I just love it!
Intel entwickelt Raytracing-Engine für Computerspiele | heise online (german)
Pff… raytracing… pff…
Please no flamewars. I’m just expressing my opinion 
If this is running in decent framerates on standard cpus then this would rock so much. Getting rid of all the crappy shaders, drivers,… shadows for everyone! I love the theory behind.
Why do you think that raytracing isn’t good for games?
Don’t worry, no flamewar, I am just interessted in your experts-opinion.
I never said it’s bad. I just said “pff” ducks
I think that raytracing is not a magic bullet to solve all the problems. Sure, it can do perfect chrome spheres nicely… but dynamic scenes, non-direct illumination and so on are still hard. At some point it will be viable for games, sure, but at that point good old rasterization might also have improved in speed by leaps and bounds. I’m not sure which one will be more widely used in ten years. Animation houses to this day mix both and only use raytracing when they really have to…
Pfff… no fart by the way will be the future in x years.
I think that’s the same crew that’s been demoing a real-time ray-tracing demo for some time (I downloaded it several years ago, it was quite impressive then, too).
It doesn’t really reduce the need for shaders, etc. – it just pushes the hard work into other nooks and crannies*, and the overhead is quite stiff.
*It’s very sensitive to complex geometry (relative to the current realtime rendering systems) so you end up spending a huge amount of time painting textures and simplifying geometry).
I thought one of the benefits of these new real-time raytracers (Saarland’s, specifically) was their ability to handle massive polycounts. The graphs I saw showed a logarithmic correlation between polycount and framerate. There’s a big performance hit initially, but the next few million polys make a minimal impact on fps.
Their big selling point is that you won’t have to worry about optimizing geometry. One of their demos is a real-time visualization of a Boeing 777 CAD file. It’s completely interactive with every bolt and doohickey rendered for your examination. http://www.intrace.com/gallery/boeing777.php
Thanks for the explanations.
What impresses me so much is the statement that raytracing scales logarithmic in case of increasing complexity of the graphics. The raster method scales linear. So, at some point you will nead less power to create more complex graphics with the raytracing approach then with the raster method.
I definitely agree with Aras.
I’ve been in the CGI business for 15 years, and way back then glorified phong shading was the order of the day. I used Electric Image Animation System, which basically rendered using a glorified phong shader. However, it produced a much better and more natural results than the raytracers back then - and it did so in a fraction of the time.
So, what I’m saying while rambling, is that I think realtime raytracing will have to improve massively before being able to generate as natural and good looking images as our current methods do. Sometimes cheating looks better than the cold mathematical approach that is raytracing.
My 0.02 swedish kronor. 
Killer question: now make that geometry move.
Most raytracing performance graphs our there are performed on static scenes, where all the geometry is nicely put into kd-trees, hierarchical grids and whatnot.
Raytracers scale better with geometry count, but moving that geometry is a hard problem (unless you stop using any acceleration structures… and your performance falls through the floor). Rasterizers scale worse with geometry count, but they don’t care how the geometry moves.
Of course, there is a lot of active and interesting and promising research going on about how to efficiently raytrace dynamic scenes, but they still have some way to go until that is actually doable in practice.
thX! As a total beginner in computer graphics I believe almost anything some experts tell me. We will see what the future will bring but I like the thought of high scalability and independency of graphic cards. More CPU’s = more power.
And I am glad that users of game engines don’t need to know much about the render techniques behind. :twisted: In a perfect world OTEE can switch to a raytracing renderer in a few years and the Unity users even do not recognize the switch. 8)