>> I was wondering if Povray rendering speed could be increased by using the
>> graphics card or a customized fpga if available. The graphics card may be 
>> a
>> little difficult to utilize (although NVidia has written their Gelato 
>> renderer
>> to use their GeForce cards), but how about an fpga directly connected to 
>> a
>> computer?
>
> Hi,
>
> this has been discussed several times before; a search on this site for
> "graphics card" gives more than 100 hits ;-). Read THIS
>
http://news.povray.org/povray.programming/thread/%3Cweb.4494e8a94ad1379aa6bb2e320%40news.povray.org%3E/?mtop=1
> as an example...
>
> Have a nice weekend
> Karl

The current methods of increasing POV render speed
by hardware are...
1) purchase a faster multi-core computer
2) set up multiple computers as a render farm
3) rent processing time via online vendors like Amazon
($ 0.80 per hour per instance for a quad core equivalent)

It's tempting for people to think that graphics cards should
be good at rendering ray-traced graphics, but they are not.
Typically a GPU is optimized around z-buffered scan-line
graphics. If you started from scratch you could put an
OK version of POV into hardware, I think FPGAs could be
used to prototype such a ray-tracing card, striping out
unneeded commands from some microprocessor design in
order to fit more dedicated floating point onto a chip, but it's
more likely that NVIDIA or someone will add in ray-tracing
features to their newest card as part of a hybrid
ray-tracing/scan-line design. It certainly would take a group
of smart electrical engineers, I'm perty sure that the expertise
around here tends to be in software not electrical engineering.

Post a reply to this message