• Join Us |
  • |
  • Sign in with:

Latest News

Next-gen of 3D acceleration

by Marcus Yam, Jul 06, 1999 12:01am PDT
Related Topics – Hardware (PC only)

Hey it's the return of Kenn (FiringSquad), discussing next generation 3D accelerators from 3dfx, Nvidia, and others. They discuss the implications of T&L, massive-multitexturing, multiple processors, and more.

One big effect of geometry acceleration is the difference (or lack of difference) on lower-end PCs. Potentially, if you've got a P2-266 and a P3-500 side by side, the graphics portion of games such as Quake 2 and Unreal will run at near exactly the same speeds on each of the systems. Unreal might lag slightly due to more sophisticated AI, but you'll no longer need to worry about upgrading the CPU to get more FPS. Yep, this is probably Intel's nightmare as the CPU becomes a non-factor in the specifics of 3D performance




Comments

10 Threads | 10 Comments






  • This is in response to #1. I recently bought a voodoo3 chip for my 300mhz machine, and I recommended one for a friend who has a 233mhz machine.
    I can run Aliens vs. Predator, halflife, Q3a, all of them at 1200x1000 resolution at 30 - 40 FPS! 10 - 30 for Q3a. I don\'t think the CPU matters at all when just the 3D chip can get this much performance. These computers are really old!
    In my opinion, 3D chips will become even less dependent on the CPU then they are now.
    If Intel forces everybody to use some crappy on-board 3D acceleration (and trust me it will be crappy!) it will be the end of all the wonderful advances in 3D rendering that games are making.



  • This is stupid. For a site like the FiringSquad to make a statement like that really shows you how mainstream derivative thinking can lead to trouble. The faster the 3D accelerator, the faster the CPU it requires. As accelerators get even faster, the CPU MHz has to be commensurate. The fact is, a large percentage of the geometry calculations are still performed by the CPU.

    In fact, I predict that Intel will eventually buy out the dominant 3D chip manufacturer and integrate their chip into their \"card-chip\" that they have now. When they do, it would only take a CPU upgrade to significantly upgrade the core MHz and the 3D accelerator at the same time-- in time, it would resemble the math coprocessor of the past x86\'s.