The current , August issue of Scientific American has a very interesting article regarding advances in real-time ray tracing for computer graphics.  I checked the website, but the full article appears in the subscriber area only.  In a year or two, the dream of no longer having to wait forever for a render may  come true.

RTRT is going to require forking out large amounts of cash for some time to come.
Significant speed up only happens when you many processors working in parallel.

is there a way for me to see part of the article even if i dont have an acount?

Quote - is there a way for me to see part of the article even if i dont have an acount?

Unlikely, but you can find some information here -> http://www.openrt.de/ and at -> http://www.tgs.com/ and at -> http://www.mc.com/products/view/index.cfm?id=20&type=software
Take note of the hardware setups on the examples page at the Mercury site (tgs.com).

The beginning of the article does appear in the regular area of the website.  I'm not a subscriber--I found the article in the August print issue of Scientific American

greg

I posted this also in the Bryce forum--but I thought it would be helpful to have it also here for Vue. 

The article has a box comparing traditional ray tracing and the real-time

traditional----Ray tracing renders a 3D scene by shooting vitrtual rays through the pixels of the 2D display.  The scene is stored as a database of objects, which can include curved as well as flat surfaces.
When a ray hits an object, the system launches "shadow" rays toward each light source in the scene and rays to test for indirect lighting by other objects.  The ray tracer also checks whether the surface will reflect, refract or simply change the color of the original ray.
Thanks to its recursive nature, this method can render a scene accurately in just one pass.

And now the Real-Time--the article describes 3 kinds of advances
1-Running rays together in parallel now happens at several levels within real-time ray tracers on desktop PCs.
Programs group similar rays into "packets," then march all the rays in a packet in lockstep through the same set of computations.

2-Acceleration structiures split the 3d scene into a hierarchy, called a kD-tree, organized so that each section carries roughly equal computational cost.  Rather than testing every ray against every object, the renderer follows the tree from its trunk to the appropriate "leaf" to find the few objects that the ray might hit

3--Cusotmized microchips built last year at Saarland University run at a mere 66 megahertz in prototype versions yet can render some ray-traced scenes more quickly than a 2,600 megahertz Pentium-4 system
Once it is commercialiezed, the "ray processing unit," or RPU should run roughly 50 times as fast- more than speedy enough for interactive software.

I hope that helps clarify things.

greg