I didn't say resolution and framerate scale that way. But I see how my words could be misconstrued. The problem really isn't you or I, the problem is that review sites don't actually inform their readers. Is this a sinister corporate practice or is this because people simply want quick and dirty purchasing advice? - I couldn't say. What is certain, that we get pseudo meaningful bar charts, the likes of this:

Actually I had to search for that one and it's from a private blog. Review sites give you 3 to 4 comparable bars at most, and always (16:10) because those resolutions are never multiples of one another. The circus around aspect ratios is a separate topic though.
To see how resolution and framerate are related, the number of pixels has to be calculated to make a proper graph.

It's immediately apparent that (1) one data point is a bit off (2) the graph is more linear and monotonous than the bar diagram above, once we put actual values on an actual x-axis.

Generally when one variable increases and the other one falls, the relationship that comes to mind is "inversely proportional" (1/x). And the bar wouldn't be a line, but a butterfly-shaped hyperbola. I assume, that's what we are seeing: the almost straight part furthest to the right that asymptotically approaches zero as megapixels are increasing.

The funny thing is that a hyperbola would suggest that lowering the resolution increases the framerate more than just linearly, because it curves upwards going from right to left! Thus lowering resolution will give give one a more than linear increase in FPS.

With at a resolution of 2x2 the frame rate probably could be extremely high.
In practice the memory buffer becomes extremely large and memory bandwidth becomes a bottleneck especially if any kind of super sampling like MSAA is used. Maybe there is some clever technology to save compute by cutting visual quality, but all I see is diminishing returns and never a frame rate high enough.