|Page (3) of 3||
General Manager of Professional Graphics
What new technology do you hope to see at SIGGRAPH 2004?
We are seeing more new technology this year than in the years past, including PCI Express, 64-bit Intel/x86 based systems and the platforms to support all this technology. In addition, we are introducing SLI -- the technology to allow multiple graphics boards in one PCI Express system. Lastly, FX 4000 SDI -- for 10-bit uncompressed SD and HD graphics to video out will be shipping in volume.
Datasets are getting bigger and bigger as artists push the envelope of digital effects. How is NVIDIA working with the 3D software vendors to ensure that their applications are getting the most performance out of NVIDIA's hardware that they can?
Large model support is critical - certainly the availability of 64-bit systems to handle larger addressable memory is going to help. NVIDIA has the best 64-bit full featured graphics drivers in the industry, with full support for Win64 and Linux64 flavors. The key to large datasets is eliminating the bottlenecks in the system. PCI Express and large frame buffers are certainly going to help.
How is NVIDIA working with its customers to ensure optimum graphics output of its solutions?
NVIDIA works directly with the ISV and end user community to optimize their applications to take best advantage of the graphics horsepower resident. Some ways that we do this is through SDKs, Kitchens, texts (such as GPU GEMS and Cg Tutorial)-,as well as by having our DevTech team directly assisting with customer's code. In many cases, ISVs graft NVIDIA sample code directly into their applications.
Quadro FX 3400
Applications that are fill-bound and geometry bound (large models) will see a nice improvement in performance. The PCI Express architecture also enables much, much faster pixel readback performance. Broadcasters and other users of high-definition video will reap the benefits of this faster readback capability.
Do you see the market migrating to PCI Express or will the market still demand AGP and PCI-based solutions?
Systems builders are rapidly cutting over to PCI Express-based solutions. But we think that like most major industry transitions, it will take some time before the entire market migrates completely over. We do expect higher-end users to be the first to move over to PCI Express-based solutions. There will be a need to offer both, so NVIDIA is offering both AGP and PCI Express-based professional graphics solutions for at least the next 12-18 months.
Can you describe the notion of multiprocessing graphics scalability? And how do NVIDIA's products fit into this?
Multiprocessing is the ability to use multiple processors in one system. It's been around for a long time now in mainstream workstations with CPUs. With the emergence of a 16-lane bus architecture like PCI Express, NVIDIA is able to offer another multiprocessing alternative -- this time with graphics processors. On June 28, we introduced NVIDIA Scalable Link Interface or "SLI". This new technology enables multiple NVIDIA Quadro graphics boards to operate in a single PC or workstation for a stunning increase in graphics horsepower. NVIDIA SLI technology takes full advantage of the additional bandwidth and features of this new high-bandwidth bus architecture too. With SLI, users have the option of joining two graphics boards together immediately for a nice performance boost or buying one graphics board now, and adding a second later when it's more fiscally convenient. The user can decide.
Though professional users will want NVIDIA Quadro graphics hardware for their multi-GPU systems, SLI technology is also available to consumers with our GeForce 6 Series of consumer graphics boards. In addition to professional users, we expect gamers to be very enthusiast about multi-GPU systems. One important thing to mention is that SLI is not available for all models of NVIDIA Quadro and GeForce 6 Series graphics boards. You also need a PCI Express-compatible motherboard with multiple x16 mechanical slots to use it. You'll see systems based on NVIDIA SLI multi-GPU technology become available in the second half of 2004 from the worlds leading PC and workstation manufacturers as well as add-in card manufacturers.
What kind of graphics performance can users come to expect when running 3D applications such as Softimage|XSI or Maya?
The kind of graphics performance users can expect from running 3D applications depends a lot on the application in use. Some applications will see very big performance improvements.
In the perfect case for a very fill rate limited application, the performance is nearly linear --very close to 2x. We expect that this will also benefit applications in that higher quality FSAA modes will be usable at higher interactivety rates. Large model datasets will also see substantial improvement, however these are not likely to scale linearly. Again it is totally dependant on the applicaiton.
What does this mean for the animators who upgrade to the new graphics/hardware solutions?
An important point to remember is that PCI Express is a new architecture. Over time, the computer industry has learned to exploit AGP to its fullest. Over time, we expect the same to happen with PCI Express. There is a lot of headroom for that in PCI Express.
What does this mean for facilities that are pushing huge datasets via their renderfarms?
We are seeing more and more applications taking advantage of the raw floating point precision (fp32) and power (in pure GFOLPS) of the GPUs -- applications such as NVIDIA Gelato and mental images Mental Ray 3.3 take significant advantage of the GPUs to accelerate high quality rendering. Amazingly, we are seeing non-traditional applications also starting to use the power of GPUs across many markets, such as Financial Modeling, Bioinformatics, etc.
In addition to existing 32-bit platforms, system manufacturers are offering 64-bit-based systems using chips from Intel and AMD. There is the notion that 64-bit is more efficient, yet the market hasn't been demanding it yet, and on the operating system side, 64-bit Windows is still in beta. At SIGGRAPH 2005, do you expect to see all 64-bit solutions handling the modeling, texturing, and rendering operations of animation/effects shops or will we continue to see 32-bit systems churning through these datasets?
The OEMs are pushing 64-bit systems hard today and there certainly are segments of the market that need 64-bit capability now. I would fully expect that every workstation that is sold by SIGGRAPH 2005 will be 64-bit capable, and taking advantage of the 64-bit memory addressibility. In terms of applications, that may take longer.
PCI Express Cards Coming to Market
ATI has four PCI Express graphics cards in its workstation graphics family. The FireGL V3100 ($249) is the company's entry-level workstation graphics solution. This dual monitor capable card (DVI and VGA) is built around ATI's native PCI Express x16 lane architecture and features 128MB DDR graphics memory and uses four pixel pipelines and two geometry engines. It achieves 6.4GB/sec graphics memory bandwidth. It supports applications that are based on OpenGL or DirectX 9 on the Windows and Linux operating systems and comes with a three year warranty. The FireGL V3200 ($399) is also an entry-level workstation graphics solution. This dual DVI monitor capable card is built around ATI's native PCI Express x16 lane architecture and features 128MB DDR graphics memory and uses four pixel pipelines and two geometry engines. It offers 12.8GB/sec graphics memory bandwidth. It adds a Stereo 3D connector with quad buffered support. It supports applications that are based on OpenGL or DirectX 9 on the Windows and Linux operating systems and comes with a three year warranty.
NVIDIA sports three solutions in its PCI Express line of graphics cards. At the high end is the QuadroFX 3400. It features 256MB RAM, 28.8GB/second graphics memory bandwidth, 12-bit subpixel precision, 8 pixels per clock rendering engine, 3-pin stereo support, 128-bit color, 32-bit floating frame buffer, up to 16X full scren antialiasing and 16 textures per pixel. It supports OpenGL shading, OpenGL 1.5, DirectX 9.0, and NVIDIA's CG shading language, and can drive dual DVI displays at 1600x1200 or a single display at 3840x2400 resolution. It also supports dual link digital support for 9Mpixel displays. It outputs 117 million triangles per second and 4.2 billion Texels per second fill rate.
Related Sites: Creative Mac , Digital Producer , Hollywood Industry , Digital Animators , Animation Artist , Siggraph News , Digital Post Production , Film Imaging , Animation & FX , Oceania , CEN - Peripherals
Related Newsletter: DMN Newsletter , Timeline Newsletter , Pixels Newsletter , Renders , Digital Media Net , DMNForums