OTOY Forums

hy folks, i recently saw this sighraph vid about GPU ray tracing (http://www.youtube.com/watch?v=0IC2NIogWR4) , more in paticular about OPTEX 2.5, somewhere in the vid they talk about how they are getting rid of the memory limits of the onboard graphics card memory.
Both for data file bottleneck and the limitation off how many textures can be used in the engine...since Octane has its own SDK and does not use Optix in any way....well my question to the developers is will we have some benefits from these new techniques or are these kind off implementations impossible because off Octanes nature and different program techniques...?

High regards,
Polytek

How can you get rid of memory limits? If the card only has 2 gig of ram then it only has 2 gig of ram. You can't magically make it have 8 gig. As far a I know the only limit is the amount of money that people are prepared to pay.

steveps3 wrote:How can you get rid of memory limits?

You can't. They are talking about optimizations their SDK does and how to get around some limitations, like the limit on texture objects. So the title is kinda misleading

I'm not knowledgeable on the CUDA level, but on the user level, RS can do a lot in the field of memory optimization, by giving us moar powerful shading system - procedurals, tiling options, full transforms for all textures, per-texture UV coordinates, more pixel blending modes, RGB<->HSV, RGB<->grayscale, etc etc... stuff that has been talked over and over in the last year...

In one RGB image there is a big amount of data -> the more tools you have to extract & manipulate that data, the less images you'll need to load per material.

CentiLeo is intersting....