Technology Highlights from SIGGRAPH Asia 2008

Introduction and Gallery Highlights

For the very first time, the world renowned SIGGRAPH computer graphics conference is being organized for the Asian market and December 10th marked the kickoff for this momentous occasion. Held at the Suntec City Convention Center in Singapore, it was a logical choice that the very first SIGGRAPH Asia conference was hosted in this garden city where a number of game developers and animation/video production houses have recently been basing their Asian operations from Singapore. Some of the big names include Ubisoft, Koei and LucasArts, but even more developers are expected to set up studios here with the country's Media Development Authority (MDA) and Economic Development Board actively enticing even more developers to do so.

SIGGRAPH stands for Special Interest Group on GRAPHics and Interactive Techniques and this computer graphics conference attracts creative professionals from production houses such as graphics artists, animators, illustrators and even students from this region to come together and witness the latest trends, cutting edge of graphics and visualization technologies, illustration techniques, animation concepts, research, attend specialized talks, focused workshops and even view a gallery of interesting works of art to draw inspiration and to free one's mind in this ever-evolving trade. Previously only held in USA, the Asian extension of SIGGRAPH is a boon to the growing development in this region and to showcase and interact with matters closer to home.

First off, let's take a quick look at the Emerging Technologies corner for some cool stuff:-

The Big Supporters and Facilitators of Computer Graphics

ATI and NVIDIA - Turbo Charging the Graphics & Animation Industry

So now that you know what SIGGRAPH is about, how does this matter to our computer technology focused www.hardwarezone.com? Well the two prominent graphics visualization players are key to enabling many cutting edge technologies as well providing the means to crunch and process complex visual data, so it's only natural that AMD (ATI) and NVIDIA had strong presence to demo and promote what their solutions are capable. In the weeks building up to this conference and exhibition, both companies were busy promoting their latest developments. So we've got quite a good picture of what their goals are with their latest hardware and software solutions.

If we had to sum up what both ATI and NVIDIA have in store as their next big thing, it has got to be stream processing. Stream processing is using the hundreds of specialized shader processors available on each of the GPUs to vastly accelerate the processing of certain general-purpose floating-point workloads which were once only possible to execute on the general purpose CPU that has far fewer cores.

Also known as GP-GPU computing (general purpose GPU), thanks to high-level shader processing interfaces/languages like NVIDIA's CUDA and ATI's Stream SDK along with their Brook+ compiler, this is all becoming a reality. However both of the above approaches are vendor-specific platforms, but the good thing is there's now the recently ratified OpenCL standard that both ATI and NVIDIA support and there's also DirectX 11 next year that will also facilitate these GP-GPU computing initiatives and really open the floodgate to all developers to reach out to the full audience. Meanwhile ATI and NVIDIA will still pursue their proprietary standards for some time to come since the common standards have barely emerged and each have been independently working with several software OEMs to spruce up their processing capability with steam processing.

To give you an example of the advantages to what extent stream processing will help the masses, take for example the latest Adobe Premiere Pro CS4 that's enhanced by the NVIDIA CUDA enabled plug-in which comes in conjunction with NVIDIA's latest Quadro CX 4800 graphics card designed for the creative professionals whose work heavily involves Adobe Premier Pro CS4. If you were to encode a 2-hour high resolution project into a H.264 file with a 1080p resolution, what would have taken a standard built-in encoder more than a day on a dual-core system, would have only taken 7 hours to complete on the very same system when using a CUDA enabled plug-in in conjunction with the Quadro CX 4800 professional graphics card. Convert the time savings into cost, and you get the big picture.

Not only does it complete the job a lot faster, the system CPU utilization is also far lower and you can actually get more productivity done while the encoding is taking place. Our own trials from a demo system on the show floor using a sample file saw the CPU usage levels drop from 100% to just 50%, while getting the job done a lot faster (just 30 seconds versus the 1min 40 seconds without the CUDA enabled plug-in - that's more than three times the improvement).

Likewise stream processing and GP-GPU computing in general are picking up in the consumer segment as well with fast transcoding, smoother decoding and much more with video-centric applications from Cyberlink and Arcsoft to just name a few. This is very much an infant segment at the moment, but the future is bright as developers are opening up to the GP-GPU computing idea with strong push from both ATI and NVIDIA to educate them on the vast untapped potential.

We'll round up the coverage of SIGGRAPH with some focused highlights from the NVIDI A and ATI booths on the following pages.

NVIDIA and Leadtek Showcase Quadro and Tesla Series

Over at NVIDIA's booth, there was quite a bit of emphasis on what their latest Quadro FX and Quadro Plex are capable of achieving. To prove it, they had important demos running on the Quadro graphics bases systems and showcased applications like 3ds Max, Adobe CS4 suite and even a new shader creation and management program from mental mill. Additionally, Leadtek partnered NVIDIA to showcase its complete professional series solutions.

The screen on the left of the image was where Adobe was showcasing its new Adobe CS4 suite and how its new software greatly benefits from NVIDIA's latest Quadro FX graphics cards. On-site was Pete Brownstein, Adobe System's

Business Development Manager for the Asia-Pacific region, who guided us through three of the core applications that had the biggest advantage with the Quadro graphics card series:- Adobe After Effects CS4, Adobe Photoshop CS4 and Adobe Premiere Pro CS4.

The huge frame buffer size of the Quadro graphics cards and their OpenGL acceleration are key reasons why they help accelerate these highly used professional creative applications. As Pete showed us the performance as well as capabilities with and without the OpenGL acceleration feature enabled, there was quite a profound difference. We can clearly see that for professionals in this line of work, this advantage is certainly very much required for much better productivity. There is certainly no doubt about that. While OpenGL acceleration is also available on consumer-level desktop graphics cards, they don't have large frame buffers and neither have they been professionally certified, so these aren't solid solutions for serious work.

