WE NEED AMD ATI GPU SUPPORT

[candemirdesign]

please suıpport ati or amd gpus
they are more more faster...

LUXMARK ROOM SCENE:
by tomshardware

AMD RAdeon 7970
979 sample/second(AMAZING)
AMD RAdeon 7950
811 sample/second(AMAZING)

Nvidia Gtx 590
663 sample/second (disapointment)
Nvidia Gtx 580
416 sample/second (disapointment)
Nvidia Gtx 680
284 sample/second (wooow disapointment)

[candemirdesign]

and i dont want to wait no more ((

[a.boeglin]

The problem with your bench is that you compare nvidia openCL vs ati openCL. And as far as I know, nvidia cards are way slower than ati cards on openCL softwares. But cuda is way faster than openCL and a lot of softwares just don't or didn't base their lines of code on an archaic openCL. Maybe things have changed since then, but openCL dev seems pretty slow. Maybe you could give a link of your benchmarks that we have the whole context as well. But from here it seems that the tests aren't fair, and statistics are always saying what we want them to say, we just need to show them the right way. And the other problem when working with 2 APIs is the support of some features, it means one version would support SSS, the other one not ( as a stupid example which isn't true of course ).

[Jaberwocky]

of course there is another way.

Open ACC

Supported by Nvidia but that is also an open standard.

See links

http://www.nvidia.co.uk/object/openacc- ... es-uk.html

http://www.openacc-standard.org/

[Reggie]

It would be nice to have AMD support, but I don't see why we "need" it. Octane is still in Beta, and it needs a lot more features before it can be considered a finished product. Then it can worry about supporting two separate architectures. CUDA is also the most mature and supported form of GPU development environment.

[AndreaMannori]

I agree since AMD GPUs have much more Gb of ram...

[steveps3]

I agree since AMD GPUs have much more Gb of ram...

They do? Since when? As far as I can tell they have exactly the same amount of Ram as Nvidia cards.

I would much prefer the Otoy concentrate their efforts on developing the Cuda version of Octane than for them to spread their efforts over multiple versions.

[matej]

I agree, features are more important atm.

But if Nvidia continues this "differentiation" we are starting to see and their prices for gpgpu cards will soar, then at some point it will become viable (price/power ratio) to port Octane to OpenCL and use cheaper AMD cards with OpenCL instead of expensive Nvidia with CUDA. Until AMD also decides to "differentiate"

[glimpse]

I agree since AMD GPUs have much more Gb of ram...

Since when? Does amd even have any gpu that would have 6gb of vram? =)..

[steveps3]

So I think that all agree (well nearly all) that Otoy should carry on ignoring AMD at least for the time being.