Accelerate Your Mac!
ATI Nexus GA and IXMicro Ultimate Rez FaceOff
Review date: 01/11/1998
3D Graphics Performance Testing

As is true of most graphic applications, your processor type and speed, amount of ram and cache are more of a performance factor than your video card. For 3D applications your CPU plays the largest role in performance, as most of your time will be spent rendering, the duration of which is primarily determined by your CPU type and speed. Video cards do not speed up rendering, only the displaying of objects.

A good 3D card can provide faster manipulation of objects in shaded (vs. wireframe) mode. Working in shaded mode provides better visualization of your models and scene, resulting in increased productivity. A crucial factor in all this is how well your 3D application supports your video card and/or driver. A great card cannot make up for buggy or incomplete 3D support in the application. Many applications may not be able to take advantage of some or all of the 3D features these cards provide. Hopefully both drivers and application support will improve in the future, and 3D hardware accelerators are one of the hottest trends in personal computing today.

3D hardware is one of the few areas where the Mac trails the Intel based systems, where 3D accelerator hardware is abundant, fast and cheap. New models appear almost weekly, and the 2D and 3D performance of some of the latest under $200 cards is pretty amazing. Hopefully some of the better models will make their way over to the Mac soon. I have a keen interest in 3D hardware, and absorb every technical article, magazine review and industry newsletter dealing with the subject.

Until a benchmark like 3D Winbench comes along for the Mac (listening Ziff Davis?), we have to make do with much simpler tools. In fact many of these were never meant to be benchmarks at all. Lightwork's Walker was created to be a 3DMF viewer for instance, but has turned into something of a standard bencmark now.

Walker 1.1 Tests:

Lightwork's Walker test consists of several high polygon count 3d scenes that display in a 350x350 window. As shown in the chart below, three Walker scenes were run, with the minimum frame rates recorded for each card. Two 360 degree spins where performed, and the lowest fps displayed was recorded. The chart lists the scores for each scene. The minimum frames per second is recorded as that is the best indication of how well the card handles the most demanding part of the scene.

Walker v1.1 3D tests: 1024x768, 16-bit color, 75hz vertical refresh
Walker 1.1 scores


Note that according to the Ultimate Rez Read-Me file, 3D hardware acceleration is only active in millions color mode, and using that color depth did slightly improve the 3D scores (appx 1 fps improvement). I feel confident future driver releases from IXMicro will improve 3D performance and support over the initial release.

As soon as I load Ray Dream Studio 5, I'll see how the cards compare in that application, which is supposed to take advantage of QuickDraw 3D accelerators like these cards.

RAVE Quake:

What better way to test RAVE (Quickdraw 3D) game performance than with Quake? Unfortunately only the Nexus GA was able to run it, as the Ultimate Rez initial driver relase only supports 3D hardware acceleration in millions color mode, and RAVE Quake will not run with the display set to that color depth. Reducing color depth to thousands results in Quake failing to detect a RAVE video card. The same error occurred on the Imagine 128, with and without its 3D hardware acceleration mode enabled.

For now, if RAVE Quake is important to you the Nexus GA is the better choice. The 180mhz 604E PowerTower Pro with the Nexus GA installed ran the RAVE Quake Timedemo demo1 test as high as 21.6fps (640x480, flames on). A clear win for the Nexus GA here.

RAVE Quake (v1.08.2) 640x480 Timedemo demo 1 scores:

  • Nexus GA: 21.6 fps
  • Ultimate Rez: N/A (would not run)

I found it interesting that changing the Nexus GA's 3D quality mode to standard instead of the defualt high quality actually reduced the framerates a bit, as did changing for the default medium texture compression to no compression (framerates dropped to 20fps). The standard quality did make the floor texture in Quake appear properly, whereas the high quality made the texture appear stretched/blurred excessively. ATI's excellent manual noted that some games may look and perform better at the standard 3D quality setting.

The Nexus GA clearly wins in every 3D test I've done so for, but it remains to be seen if that is due to its more mature driver or a superior 3D engine. Future revisions of the Ultimate Rez drivers will hopefully enable RAVE game performance and 3D hardware acceleration in thousands color mode. This will allow us to really see what the Ultimate Rez's 3D hardware can do. [Update: The 3D/gaming driver situation for the Ultimate Rez (or the later released ProRez and Mac Rocket cards based on the same chip) never improved. IXmicro went out of business in 1999, and the IX3D series of cards never had a driver update to make them a good 3d/gaming card.]

3D Tests Winner: Nexus GA

On to Software/Docs page =>>


Copyright © 1998, all rights reserved.
All brand or product names mentioned here are properties of their respective companies.
Site terms and conditions of use.