Computer Games and Viz: If You Can't Beat Them, Join Them

by Forrester Cole
15 August 2001

   Apparently there are a lot of bitter people in the scientific visualization community these days. A good number of them showed up to the panel on computer games and visualization today. The topic of the panel was the relationship between the computer gaming and visualization industries, as gaming has taken over the leadership role in graphics features and hardware. The panel was organized by Theresa-Marie Rhyne of North Carolina State University, and included Peter K. Doenges of Evans & Sutherland, Chris Hecker of definition six, inc., William Hibbard of the University of Wisconsin-Madison, Hanspeter Pfister of Mitsubishi Electric Research Laboratory, and Nate Robins of Avalanche Software.
   While games have provoked higher and higher performance in PC and console graphics hardware, the high-end visualization market has been left behind. Chips designed for gaming often do not have the features that visualization requires, such as high precision framebuffers or floating point math throughout. Gaming oriented chips also tend to have a large number of unintentional "features" that visualization would rather do without. This situation leaves the visualization industry with a dilemma: sacrifice features to use the cheap, fast, gaming oriented processors, or stick to the more feature-rich, but slower and far more expensive SGI and similar machines. Doenges and Hibbard gave good overviews of the problem from the visualization side. Doenges advocated that the visualization industry become more active in getting its needs heard by the hardware companies such as ATI and NVIDIA. Hibbard acknowledged that visualization and gaming do have different needs, but claimed that given a few years the hardware manufacturers will integrate the high-end visualization features anyway.
   Hanspeter Pfister painted a much less rosy picture. After acknowledging the obvious good points about the PC graphics chips, he dived straight into their deficiencies. Not only are the graphics chips buggy and optimized for Quake, not visualization, in their drive to one-up the competition the hardware vendors are polluting OpenGL and DirectX with specific, unportable extensions. All this is happening because the gaming industry wants more frames per second. The visualization industry is paying the price for the gaming industry's short-sightedness.
   The game programmers were not going to take that in silence. Hecker put the blame for bugs and API hacks on the hardware manufacturers. He claimed that the hardware manufacturers do not listen to the game makers either, and have such long lead times that they could not integrate advice into their next chip even if they were inclined to. Game programmers also want to see good, general APIs. The solution to the argument seemed to be to offload all the blame for the current mess onto NVIDIA and ATI, with a side helping for SGI.
   A long question and answer session followed the presentations. Some discussion centered on the observed historical cycle from hardware to software rendering and back again, as the needs and complexity of graphics grew. An audience member, supported by Pfister, suggested that all the problems may be solved by a future fast CPU that would make a GPU unnecessary. Hecker and Doenges countered that even if Moore's Law continued for the next twenty years, CPUs would still not be fast enough to generally solve all rendering problems, leaving room for GPUs in the foreseeable future.

The Official SIGGRAPH 2001 Computer Games and Viz page.