AMD Boosts Application Development for AMD Fusion APU-Powered PCs

[btarunr]

AMD today announced availability of the updated AMD Accelerated Parallel Processing (APP) Software Development Kit (SDK) v2.3 with full support for the first AMD Fusion Accelerated Processing Units (APUs), OpenCL 1.1 and AMD Radeon HD 6900 Series graphics. AMD APP technology enables AMD GPUs and CPUs to work together to run many demanding computing tasks faster than possible with just the CPU alone.

AMD Accelerated Parallel Processing SDK v2.3, previously known as the ATI Stream SDK, empowers software developers to write new applications that can take full advantage of the parallel processing power of heterogeneous computing platforms, such as those based on new AMD E-Series and C-Series APUs that combine a multi-core CPU and DirectX®11-capable GPU on a single die.

“When developers harness the power of parallel processing within our APU designs, they can fundamentally change the PC experience to help not only make it faster, but also to create new possibilities in software,” said John Taylor, director of Client Product and Software Marketing, AMD. “In 2008, AMD was the first processor design company to embrace OpenCL, and as such, we have made significant progress in parallel processing innovation. Our vision has been realized with the widespread availability of the first PCs powered by AMD Fusion APUs.”

“Industry standards like OpenCL and OpenGL, common APIs, and tools like the AMD APP library and SDK give application developers a shared environment that allows them to harness maximum hardware performance for amazing application experiences across platforms,” said George Tang, vice president and general manager of ArcSoft’s Video and Home Entertainment Group. “By supporting these open standards, AMD is giving developers access to a programming environment that helps them more easily navigate the recent advances in computer hardware, including AMD’s APU technology.”

In addition to support for the first AMD Fusion APUs and OpenCL 1.1, AMD APP SDK v2.3 offers improved runtime performance and math libraries for OpenCL.

AMD Developer Summit
Improving OpenCL performance and programmability on AMD platforms is a key initiative for AMD, and this summer, AMD technical personnel, AMD executives, developers and partner companies will gather at the AMD Fusion Developer Summit to discuss industry standards including OpenCL, and to explore how to best optimize applications for heterogeneous computing.

The inaugural AMD Fusion Developer Summit (AFDS) will be held June 13-16, 2011 in Bellevue, Washington. Presentation proposals are due by February 4, 2011. AFDS will gather developers, academics and emerging innovators to learn more about heterogeneous computing, APU technology, parallel processing and developer programs like the AMD Fusion Fund.

Get the Download AMD APP SDK v2.3 here.

[meirb111]

amd aims high but lets hope it will not end up as useless as Stream Technology

[TheMailMan78]

Is Nvidia countering this in any way? Honest question.

[Imsochobo]

Originally Posted by meirb111

amd aims high but lets hope it will not end up as useless as Stream Technology

Stream is almost ~as good as cuda.
Marketing... marketing..

Amd's tech is really good, but Opencl is what i'm rooting for!

[meirb111]

Originally Posted by Imsochobo

Stream is almost ~as good as cuda.
Marketing... marketing..

Amd's tech is really good, but Opencl is what i'm rooting for!

cuda has many software support stream has very few

[TheLaughingMan]

Originally Posted by TheMailMan78

Is Nvidia countering this in any way? Honest question.

Well they did license almost all their tech to Intel, so Intel may counter with an Nvidia right hook. We will have to see though.

I wonder. Does this mean you can build one program and run part of it with the CPU and then run another section of the code with the GPU, so you can benefit from whichever one gives the best boost.

[streetfighter 2]

Originally Posted by meirb111

cuda has many software support stream has very few

I thought Stream was just ATI's implementation of OpenCL? IMO OpenCL, by virtue of it's compatibility, is better than CUDA. However I am aware that CUDA has achieved far (far) greater industry acceptance.

Anyway about it, the Fusion platform looks really nice and I'll probably do some tinkering with it in the next several months.

Originally Posted by TheMailMan78

Is Nvidia countering this in any way? Honest question.

Well they did announce that they were making an x86 CPU (I think). I don't think it's a direct competitor, but it's still something (quite) different from Nvidia.

Additionally Intel is making Atom CPUs with an FPGA. I figure that will compete a bit with Fusion in the embedded devices market.

[Cheeseball]

Stream is a piece of crap. Burst reading/writing calculations are seriously in pieces, the way it handles global buffers is a pain, and their ISA progress (which is NEEDED in multi-GPU computing) is so damn slow.

What makes it worse is that their progress for shifting to OpenCL is even slower.

This is probably the only reason why I have a GTX 460 in my 3rd PCI-E (4x) slot along with my Crossfire'd HD 5770s, because CUDA and it's compatibility to OpenCL is progressing pretty damn well. And NVIDIA actually HELPS developers out directly through e-mail and the official forums.

C'mon ATI, stop concentrating on games alone and fix your drivers and SDKs.

[pantherx12]

Originally Posted by Cheeseball

Stream is a piece of crap. Burst reading/writing calculations are seriously in pieces, the way it handles global buffers is a pain, and their ISA progress (which is NEEDED in multi-GPU computing) is so damn slow.

What makes it worse is that their progress for shifting to OpenCL is even slower.

This is probably the only reason why I have a GTX 460 in my 3rd PCI-E (4x) slot along with my Crossfire'd HD 5770s, because CUDA and it's compatibility to OpenCL is progressing pretty damn well. And NVIDIA actually HELPS developers out directly through e-mail and the official forums.

C'mon ATI, stop concentrating on games alone and fix your drivers and SDKs.

Perchance have you been on the opencl and stream dev forums on the AMD website?

Because crap damn I see a lot of ATI folk helping directly there

[Cheeseball]

Of course. That's where I have to post my efforts in case something goes wrong. If you read most of the posts they are complaints (e.g. lack of certain functions) on the bad implementation ATI is using for Stream-to-OpenCL. It's good that they're adopting OpenCL, but they're very slow at doing it.

If they would fix their efforts on driver optimization and SDK integrity I would be using my 5770s for programming, but at the moment their implementation actually adds overhead.

What makes it worse is ATI's support for GNU/Linux. It was going great at first but then they dropped the ball after the 4-series, particularly a lack for support of UVD.

[leonard_222003]

I was just bitching in another thread about Ati not doing anything for accelerating some apps with stream/opencl and i was wondering if you guys know any program that can use Ati gpu's to encode , or some pro editing app that use Atis GPU for preview and exporting.
I'm really pissed i have this crappy HD4850 in my PC and render a 6 hour footage in 1,5 hours or 3-4 hours for HD material , friends with GTX285 and adobe premiere pro do this in half an hour , WTF Ati ?
So any app with real usefullness ( not tech demos and software in alpha stage or some it can do on paper , right now what it can do ? ).

[Wile E]

All I want to know is, will this finally get us some decent GPU accelerated encoders on ATI cards? nVidia stomps all over ATI for that.

[jmcslob]

Sure they will by the time the 12xxx series cards are out
You know within a year..LOL

[ViperXTR]

CUDA seems to be all over the place indeed.

http://forum.beyond3d.com/showthread.php?t=58195

although or most of the software there are just spinoffs and uses the same CUDA engine parameters