Some people have noted that our “Fusion” brand reminds them of the “fusion” code name for a piece of silicon containing both a CPU and a GPU. At the time we introduced the Fusion brand we conveyed that it was more than a name, that it was about our technology partnerships, how we engage with our customers and how we help innovate the industry.
We still believe that.
One of the ways Fusion comes to life is in how a PC is much more than an X86 CPU. The various elements of the PC platform coalesce to improve the user experience and enable the system to get its work done, seamlessly. In that same sense, let us consider how the Fusion brand is being made real every day with ATI Stream technology.
Our ATI Stream technology, which is about sharing workloads between the CPU and the GPU, divides software execution between the various compute engines at the platform level. Today, more and more applications are highly parallel in their execution, either because the task at hand lends itself to this approach or because software programmers want to take full advantage of the increasingly popular multi-core processors, or perhaps a combination of both.
This may be a behind the scenes, but its impact is likely very real for you.
A good example is how video encoding and decoding − or “transcoding” as it is known − has become a mainstream, daily event for the YouTube generation.
It has fast become central to many consumers’ experience to make digital video content available on any screen size and resolution, high definition or standard, regardless of how the original content was formatted. What many consumers don’t realize is that this is a highly complex problem that can consume huge amounts of compute resources, often requiring hours rather than minutes or seconds to complete.
As it turns out, transcoding is a highly parallel computing problem, with relatively few instructions being hammered against a huge amount of data, something that graphics processors are particularly good at doing. To that end, we are working closely with ISVs like Cyberlink and ArcSoft to help them embed a video transcoder in their applications that can leverage ATI Stream technology to greatly accelerate transcoding. We even offer a free, basic utility that performs video transcoding on many of our latest ATI Radeon™ graphics cards, using the latest ATI Catalyst™ 9.5 driver. Click here for a blog that includes a “how to” video showing how easy it is to work with video files and the ATI Video Converter.
Beyond transcoding, programmable ATI GPUs are rapidly expanding their role in compute platforms to perform diverse tasks, such as taking advantage of their hundreds of cores for protein folding in life-science research, seismic simulations for geophysical exploration, data encryption and many more applications through ATI Stream. You can see here how financial data analysis is greatly accelerated with a GPU.
What is particularly exciting about ATI Stream technology is where it is going. Programming multi-core architectures is getting more standardized and straight-forward through the OpenCL specification, something AMD is already leveraging. AMD showed an OpenCL demonstration of Havok Cloth™ game physics at the recent Game Developers Conference (GDC).
And because OpenCL is exactly that − free and open − the barrier to entry for programming multi-core systems is getting lower.
Later this year, AMD plans to release a new ATI Stream Software Development Kit which will make OpenCL widely available for AMD hardware. Also planned are modifications at the chip-level designed to leverage the OpenCL specification.
I will talk more about ATI Stream throughout the year with other specific examples of how AMD’s unique combination of CPU and GPUs are supercharging performance. In the meantime, you can read more about this here.
And as we like to say, The Future is Fusion.
Nigel Dessau is senior vice president and chief marketing officer at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such links sites and no endorsement is implied.


(4 votes, average: 4.00 out of 5)
(4.69 out of 5)
#1 by Surya - May 30th, 2009 at 01:13
Nigel, I want to build my home based supercomputer using AMD Graphics based. I want to connect two PC using dual gpu Radeon 4870X2 based but i could not find a card containing connector that could connect 8 lanes PCI express like ATI XGP. I just want a converter card that plugged in an empty pci express slot with 8 lanes external connector. Where I could find this solutions?
#2 by Nigel Dessau - June 3rd, 2009 at 12:05
At CES we used a Phenom-II based motherboard all sitting on internal PCIe to create a “Fusion Render Node” configuration for a technology demonstration. After CES I moved toward an Opteron based enterprise class design which employs InfiniBand( above 2Gig) to gang up multiple FusionRenderNodes but this maybe outside your budget. Bottom line is we don’t know of a card that can be hosted on the PCIe and then fan out into multiple lanes.
#3 by Andre Straker - October 2nd, 2009 at 09:16
Nigel,
I am of the opinion that GPGPU computing reaches into the embedded arena. From robotics to security to automotive, GPGPU has the ability to set some applications on fire. I am still amazed at the performance improvements obtained.
What is AMD doing to address this, potentially huge, market?
Regards,
Andre