Do You Think Ray-Tracing Is The Future?
Posted: Thu May 22, 2008 5:25 pm
Just as Pixel/Vertex Shader revolutionized graphics in the late 90's - do you think Ray-Tracing will do the same for the 00's and 10's?
Personal computing discussed
https://techreport.com:443/forums/
mortifiedPenguin wrote:While I think that ray-tracing will bring good stuff to the table it isn't the be all and end all of graphics. There are still a few things ray tracing is inefficient at doing that could be done with the same quality using rasterization. So I thing that a hybrid approach is the most likely solution for efficiency balancing.
NeRve wrote:Also from what I understand Raytracing requires a complete overhaul of all current GPU architectures because they are all based on Rasterizing (all current gens down to the old Radeon 7000 and TNT2 cards.)
BoBzeBuilder wrote:I'm still not sure what ray-tracing does. Better lighting? I looked at some screen shots and they didn't wow me.
mortifiedPenguin wrote:NeRve wrote:Also from what I understand Raytracing requires a complete overhaul of all current GPU architectures because they are all based on Rasterizing (all current gens down to the old Radeon 7000 and TNT2 cards.)
This is true, although I believe there are some efforts to run ray-tracing on GPGPU systems. Wouldn't be quite as fast as a "pure" ray-tracing system though. I figure such an architecture would have both dedicated ray-tracing and raster components with some sort of interconnect and a third processor to combine calculations into the final image.BoBzeBuilder wrote:I'm still not sure what ray-tracing does. Better lighting? I looked at some screen shots and they didn't wow me.
Not necessarily "better lighting" but a more accurate way of calculating lighting as well as things like refraction and reflection. Essentially, a ray-tracer shoots out rays of "light" and calculates their exact bounces as they hit objects and every time it bounces it would change the color/intensity/hue of the pixel that it hits. Makes for some pretty accurate lighting (in terms of physical correctness) if there are enough rays being cast from the light source.
edit: A good example of what ray-tracing can do.
mortifiedPenguin wrote:To be clear, it traces the light "backwards" from the eye (or, more exactly, the viewport representing the screen) into the scene to see where that light came from. Since each pixel on the screen represents one potential ray, by following it backward to a light source you can accumulate all the effects of reflections, translucency, and opacity of the materials in the scene it might hit along the way.Essentially, a ray-tracer shoots out rays of "light" and calculates their exact bounces as they hit objects and every time it bounces it would change the color/intensity/hue of the pixel that it hits.
UberGerbil wrote:In case you haven't noticed, the 00's are almost over.
Meadows wrote:I won't be a believer until I see it work, but I'm pretty sure the big dogs are working hard to make it happen.
Meadows wrote:A very good example is reflection itself. Not all surfaces are perfectly reflective, which means over-using it on any surface would create an unrealistic ray-traced image with a worse impression than what the old ways (old in practical terms) could have given.
Meadows wrote:As JBI says, this has been a solved problem for over 20 years, which is why all the ray-traced images you can find (there are lots on the net, just look) don't make everything look like they're made out of liquid metal (unless that's the effect they're going for, in which case it's much easier to achieve than with rasterized graphics).Not all surfaces are perfectly reflective, which means over-using it on any surface would create an unrealistic ray-traced image with a worse impression than what the old ways (old in practical terms) could have given.
liquidsquid wrote:Quality of the game and quality of the graphics are (or should be) largely orthogonal issues. Look at Hollywood. You can have crappy movies full of gee-whiz special effects, and wonderful films that use nothing more than a single camera. But does it make any sense to say "Nobody can use any special effects in a movie because every movie that uses them is crap"? From what I've seen. the movies that manage to have fantastic effects and a fantastic story become true classics. "Hey, Steven Spielberg, you only get to use a black and white camera to make sure you focus on a the story!" What if somebody told Lucas he couldn't do the original Star Wars because a movie with lots of special effects is certain to be "stinky"?The point that needs to be made: Why? Will games/software truly benefit? As it is now, it seems the prettier the game/OS, the lousier the game/OS plays/is. All we need now is PCs that can render beautiful, stinky games. "Gee, look at how these Christmas Tree Ornaments sparkle! OMFG, who shot my head off!" or Look, my OS is prettier than yours, but I can barely run applications on it since it eats all of my GPU resources and memory!
