MacTell Vision 3D Pro II Video Card Review (1998) 3D Tests Results Intro | 2D Performance | 3D Performance | Game Performance | Movie Playback | Features | Summary 3D Performance
Evaluation of 3D performance was based on using the card in several real world 3D applications and a series of benchmarks to provide more accurate comparisons of performance than my "seat of the pants" judgements from using standard programs.
I tested the card with both Infini-D 4.01 and Ray Dream Studio 5.02. Infini-D 4.01 tests were run with the camera view set to "best interactive" or "HW interactive" rendering engine to allow shaded objects to be moved in real time (using models from the Ch. 7 tutorial file). It seemed to me the "HW Interactive" was the faster and better quality option, but both worked fine. Infini-D's camera view seemed to be slightly smoother with the Vision 3D Pro II with the 180mhz 604E CPU than any other card I've tested, and the lighting seemed more accurate as well. I noticed slightly less hesitation in moving the rendered objects than I'd seen with the other cards I've tested.
With PowerForce G3 or XLR8 G3 CPU cards installed in the PowerTower Pro vastly improved performance and responsiveness was seen (as expected). However since I'd tested the other cards with the PTP180 stock CPU (180mhz 604E), I used that CPU card for all the comparison tests.
At 1024x768, thousands colors, the 3D performance in all the applications I tested was virtually identical to the Nexus GA card, maybe slightly better, based on observations and "feel" in normal use. For those looking to get the best 3D application performance, make sure you have the fastest CPU you can afford and lots of ram in addition to a good video card.
As I've said in past video card reviews, I'm not alone in being frustrated at the 3D hardware support in Mac applications. Often times it's a combination of too little memory on the video card (3D buffering), a driver/hardware bug, lack of full support for Quickdraw 3D in the application and/or a limitation/bug in Quickdraw 3D itself. If you check reviews of most popular 3D packages you'll see many have a track record of bugs and limited 3D hardware support. This is improving, as is the speed and quality of QuickDraw 3D itself. BTW - the Ray Dream 5.02 update seemed far more stable than the initial release.
Although I've yet to see any huge differences between the 3D hardware accelerated video cards (with proper drivers) in 3D applications, I do have an appreciation for the texture mapping they provide when used with a fast CPU card. That combination really makes a big difference in productivity, just don't expect a video card alone to make a huge improvement.
I've had discussions twice with NewTek on obtaining a evaluation copy of Lightwave 5.5 for a complete performance review of all the cards I currently have on hand, but once again they showed some interest and then never followed up. They seem to send mail, ask questions, sound interested, then I don't hear from them for months. I'll certainly have to return the Vision 3D Pro II card long before I ever see a copy of Lightwave (if I ever do).
I used Village Tronic's RaveBench 1.1.1 as both a benchmark and visual features check of the card. It runs several tests of texture mapping, transparency, movement, and environment mapping functions. Happily the Vision 3D Pro II supported all the features well, and I noted the latest driver (5.1.2) provided much better transparency image quality. I compared the results with the top two performers of this test - the ATI Nexus GA and Village Tronic MacPicasso 3D Overdrive (3Dfx).
For the RaveBench tests, screen resolution was set to 1024x768, thousands colors and the test window was 640x480 pixels. The Nexus GA was set to best 3D quality in the ATI control panel.
Since RaveBench allows only two scores to be compared and due to the lack of Rave support in thousands color mode on the Ultimate Rez card, only the 3D0 and Nexus GA comparisons are shown. The following chart lists the results of the tests in frames per second for each card.
In tests other than the Mesh function, the Vision 3D Pro II was slightly faster than the Nexus GA, but I was disappointed in the results, since the Nexus GA is a 64-bit card and the Vision 3D Pro II is a fully 128-Bit card (twice the available bandwidth). This indicates a high level of optimization of hardware and software on the Nexus GA along with possibly a much less optimized driver for the Vision 3D Pro II.
The Vision 3D Pro II was consistently faster than the 3D Overdrive, not bad considering the reputation of the 3Dfx Voodoo chipset. I think the VillageTronic viewport scheme must extract some performance penalty under Rave, or else it's a lack of fully optimized driver.
The results of the Ultimate Rez are not shown, as the current driver release does not support Rave acceleration in thousands color mode. I did run the tests in both software mode at thousands colors and in hardware assisted mode at millions colors. In both cases the Ultimate Rez was 3 to 10 times slower at each test than either card and did not support some features such as transparency.
Walker 1.1 Tests:
I ran three scenes in Lightwork's Walker 3D viewer with the 2D screen set to 1024x768, thousands colors (Walker used a 350x350 default window size). I was very shocked at the low Walker scores, - to the point where I reran the tests several times to ensure I didn't do something wrong. Since the more complex RaveBench tests showed much better performance, I didn't put a lot of weight on the simple Walker spin tests after using the card in Infini-D. Again it looked like the lighting and shading in the scenes was more realistic with the Vision 3D Pro II, but I'd have to take screenshots of each to compare them (I plan to).
I list the lowest framerate seen in the rotation of the scene, as that indicates how the card handles the toughest part of the scene. I saw very high framerates on the Vision 3D Pro II on the less detailed parts of the scenes (as high as 180fps, but these sections were usually just simple walls, or blank areas).
The minimum framerates seen in two 360 degree spins for each of the three Walker scenes is shown below:
- Vision 3D Pro II: 4.29 fps (driver 5.1.2)
- 3D Overdrive: 6.00 fps
- Nexus GA: 8.00 fps
- Vision 3D Pro II: 5.81 fps (driver 5.1.2)
- 3D Overdrive: 7.06 fps
- Nexus GA: 9.38 fps
- Vision 3D Pro II: 9.09 fps (driver 5.1.2)
- 3D Overdrive: 9.68 fps
- Nexus GA: 12.73 fps
In summary, 3D performance of the Vision 3D Pro II was good (except for the Walker tests), but with the current drivers in normal use it seemed no better than lower priced cards such as the Nexus GA. If I had more time I'd like to retest them all more extensively with the G3 CPU card and more detailed scenes to try and see if that would be a better discriminator between the cards. I must point out however, that the Vision 3D Pro II is the only card that combines best of class 2D speed with good 3D support at this time. I sincerely hope that future driver releases will provide even more speed in 3D for the Vision 3D Pro II.
I put more weight in real world apps and the RaveBench test rather than the simple spin test of Walker. On a scale of 1 to 10, I rated 3D speed a 8.
UPDATE: I just finished checking the latest beta sent by MacTell (GA-220.127.116.11) and it performed identically to the Formac released GA-5.1.2. In fact Get Info and Extension Manager both show it is version 5.1.2, and the modified date was one day before Formac's 5.1.2 version. Performance in RaveBench was identical to 5.1.2 on all tests but was 1% slower on the Mesh test, which is within the run-to-run variation. MacBench tests showed the same results as 5.1.2 also.
Although some may ask why I bother with games on a card like this but I know the majority of readers would want to know (and I don't need 500 mails asking why I didn't included it ;-). The next page shows results I had running several popular games.
Index of Vision 3D Pro II Review Pages
Copyright © 1998. All rights reserved.
Brand or product names mentioned here are properties of their respective companies.