For creative professionals dabbling with Adobe Premiere Pro CS4 in particular, NVIDIA has recently launched a special graphics accelerator, the Quadro CX. This card is equivalent to the Quadro FX 4800 which was also recently launched, based on the original GeForce GTX 260 GPU. However, the CX comes exclusively bundled with the CUDA accelerated RapiHD plug-in developed by Elemental Technologies to greatly turbo charge the render time required and vastly cuts down on the CPU processing power required as we've highlighted earlier. This plug-in is not available by any other means besides purchasing a Quadro CX.

While both the FX and CX variants have an equivalent suggested retail price of US$1,999, certainly we would recommend the CX since it can do all of what the FX can perform. This weird pricing strategy was of course a move to promote the new Quadro CX, however as we spoke to NVIDIA, they expect the retail/partner solutions to mark the Quadro series pricing up or down by demand and interest. This would likely cause the CX to cost more than the SRP or the FX to be priced lower than the CX in the long run.

For those who are more focused on other kind of illustration, animation and design applications like AutoCAD, Maya, 3DS Max and several others, the Quadro FX series has several models to cater to the level of usage and requirements. Additionally, the Quadro series come with specialized drivers for these applications to provide performance boosts. The top dog here is the recently launched Quadro FX 5800 that's based on the GTX 280 GPU and has a whopping 4GB frame buffer.

To get a better idea of who exactly requires that large a frame buffer size, we spoke more to Tyler Worden, Market Development Manager of NVIDIA who's in charge of the professional visualization solutions. From our discussion with him, it seems that while 4GB sounds a whole lot, the medical, engineering and especially the oil and gas industries can eat volumes of processing power and memory requirements in a jiffy. He explained the data which these industries handle to churn out their models for visualization is so immense that even if the Quadro card had 16GB of frame buffer, that wouldn't satiate their requirements.

For example in the oil and gas exploration arena, seismic charges are set off and numerous sensors are placed across a huge area to map the terrain to form an underground 3D map of sort. No longer is data just analyzed on specific areas, but these days a much larger area is mapped to get a clearer view of the macro scale of things and better determine the aspects of the entire terrain. This could be several kilometers in every direction and correspondingly requires a terrific amount of data manipulation and crunching power to generate the required analysis models. The huge frame buffer of the latest generation of Quadro FX cards help on two grounds:- one is on the CUDA-enhanced specialized programs used to tap on the GPU's crunching power and thus a huge streams of data are fed thru the GPU for accelerated processing. The other is on the actual handling of the visual model which is extremely huge and requires a boatload of memory to handle it fluently.

Over on the high-performance computing (HPC) segment, NVIDIA's targeting the Tesla series of stream processing cards, which the latest is also based on the GTX 200 series of GPUs such as this NVIDIA Tesla C1060 stream processing card shown below. No display outputs are featured, so it's likely a case of the removal of the NVIO chip, but then again, that's not required for a stream processing card.

Last but not least was a demo from mental mill that's related to the creation and managing shaders that can be re-used to create more complex shaders and can be used in several commercial design applications like mental ray, Catia, 3DS Max, Maya and Softimage XSI. Best of all, shader artists don't really require programming know-how to create, manipulate and incorporate the shaders as mental mill has a very graphical, intuitive and flexible GUI interface with real-time preview. Artists with programming skills can also directly write shaders within the program for further fine tuning, but with the real-time preview and debugging, the task at hand is far easier to manipulate. Here's how the node-based GUI looks like with each node having a list of attributes and numerous controllable parameters:-

ATI and Sapphire Showcase Professional Visualization Solutions

Over on ATI's corner, one of the gold supporters of the SIGGRAPH Asia 2008 event, they had their own professional solutions on display. So here are some cool stuff we found on their booth:-

Just when Full HD is getting into gear of mainstream, the industry is already pushing that to Quad Full HD - a resolution of 3840x2160 pixels that's capable of displaying close to 9 megapixels in 1:1 ratio. More resolution equates to getting more work done as you can squeeze more out of the same display area. To fulfill the high requirement needs, Sapphire, one of ATI's most loyal partners that also handle professional solutions offers its clients a 56-inch Quad Full HD LCD solution that comes bundled with an ATI FireGL V8650 professional graphics card that uses its dual-link DVI outputs to power the display. While not the newest, it is the only card from ATI with a 2GB frame buffer for processing the huge workload sizes that comes with such high resolution requirements.

ATI is also very focused on enabling stream processing for the masses and calls its technology ATI Stream. Its full potential is released when paired with the Radeon HD 4000 consumer series or with the AMD FireStream 9000 professional series as they utilize the latest GPU series. The ATI Stream SDK in combination with their Brook+ compiler enables easy high-level programming to unleash the GPU's full processing prowess for GP-GPU processing. Like NVIDIA, ATI too supports the OpenCL API and is probably an even stronger supporter of it. OpenCL is still in its early days and while there hasn't yet been a demo or program to showcase an OpenCL application running off the GPU, AMD/ATI is the first to showcase a demo running off the CPU and was displayed at their SIGGRAPH Asia booth:-

However ATI Stream isn't just for the professional market; it's just as apt in the consumer realm as well. ATI is working with several video player/creation application vendors to help accelerate video encoding, transcoding, up-scaling and more. On demo at the show floor was ArcSoft TotalMedia Theater with SimHD DVD up-scaling technology (up to 1080p) that's enabled by ATI Stream technology:-

Well, that's it for the SIGGRAPH Asia 2008 and hopefully we'll be there in the 2009 conference to catch next year's updates in this field in Yokohama, Japan. See you then!

Our articles may contain affiliate links. If you buy through these links, we may earn a small commission.