|
|
Ken wrote:
>
> Mark Palmquist wrote:
> >
> > Just a note, IBM's Pacific Blue posted 3.9 TerraFLOPs when running for Los Alamos
> > Labs. and
> > I just saw a note about a computer that did 12 TerraOPs (4 bit intiger math NOT
> > Floating point). Port POV to Pacific Blue and I think most scenes would render in
> > 20 Minutes or less, even 'STRIKE' at 1024x768 AA3, Focal blur and all lights
> > converted to area lights! It would be fun to watch it anyway.
> >
> > Just my 2 cents......
>
> I doubt the majority of scenes would take even 5 minutes with that
> kind of power computational power to work with. Only in your last
> examples would it start to slow down any. It would be nice to try
> that 100,000 overlapping transparent spheres, with the max trace
> level set to a few thousand, to see what would happen to it's render
> speed.
>
Reviving an old question about using raytracing/media/radiosity etc. in
realtime renders...
I calculated that the computer would have to be ~18,000 times faster. Well,
consider FLOPS: an average PII curretly gets about 300 MFLOPS. Pacific Blue
at 3.9 TFLOPS would be about 10,000 times faster. Getting close... Now if
only I had some extra money!
Margus
Post a reply to this message
|
|