A Click on this Banner shows your site support to my Sponsors

Accelerate Your Mac! logo
The Source for Mac Performance News and Reviews

Preview: ATI's OEM Radeon 9700 Pro Graphics Card
By Mike
Published: 4/4/2003

(Update: For performance tests of the later 9800 Pro card, see the Video topics page)

Intro/Specs | 2D Performance | 3D Performance | Game Performance

3D Apps and Benchmark Performance

This page lists results of tests with non-game 3D applications and benchmarks like Lightwave 3D v7.5, ATI's Shader Demos, CineBench 2003 and an Apple developer Vertex Performance test program.

Lightwave 3D v7.5 Tests

I'm a novice at Lightwave 3D, but used a test suggested by a LW mac user group in the past. I loaded the "Hummer" scene (the longest and most involved of the samples I had), set the display options to 50,000 polygon bounding box threshold and timed how long it took to play a large (single window/max size) preview of the complete animation. (1600x1200/32bit monitor resolution.) Hopefully more software will take advantage of ATI's shader capability in the future to allow cards like the 9700 Pro to distance themselves from previous models more so than they do with typical current applications.

Considering a stopwatch was used to time this (I'd guess +/- 1/2 second margin of error for stop/start) the 9700 pro and GeForce4 Ti tests are practically a wash (less than 2% difference).

Geforce3 results are not shown since playback of this scene exhibited the same problem noted with the first Geforce2MX tests a couple of years ago (tracks disappearing during playback, reportedly due to the near/far scaling in this scene.)

ATI Shader Demo Tests

Since I have literally no applications or games that currently support shaders, I used ATI's Shader Demos (requires 8500 card or better) to record the lowest framerate seen during a 1 minute period to compare shader performance of the 3 ATI cards. The 9700 Pro clearly has much higher shader performance than previous models. (I just wish I had more software that utilized this feature. A test copy of Unreal Tournament 2003 should arrive next week for testing.)

The image quality in these demos is very impressive. The previous 9000 and 8500 cards trail far behind the 9700 Pro in these tests.

CineBench 2003 Tests

I used Maxon's Cinebench 2003 benchmark (available here) with the desktop set to 1600x1200, millions colors. I've included both the results scores as well as the detailed info on the two 3D scene fly-by tests.
Here's a graph of the overall Graphics scores - OpenGL hardware Lighting test scores are the ones that show the best card differences; the software scores are primarily CPU affected and therefore were similar for all cards in the same system.

Cinebench 2003 results

The 9700 Pro scored a bit below the Geforce4 Ti, although the difference in that score isn't really significant.

I also graphed the detailed results from CineBench regarding the Hardware OpenGL framerates for the Scene1 and Scene2 fly-bys.

Cinebench Scene FPS rates

The fact that the 8500 and 9000 did better in Scene2 makes me think the 9700s initial drivers are the reason for its lower test results in that scene. (I can't think of another reason to explain why the 9700 Pro would score lower than the 8500/9000 cards.)

Vertex Performance Benchmark

Vertex Perf. Screenshot

I used version 1.0 of the Vertex Performance Test at Apple's developer site to measure triangle throughput. I want to stress (especially since the GF4Ti and GF3 rates were so low in comparison) that the Nvidia cards were tested using the standard drivers in 10.2.4. The ATI driver update used for the ATI cards (build dated Apr.3, 2003) included revised OpenGL files. I was tempted to remove the Nvidia results.

The graph shows results with 2 to 8 textures (2T-8T). Only the 9700 Pro allowed tests with more than 6 textures, therefore many of the cards have no results possible for higher numbers of textures. In some modes the rates vary a bit but I tried to record the most consistent observed rate.

The highest performance mode for all cards was the "Vertex Array Range" mode. Normals were enabled for this test.

Vertex Array Range Normals

Chris, an ATI programmer, had told me that Doom III uses CVA mode, no normals (I guess as with previous Quake versions, it will have it's own lighting code, rather than using the card's hardware lighting) so I used those settings to compare the cards. I'm going to repeat this test again to double check this mode's results, but since with the drivers I have the 9700 Pro was actually slower in this CVA mode (no normals) than the 9000 and 8500 cards with 5 or less textures enabled. (Chris told me that the current drivers were not optimized yet for slower paths like this, but I still would like to swap the cards in and repeat it to double check the results.)

Vertex CVA mode, no Normals

For a table of GPU specs comparing the 9700 Pro to the 9000 and 8500 cards, see the Intro page. The next page compares game performance.

Index of OEM Radeon 9700 Pro Review

Intro/Specs | 2D Performance | 3D Performance | Game Performance

- or -
Back to www.XLR8YOURMAC.com

Copyright © 2003.

No part of this sites content or images are to be reproduced or distributed in any form without written permission.
All brand or product names mentioned here are properties of their respective companies.