WE NEED AMD ATI GPU SUPPORT

[GR1F1TH]

You know it's funny that since Nvidia went to single precision they are going to similar architecture that AMD uses. I don't doubt that Cuda 4.0 will unlock the potential in their cards for octane it's just a shame that it's so far out when the cards are here now (well at least the 670's are in stock). Oh well, I'm eying those 590 prices right now.

[unit102]

I own "ATI sapphire vapor-x 7970 6GB" & MSI Lighting GTX 580 3GB

I wish one day and to be soon to make my investment on sapphire 7970 6GB worth it because the processor is good and plenty of ram on board. just imagine having
2 of these or 4 .. my job is to create 3D visualization and I work closely with engineers (create 3d pipes, structures, electrical cables, etc.) and it
becomes a very heavy scene.. This is another market comparing to small scenes. so larger memory makes things more possible, atleast for me if it comes to GPU rendering.

I would like to thank the entire team for Octane Renderer, I love Octane for 3Ds max plugin its just I wish I can take it further with me (to be part of my job's work) for now I am using it for my personal use.

[Faizol]

It's not true to say that OpenCL is slower than CUDA. In a properly designed architecture using OpenCL, the performance should be comparable to the CUDA version. Documented test on OpenCL vs CUDA and profiling with use case analysis are available on the web if one wants to check on OpenCL and CUDA performance comparison.

However just by going the OpenCL route, there's no guarantee that the current Octane performance on Nvidia card can be replicated on AMD's gpus (most likely the performance will drop). And also OpenCL doesn't have the granularity as provided by CUDA in terms of the framework features and accessing hardware details.

AMD's cards are good for games, but their hardware are not designed for gpu compute from ground up unlike Nvidia's card. More or less it'll be like how gtx580 compares to gtx680.

[matej]

I don't know if OpenCL could be as fast as CUDA, all I know is that Cycles which is opensource and part of a program built around non-proprietary standards, doesn't have functioning OpenCL rendering. Supposedly there is support for OpenCL in Cycles and maybe 1 out of 10 people got it working on their machines (I never did), but nobody is using it in production and the option is there more or less just for geeks to play with it and do some testing (even the Cycles lead programmer says so)...

And that tells a lot about OpenCL (non)usability in rendering (maybe in other fields like science computing, OpenCL is as viable as CUDA, but we don't care about those fields).

[3dworks]

yes, please support ATI cards in the future as well, especially since ATI cards are still the most popular mac pro cards around. support for nvidia on macs has been always weak, i replaced my 285GTX card for mac long time ago with an ATI 5870 which is way faster when used with adobe products like after effects and photoshop.

cheers

markus

[kavorka]

I wouldn't hold your breath for ATI support. If they ever do it (which I wouldn't be surprised if they didn't), it just wouldn't be as fast. There is a reason why Refractive software (and other companies) support CUDA cards for rendering. Unless something in the future changes with ATI cards, CUDA (GeForce) is just far superior.

[matej]

I don't want to wait for a fully-fledged Octane renderer until 2025, cause I'll be in my forties then, so please don't bother with OpenCL.

I would prefer to use ATI cards too to run my programs & games, but not if this means crappy / unsupported / featureless / buggy / slow rendering.

[Goldorak]

Moving from CUDA to CL is not trivial, nor is supporting both codebases at the same time (at least at the quality and feature level we all want for Octane). It is not a step to be taken lightly. Even renderers that may start on OCL often end up adding CUDA backends down the line to get better support on current and future NVIDIA hardware.