These aren't necessarily conflicting goals. Much of the GPU development that would be necessary for real time ray tracing -- lots of parallel floating point execution units -- are exactly what you need to do physics. From the standpoint of the math that needs to be done, there's very little difference between projecting a light ray through space and projecting an object like a bullet. Whether it's your railgun or a light beam, you want to know where it goes and what it hits. The scatter of light off a diffuse surface is similar to the scatter of debris from an explosion. And so on. Something like Larabee, or it's successor, should be an awesome physics accelerator even if it never pans out for ray-traced graphics.To be honest I do not see how a card that can ray-trace has any real use on the computer other than a nice g-wiz. I would much rather have time spent on a more common-place physics accelerator to have interactive environments. Now that to me makes for a positive gaming experience. Besides, with most games I play, things are going by so fast, all you care about looking at it where the enemy is, and how to get 'em. Not that refractions through their blood plops are accurate.
just brew it! wrote:Meadows wrote:A very good example is reflection itself. Not all surfaces are perfectly reflective, which means over-using it on any surface would create an unrealistic ray-traced image with a worse impression than what the old ways (old in practical terms) could have given.
Only a very simplistic ray tracing algorithm would treat a surface as 100% reflective. Control of what colors are reflected (and how much) would of course be necessary. Surface normals can also be applied so that the surface is not perfectly smooth...
Meadows wrote:Can it model blurs? Yes, but it isn't terribly efficient. It would essentially be the projection of an identical or nearly identical ray onto a pixel offset of the "real" location of where it would be in a pure reflective object and blended with the color of the ray/pixel already there. It gets pretty computationally expensive at that point though, since we'll need to shoot out more rays to get the blending properly or we get some aliasing artifacts. That is actually one of the weak points of ray-tracing, when things are blended since we need more samples and more samples means more computation time. This would be a good example of where the hybrid approach makes sense, trace the reflection for accurate positioning/color then blur it using raster techniques.I'm more interested in whether it can model objects which reflect the world blurred, that is, have flawed reflection as opposed to flawless.
Meadows wrote:Of course (but it's more expensive). The issue is not just blurred reflection; it's also refraction. In fact the key to doing realistic-looking flesh is to do some internal refraction (subsurface scattering (PDF -- see images at the end) -- some light penetrates the skin and reflects out from underlying flesh/blood, which is also what gives the finest marble its lustre) as well as reflections from sweat/oils on the surface. (Here's a small PDF showing another example). Interestingly, many of the medical imaging techniques (CAT, MRI) use the same techniques -- they just are tracing back real rays, not imaginary ones. In fact, one of the origins of raytracing was a team of physicists ooking at working backwards to trace the path of subatomic particles, and when they showed off what they were doing to computer graphics professionals the CG guys said "You've solved the Hidden Surface Problem!" And the physicists said "What's the Hidden Surface Problem?"I'm more interested in whether it can model objects which reflect the world blurred, that is, have flawed reflection as opposed to flawless.
bogbox wrote:blablabla (...) the pc is dying (...) laptops with IGP are the future (...) blablabla
Meadows wrote:bogbox wrote:blablabla (...) the pc is dying (...) laptops with IGP are the future (...) blablabla
How can the PC die, if laptops are the future? A laptop is a PC.
bogbox wrote:good point !what about a ps3 with mouse and keyboad and a emulated xp is a PC too? yep:D
my bad ! labtops are not for games ,just ask a igp from intel
Meadows wrote:bogbox wrote:good point !what about a ps3 with mouse and keyboad and a emulated xp is a PC too? yep:D
my bad ! labtops are not for games ,just ask a igp from intel
In the broad sense of the phrase, a PS3 qualifies as a PC as well, and you don't need to emulate Windows on it and you don't need a keyboard/mouse for that.
Laptops can be used for gaming and are very advantageous for LAN party travelling. "Ask" any discrete GPU from nVidia, to put it your way.
You continue to make moot points.