OpenRT/RTSL - competition for CUDA/OpenCL

A few might remember the startup Caustic Graphics which announced a Raytracing Addoncard about 3 years ago to “revolutionize” raytracing.
They planned to release it along with CausticGL a software library, targeting the automotive and movie segment.

Then it became silent. In 2010 Caustic Graphics was bought by Imagination Technologies (leading in SoC GPUs - PowerVR).

On the CES 2012 it was announced by Imagination Technologies that they’ll release a card this year.

It’ll use PowerVR OpenRL as raytracing interface, which is going to be platform independent. It’ll be coded in RTSL (Open Raytracing Shading Language) derived from OpenGL/GLSL.
Currently there’s only a Windows download for V1.0 but OSX and Linux are on the way.

Rumors have it that there was a prototype card already in a backroom at the CES 2012 to preview for some selected people, and that in August at the Siggraph 2012 they’ll announce it officially.

At the moment they offer Brazil 3.0 as interactive realtime rendering engine and plan to support 3D apps via plugins.

https://vimeo.com/31864755

I think this is quite intresting, if it really becomes all that open it would be nice, so to say an enhanced version of OpenCL, tailored soley for raytracing.

It’ll depend on a particular hardware from one manufacturer in the start, true that, and it might take some time until other manufacturers pick up the production of cards supporting it.
However I enjoy the thought to be able to choose a good card for viewport performance and have the choise to buy a card soley for raytracing completely independent.
And on top of that, if you watched the video, you see that OpenRT works on CPUs, GPUs and RTUs, even supports IGPs of Intel CPUs.

Remains to hope that OpenRT will work out and there’ll be a healthy competition with raytracing cards.

The sad story with new standards usually is:
http://imgs.xkcd.com/comics/standards.png

Anyways, it might be intresting to already look into it a bit with Cycles in the drydock, can’t be the worst idea to plan ahead and look into new promising technologies, or like I say, show some sceptical interest :wink:
Better to work now towards the possibility to implement OpenRT easily, rather than rewriting everything later.

I don’t think this will be relevant in the future. First of all, calling something “OpenXYZ” by providing a specification but releasing the only implementation as proprietary doesn’t really fit my idea of “open”.
Also, there is a longlasting trend in computing towards de-specialization. There’s little point in creating dedicated raytracing hardware, and the release of this SDK for heterogeneous platforms is a testament to that. Raytracing isn’t a problem that benefits greatly from being implemented in hardware in the first place - and even if you do you are guaranteed to be outperformed by the raw power of more general solutions pretty quickly, simply because of the economics of it.
As for the shading language standard, I ask: Do we really need this? Very rarely a shader can truly be re-used in a different environment (this is especially true for something like GLSL) and therefore a specialized language has little benefit over a more general one.

Remember AGEIA PhysX, separate device dedicated to calculating physics?

Yup, now it’s in Nvidia’s cards…

Yup, now it’s in Nvidia’s cards…

Technically, they’ve just rewritten the software to run on the GPU. There is no special hardware for physics on any GPU.

Well, that was my point, but I guess I should have explained better :o