- Date:
- Monday , August 19, 2002
- Author:
- Kyle Bennett
- Google +1

ATi Radeon 9700 Pro Benchmarks
We show what the 9700 Pro is capable of on a Pentium 4 2.53GHz, an AMD 2.5GHz, an AMD 1.5GHz and throw in the Ti4600 for competition and the Parhelia just for fun.
Test Systems:
P4 2.53GHz - Gigabyte Titan533, 4 x 128MB Kingston PC1066 RDRam, Maxtor 40GB ATA133 HD, Windows XP Professional; Intel Chipset Driver version 4.00.1009; ATi Catalyst Driver v02.2; NVIDIA Detonator Driver v30.82.
AMD 2.48GHz (15*165) - EPoX 8K3A+, 1 x 512MB Corsair PC3200 DDR Ram, Maxtor 40GB ATA133 HD, Windows XP Professional; VIA 4n1 Driver v4.42; ATi Catalyst Driver v02.2; NVIDIA Detonator Driver v30.82. This is an overclocked system running with an overclocked AGP bus at 82MHz and PCI bus at 41MHz.
AMD 1.533GHz - EPoX 8K3A+, 1 x 512MB Corsair PC3200 DDR Ram, Maxtor 40GB ATA133 HD, Windows XP Professional; VIA 4n1 Driver v4.42; ATi Catalyst Driver v02.2; NVIDIA Detonator Driver v30.82.
All GeForce 4 Ti4600 and Matrox Parhelia benchmarks were run on the Pentium 4 2.53GHz system. The 9700 Pro system is noted on the graphs. AA On - AF On denotes the following: the 9700 Pro at 4XAA and 16XAF, the Ti4600 at 4XAA and 8XAF (no AF in D3D though as it is not supported in the current drivers), the Parhelia at 16XAA and AF set to on in the driver panel. We always used the driver control panels to turn features on and off as we see it a being a bit more controllable when forcing the settings through the drivers.
Applications used for Benchmarking include Quake 3 (OpenGL), MadOnion.com's 3DMark (Direct 3D), Serious Sam 2 (OpenGL), Jedi Knight 2 (OpenGL), Comanche 4 (Direct 3D), Code Creatures (Direct 3D). In Quake 3 we used all settings placed at their maximum setting. Serious Sam 2 was run with Beyond3D's benchmark script turning on maximum visual settings and maximum AF. Jedi Knight 2 was run with maxium visual settings as well. Basically we turned on everything we could in order to enhance the visual features of the game.
Also we did not run benchmarks below 1024x768 resolution as there is simply no need to purchase this card if you are going to be running low resolution settings in our opinion as it defeats the purpose of what the card is built to do; and that is deliver great visuals and along with great frame rates.
9700 Pro Vs. 9700 Pro Benchmarks:
We thought we should start out pitting the Radeon 9700 Pro against itself. These benchmarks will show you specifically how the card scales when you start turning on Antialiasing and Anisotropic Filtering (AA and AF) while going for mid to high resolution settings. All of the follow 9700 Vs. 9700 benchmarks were performed on the Pentium 4 at 2.53GHz.
Quake 3:

While AF is not exactly free, you can see from the first two sets of bars that it is certainly not expensive. Amazingly, we can run Quake 3 at 1600x1200 with every bit of eye candy turned on exploiting 4XAA and 16XAF and still have a single player or deathmatch game that most players are going to be very happy with. Continuous frame rates in the 80s and 90s are what can be expected at that level of quality.
MadOnion.com's 3DMark:

The Radeon 9700 Pro turns in the highest stock 3DMark score that we had ever seen...until we ran it on our AMD CPU at 2.5GHz (next page). Under this synthetic benchmark we also show AF at 16X to be a bit more intensive to run than previously suggested by Q3.
Comanche4:

Comanche shows that at pretty much any resolution with no AA or AF, we are CPU limited. When the 4XAA is turned on, we will don't see a big hit. However the 4XAA and 16XAA give us about a 40% decrease in score at 1280x1024.
Code Creatures:

While it is hard to state the validity of running this benchmark on the Radeon 9700 Pro because this is another synthetic benchmark designed to run on NVIDIA hardware it does certainly show some possible limitations that can be presented by current DX8 technologies. It should be kept in mind that this is benchmark that was meant to stress the GeForce4 when it came out earlier this year.
