What I also get is – it’s about so much more than playing games.

The “more” is around something called DirectCompute. In many ways it’s the feature that my favorite applications are going to benefit from more than any other so I thought it would be worth spending a blog on.

DirectCompute is a component of the DirectX11 API to be released with Windows 7. It is fundamentally designed to enable GPU compute and enables applications to take advantage of the massively parallel computing power of the GPU.

What does DirectCompute mean for users? Simple: you have two great processors in your PC – one CPU and one GPU. And your system can use both to solve problems. For some workloads, like GPU-accelerated video transcoding and rendering, this combination can really speed up your throughput.

So as you begin compiling your holiday wish list, keep your eye out for systems with this astounding GPU compute capability. But here’s my caveat emptor. When you look at specifications of GPUs, it is important to note there are different levels of support provided for DirectCompute.

One level, which AMD terms DirectCompute 10, runs on the legacy path of the DirectX 11 API to support previous generation of DirectX 10/10.1 GPUs. If offered a system with this feature, just say no.  You should demand what we at AMD call Direct Compute 11. This is the only version designed to unlock the full feature set of DirectX 11 and, as I mentioned previously, the only architecture that is fundamentally designed to enable GPU compute. 

For developers, DirectCompute 11 represents a paradigm shift for GPU compute development. It will now be much more straightforward for developers to code using this architecture. No longer will they have to do contortions and jump through hoops to code what they want to. DirectCompute enables new algorithms that were not possible previously. Some examples of these advanced techniques are order independent transparency, ray tracing, better shadows, and depths of field. If you are interested in more details, you can read this white paper. DirectCompute is a de-facto  industry standard for developers of GPU-compute applications and as such, it should not be confused with  proprietary APIs  (and by “proprietary” I mean supported by only one supplier’s hardware).  

These are some of the reasons why at AMD, we are excited about the ATI Radeon™ HD 5800 Series graphics products. It’s not only the first and only GPUs in the market with full DirectX 11 support it is the only one to unlock the full feature set of Windows 7 and DirectCompute 11.We are proud of this technology leadership. We believe this industry standard will accelerate industry adoption of GPU compute applications running on Windows7 and add a new dimension to the end user’s computing experience.

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.

I remember the days of SNA. It ruled, it worked, it offered a huge jump forward and it got killed by the Internet, or more exactly TCP/IP. I remember the debate at IBM about the difference between “open”, “proprietary” and “de facto” standards. It seemed to make a big difference (at the time).

In those days, hardware or software specifications that were controlled by one company were what today we call “proprietary” standards. When a proprietary standard becomes widely used, it generally becomes a “de facto” standard even though it is not governed by a standards organization. “Open” standards, on the other hand, are typically developed by a standards organization or a consortium and are (with thanks to techweb), “available to the public for developing compliant products, open standards imply ‘open systems’; that an existing component in a system can be replaced with that of another vendor.”

So, to the point of this blog: graphics and Stream computing standards. There are many, but I’d like to compare one that is widely considered de facto, one that is currently proprietary but would like to become de facto, and one that is open.

Let us start with DirectX

Among other things DirectX is a Microsoft technology that gives the game player or video watcher accelerated graphics, video and sound performance within Windows. Today we are at DirectX 10.1 and heading to DirectX 11 later this year. DirectX and the similarly widely adopted Khronos Group’s OpenGL have generally replaced proprietary standards like Glide from the former 3dfx.

While you could argue that DirectX too started as a proprietary standard, mass adoption has made it a de facto industry standard – like it or not.

Next up: CUDA vs. OpenCL

There is a need in the PC world for a programming model that allows both the CPU and the GPU to work together to excite applications. At AMD we call this Stream or Accelerated Computing. NVIDIA’s focus is on its proprietary standard CUDA, which competes with OpenCL ― which is the open standards-based approach, again governed by the Khronos group.

Recently at GDC, we demonstrated an OpenCL version of the Havok engine.  And we expect to see more companies embracing OpenCL over time.

In the area of physics simulation, which is an example of Stream computing, NVIDIA links CUDA with its proprietary physics engine, PhysX, while AMD has chosen a different path in favor of open standards. OpenCL, which could also be described as a language and a set of APIs, ushers in a new era of computing by allowing applications to call on CPUs and GPUs in unified manner, resulting in the right processor accelerating the workload. And unlike the proprietary PhysX, OpenCL also allows many companies to develop and offer physics engines and other plug-ins of their own. 

In an industry that loves a war, the new one to take note of may be between CUDA with PhysX against OpenCL and an army of numerous engines & plug-ins from independent companies.  

In the end, the industry almost always comes to agreement on standards. In the PC industry we are currently driven by “de facto” standards, which generally come into use faster but can limit choice for users. When we all agree on “open” standards, on the other hand, we differentiate on a level playing field.

And history suggests that is really good for consumers.

The CUDA and OpenCL battle will be fought over the next few years, with applications and ― I suspect ― users as the battleground. In an ideal world we could all save time and money by agreeing on one or the other. Given that we believe in open standards, we vote for OpenCL.

Which do you vote for?

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.