Linus Torvalds (torvalds@linux-foundation.org) on 10/2/10 wrote:
---------------------------
>Rohit (@.) on 10/2/10 wrote:
>>
>>May be increasing the ppi (a la iphone 4) will help?
>
>I'd love to see more high-resolution displays, and it
>is going to drive some graphics performance work
>in the future, but it won't necessarily really do much.
>
>For example, say that you currently run at 1920x1080 or
>something, and move up to 2560x1440. You're doubling your
>pixels. So add a second (or third) screen to drive things
>up further still.
>
>What happens? One thing that happened is that you're just
>talking about a few years of graphics card performance
>improvement. Think about it. And think about how quickly
>people upgrade their monitors. End result: your resolution
>isn't growing fast enough for it to matter.
>
>The other thing that happens is that the quality of those
>pixels don't need to be as good. Let's face it, even when
>you run at 1920x1080, things like 4xAA and fancy filtering
>don't matter hugely in moving pictures. There's a reason
>why video is considered "HD" when you do it at 1280x720p,
>but a still picture isn't considered high quality unless
>it's in the ten megapixel range.
>
>Yes, yes, graphics card vendors (and game review sites)
>do the whole "compare screenshots" thing to see if you
>can tell the difference between the quality of the
>anisotropic filtering. But nobody actually tends to care
>during any real gameplay.
>
>The big deal about things like good AA/filtering is that
>the human eye is actually good at seeing certain kinds of
>pixel effects. Things like the "marching pixels" effect
>when moving around a bit and you see creeping edges or when
>you can see odd sampling effects from textures etc.
>
>But at high resolutions and moving video, you don't need
>to do expensive multi-sampling etc. You can choose to just
>blur the picture, and give it a lower resolution. And no
>human will ever tell the difference.
>
>So as pixel counts go up, the opportunity for just faking
>it goes up too. So while you might see an order of magnitude
>more pixels on your screen some day, I doubt you'll need
>even an order of magnitude more processing power to make
>the picture look fine.
>
>So no, extra resolution won't drive an infinite amount of
>future graphics performance.
>
>Other things will do it, of course. People still do want
>more realistic effects etc. So I'm not saying "640kB of
>memory is enough for anybody", I'm just saying that there
>are lots of diminishing returns. I also guarantee that
>noise from your computer is a real issue, very much during
>gaming too. Not all games are all about loud explosions,
>and not all of the experience is about how many pixels you
>can push and how quickly.


But yes, human physiological limits are going to bound the growth of a number of things like that. For example no one needs audio sampled at 1MHz.

Human vision has a resolution of about 1 arc-minute (with a few people able to do a bit better than that), or about .15mm (about 180dpi) at half a meter (approximately what you'd be doing on a laptop. On a desktop the display is typically a bit further away, especially larger displays.

You need pixels a bit smaller than that because of the various edge effect from the square pixels mapping onto an analogue field of vision, but still, the limit is not all that far away.

Likewise, printers are probably not going to be seeing huge increases in nominal resolution, even though printed material is often looked from a much shorter distance. If printers do increase resolution it'll be to further color rendition.

On the color depth front, we could use a few more bits, but again, it's limited by the dynamic range of the human eyeball. Something on the order of 10**12:1, from maximum brightness to minimum night-adapted. Of course the short term range is much narrower than that.

Nor are we likely to see huge size increases in displays (I've got two 24" monitors on my desk, how much bigger can I use?), excepting displays meant to be viewed at longer distances.

But for monitors we're definitely on the diminishing returns slope, even though we might ultimately see an order of magnitude increase in the number of bits on the display.

With the usual caveat that there will be specialty application that need more.