I happened to find an article on Tom's Hardware dated July 22, 2009 that I
have not yet seen mentioned here in the povray groups. "When Will Ray Tracing
Replace Rasterization?"

http://www.tomshardware.com/reviews/ray-tracing-rasterization,2351.html

Gilles Tran's "Glasses" image is used. One he rendered with povray, though I
saw no attribution to Gilles or povray.

The article is not that deep, but perhaps of interest to some here.

Post a reply to this message

William Pokorny <pokorny_epix_net> wrote:
> http://www.tomshardware.com/reviews/ray-tracing-rasterization,2351.html

"To create soft shadows or diffuse reflections (like those you see in
brushed metal, for example), more advanced ray tracing techniques like
path tracing or distributed ray tracing are needed. But such
techniques require a much greater number of rays and are still far
from being feasible in real time."

  Actually "sending more rays" to simulate volumetric rays (which is
basically what those techniques are trying to achieve) is not the only
way.

  AFAIK Pixar used true volumetric raytracing in Cars, rather than classic
raytracing. In other words, rather than shooting rays (which are infinitely
thin lines), they shoot cones. Then they calculate the intersection of these
cones with the scenery/lights. These intersection calculations return the
percentage of how much the cone covers the tested polygon, which allows for
easy blurred reflections, shadows and other similar effects. The number of
"rays" remains the same.

  Of course the problem is that calculating the intersection of a ray and
an object is relatively simple, but calculating the intersection of a cone
and an object (so that you can calculate how many % is covered) is very
complicated. It simplifies things when you only have to do cone-triangle
intersections (basically you only have to calculate how many % of the
triangle resides inside the cone), but still.

-- 
                                                          - Warp

Post a reply to this message

Warp wrote:

>   Of course the problem is that calculating the intersection of a ray and
> an object is relatively simple, but calculating the intersection of a cone
> and an object (so that you can calculate how many % is covered) is very
> complicated. It simplifies things when you only have to do cone-triangle
> intersections (basically you only have to calculate how many % of the
> triangle resides inside the cone), but still.

Pixar's renderer turns everything into sub-pixel sized quadrilaterals, 
so they can definitely put this technique to work.

On the other hand, since Pixar has added ray-tracing to its in-house 
renderer in order to achieve certain effects, it's safer to say that 
pure tracers and pure rasterizers will both be replaced by renderers 
that do both.

Regards,
John

Post a reply to this message

"This is the worst possible scenario, because of all memory 
characteristics, latency is the one that has made the least progress in 
recent years, and there's no indication that's likely to change any time 
soon. It's easy enough to increase bandwidth by using several chips in 
parallel, whereas latency is inherent in the way memory functions."

I guess we'll have to wait for carbon nano structures, memristors, 
diamonds and/or super conductors to replace silicon and gold...

-Mike

Post a reply to this message