On the day prior to Computex NVIDIA held a press-conference at the NYNY, a venue close to the TWTC exhibition grounds. The presentation was led by NVIDIA's CEO Jen-Hsun Huang.

World of GPU computing

Mr. Jen-Hsun gave two conferences during Computex, the first presentation I had time for, and for you guys was of course the most interesting. With the graphics industry on a slope right now we really wanted to see what Jen-Hsun had to evangelize here, and perhaps talk a little about the next generation products.

NVIDIA's CEO talked about new ways GPU Computing is changing the way we work and play. As you can guess already, NVIDIA talked here about CUDA mostly. If you don't know what CUDA is, please read up on some of our latest GeForce reviews where we explain it in detail. CUDA simply put is a programming interface to utilize the GPU for other things than games. We have seen a good number of applications already, but some more prominent ones allow us to transcode videostreams over the GPU with a nice performance increase. Parallel processing ladies and gentlemen...  some applications are just extremely suited for it.

Jen-Hsun was particularly excited about the pending release of Windows 7, Windows 7 has DX compute embedded in its core and as such the GPU can now completely be opened up to the operating system, assisting applications that would like some parralel tough love. It's very abstract right now though so let's fall back to a more simple example of what you can achieve with CUDA.

Another nice example is the recent release of CoreAVC. CoreAVC is a video codec allowing playback for, for example, MKV (Matroska) x.264 encoded content. CoreAVC recently became CUDA ready and as such the x.264 content can now be completely accelerated and, to a certain extent, enhanced over the GPU, no longer the CPU. The advantages here are simple, you leave more processor resources available to your PC, the GPU can do the job with significantly less energy than a CPU and, often... does a better job doing so.

But also, let's not forget about PhysX, PhysX is actually at this stage CUDA driven. And you can argue about the additional thrills it brings to gaming, my point of view remains simple; it is an extra feature that allows for some pretty cool effects. With a CUDA ready graphics card, that only costs ~10%  performance of the GPU + it has to render some extra objects... but still. It's a technology that i like very much and even after 2 years, PhysX is still in the early stages of being adopted. Five years from now we'll take it for granted as much as we now take 4xAA completely for granted. Whether or not PhysX remains a CUDA only feature can be doubted. Lately a lot of rumors are hinting at OpenCL PhysX, making things very interesting. Also of course keep an eye out for the physics GPU implementation of Havok.

So CUDA is something that NVIDIA keeps emphasizing no-matter where you are and where you go. I do hope though that NVIDIA realizes that people buy their graphics products to play games... everything else is just a nice little extra.

The presentation was dominant on the compute side of things and the upcoming ION products. There was very little to no mention of DirectX 11 compatible graphics adapters, GT300 and future architecture. As always, Jen-Hsun is a visionary and did talk a little more about NVIDIA's strategy and made sure examples were shown.

Interesting was a presentation of NVIDIA's Drew Henry, obviously very proud to deliver ION as promised, on time.