First I’d like to start by trying to explain a difference between Open Standard and Open Source for those of us who are getting confused. Yes, I may have been part of it myself in my last article. I kept on mentioning Open Source and Open Standard as if they were the same, even though while I was composing the Editorial I knew what I was trying to get at.

Open Standard, which would be OpenCL, is something that is available for all to use, but even though it can be used by everyone, the actual source code (SDK) and it’s tools are not open. They cannot not be edited by the user nor can it be seen. Open Source, on the other hand, is something that can be used and edited by anyone and it is welcomed, since more sources can be added and, in theory, will create better development, thus increasing the amount of resources available to use.

What is the open standard: OpenCL or the OpenCL SDK from either ATI or NVIDIA.

What is open source: the physics library that Bullet has compiled that will work in unison with OpenCL.

In November of 2007, ATI announced, with their Spider platform, Stream, which was a parallel processing application used to transcode video via AVIVO. I was present for the launch of that platform and can remember Pat Moorehead (AMD) sitting down at the computer showing me how much quicker transcoding could be with this breakthrough technology. Honestly, that is probably the last time I heard about it. It seems ATI just dropped the ball and it was pushed to the side. While having my first briefing with ATI, I asked Eric Demers about this and he agreed that, since the launch, there was not much further development with Stream and it was a mistake.

Stream, as we knew it, was proprietary, the same as CUDA, which is by NVIDIA. Now, ATI is stating that with OpenCL, Stream will take on a whole new meaning, it will be something everyone can implement and can be used with any GPU. In theory, yes, but we need to remember Open CL is only an open standard. The SDK is closed, so if a program is developed using the ATI Stream SDK and ATI created that SDK, don’t you think there would be something in those development tools that would make the code be more favorable to ATI than NVIDIA? The same could be said if the NVIDIA SDK was being used. To date NVIDIA has the only certified and working OpenCL SDK for GPU.

This is the question that I pose; ATI has said that they were in the process of developing a Stream SDK for GPU acceleration in OpenCL, which is an open standard, but the source code is closed. To me, this is the same as proprietary. If I can’t change it, even though I can freely use it and am limited to use the development tools in the package, I would have to say proprietary. The reason for this question is because I have been hearing the word proprietary being used a lot lately when ATI talks about PhysX or CUDA. OpenCL is an open standard but I still see proprietary parts to it and that would be the SDKs.

I am trying to line up another briefing with ATI to get further information on this issue. I have received the acceptance and will be allowed another briefing but the date and time has not been set, as of yet. I had hoped to have these questions answered and also give ATI the chance to possibly give a clearer explanation without the marketing hype. I will amend this part of the editorial after the briefing.

On a side note, I would like to refer to my third paragraph about Platforms. I have been reviewing AMD products for a while. I started with a site I worked for and now have continued with Hi Tech Legion. I have been reviewing AMD products as a platform for quite some time but it was brought to my attention, by a colleague of mine, that no one spoke of any AMD platform at all for this launch. As a matter of fact, after I checked out if the information was true, I came to the same conclusion. No one that reviewed the 5800 series card tested it on a platform with an AMD processor. What happened? I can remember getting lambasted for not testing a video card cross platform on a launch. I know for a fact that these same sites have reviewed the platforms, so they have to have the processors, so why not? Is there something we don’t know about? Could there be a possible bottleneck with the new ATI card used in combination with an AMD processor?

Or is it something else? I feel that was an interesting question posed by Ed Crisler and here is link to the article on his blog. The Card that Killed the Dragon?

Part 1

Part 2

Part 3