Hardware architecture for real-time ray tracing

[Anonymous]

For the first time SaarCOR allows for displaying highly complex objects with millions of triangles with photo-realistic image quality at realtime. The architecture is highly flexible and scalable and can be optimized for different application scenarios. A first prototype of this graphic board purely based on ray tracing is available.

The chip is 3-5 times faster in raytracing than a Pentium 4 CPU, although it clocks only at 90 MHz.

http://www.saarcor.de/

[Anonymous]

Ah, raytracing rears it’s head again. (Actually, if it had a head, it’d probably be bald and really really polished…)

I’ve been told that there’s some schools of thought which reckon raytracing will ultimately be what wins out, from a graphics technology point of view, partly because of the fact it’s so highly parrallisable.

It’s an interesting argument, as the trend in hardware R&D is increasingly towards lower clockspeed multiprocessor architectures, and hench algorithms that parrallise well. (It goes like this: Clockspeed seems to be hitting physical limits, and, for example, both AMD and Intel are starting to go dual core – how long before they go quad, 8-way,n-way, oh, and look at Cell etc).
In a world where the number of available processors is not really limited, but the clockspeed of each processor is, maybe raytracing is the way to go.

Still, there’s been so much research into the evolution of a modern gpu, I imagine it’s hard to catch up to.

The chip is 3-5 times faster in raytracing than a Pentium 4 CPU, although it clocks only at 90 MHz.[/quote:5e437e821a]

Interesting… but I have to wonder, how would a modern Nvidia (for example) GPU, but clocked at 90Mhz, compare to a P4 running the same graphics pipeline in software?

[Author]

Posts