The Top-End Mac Graphics Cards, ATI Radeon X800 XT vs Nvidia GeForce 6800 Ultra--How Fast Is It? [Updated]

We have been waiting long time to pit production units of the two top graphics cards for the Macintosh against one another. Last October, we received the Apple-made Nvidia GeForce 6800 Ultra DDL graphics card for our dual 2.5GHz Power Mac G5 needed to drive our 30" Apple Cinema Display. All three pieces of our system had been ordered at the same time from Apple right after they were announced but they arrived separately over the month of October 2004--life on the cutting edge.

Our review of the new Nvidia graphics card was posted on November 8. Not only did it drive the fabulous 30" Cinema Display beautifully (it can actually drive two at the same time), it provided exception performance for 3D games. The Nvidia GeForce 6800 Ultra DDL with 256MB of VRAM is a behemoth graphics card, taking up the full length of the 8X AGP graphics slot and covering the PCI-X slot next to it as well. Still, its performance on other graphics benchmarks was disappointingly on par with the ATI 9600 XT that came stock with our Power Mac G5.

In the works at the time was ATI's response, a Radeon X800 XT for the Mac that would also have 256MB of VRAM and drive Apple's 30" Cinema Display through a multi-link DVI port. Early reviews of the Windows version were mixed. They first gave the crown to ATI based on tests with Unreal Tournament 2003 but then it shifted to Nvidia based on tests with DOOM 3.

Prior to Macworld in January, only a few pre-production units had found their way into the hands of selected reviewers. We first got our hands on ATI's much-touted Radeon X800 XT for the Mac in January at Macworld San Francisco and were very impressed with its performance. We were told by ATI executives that commercial production was underway and that the Radeon X800 XT would be in the retail channel the following week. To make a long story short, despite ordering right away our Radeon X800 showed up in late April just few days before ATI released a firmware v107 update and an ATI Displays 4.5 update that immediately rendered our Radeon X800 XT useless.

By the time ATI thought they had the firmware and Displays issues fixed, Mac OS X 10.4 and 10.4.1 were released and its late May. So ATI released another firmware update v108 and it also had issues with our setup, i.e. unable to run 3D games full screen and unexpected quits and freezes using ATI Displays 4.5.1 to adjust anisotropic filtering. We waited awhile for ATI's support to get back to us and ATI's advice was to downgrade back to the original firmware v104. It's now early June, the WWDC 2005 is on and there are rumors of an ATI Radeon X850 XT for the Mac. While we still don't have an ATI Radeon X800 XT that works with the latest firmware and our system, it's been 6 months since we were promised one and 7 months since we got our Nvidia 6800 Ultra so we ran our shootout anyway. [Note: Later in the day, June 13, ATI released firmware 109 which fixed the problem of running 3D games at full screen.]

3D Games

First, we ran benchmarks on four popular 3D games: the venerable Quake III Arena, Call of Duty, Halo and the most recent "system-buster" DOOM 3. Our 30" Cinema Display has a native resolution of 2560x1600 pixels and Halo is the only game that supports it. We normally play full screen at a resolution of 1024x768 to allow enabling most eye candy and still get a frame rate of 60 or better. The way we test and the settings are listed in our SpeedFAQ. For this shootout we retained all setting but simply increased the resolution to stress the cards' performance. Then we used ATI Displays to alter the Radeon X800 XT's OpenGL settings to illustrate the excellent degree of control that you get with an ATI Radeon graphics card.

Here are the results using our dual 2.5GHz Power Mac G5 with 1.5GB RAM running Mac OS X 10.4.1:

3D Games--ATI X800 XT vs Nvidia 6800 Ultra Graphics Cards (frames per second)
Game/- Card
Resolution (Pixels)
1024x768
1600x1200
2048x1536
2560x1600
Quake III Arena
- ATI X800 XT
400
365
204
n/a
- NV 6800 Ultra
401
381
358
n/a
Call of Duty
- ATI X800 XT
86
85
76
n/a
- NV 6800 Ultra
86
84
81
n/a
Halo
- ATI X800 XT
74
50
36
26
- NV 6800 Ultra
67
38
26
20
DOOM 3
- ATI X800 XT
53
39
n/a
n/a
- NV 6800 Ultra
61
53
n/a
n/a
*Blue is better in each card comparison within the same resolution.
n/a means that the particular resolution is not available for that game.

In the 3D game tests there is no clear overall winner in speed. Nor is there a clear winner in appearance--both do a terrific job of rendering excellent graphics in these games. For the older 3D games (Quake III Arena and Call of Duty), the ATI mostly does better than the Nvidia by a few percentage points at the lower resolutions tested but runs out of gas and is bested significantly at the highest resolution that the game supports. If Halo is your game, however, the ATI really shines all the way up through the native resolution of our 30" Apple Cinema Display. On the other hand, if DOOM 3 is your game, the Nvidia significantly bests the ATI at both supported resolutions.

Other Graphics Card Benchmarks

Next we ran our other benchmarks that relate to graphics performance. We also ran a benchmark not usually associated with graphics performance, the Startup time, when we noticed a significant difference during use and testing of other benchmarks They are described in our SpeedFAQ. Here are the results, either as a score where higher is better or as time in seconds where lower is better:

Other Graphics--ATI X800 XT vs Nvidia 6800 Ultra Graphics Cards (score or seconds)
Test
Graphics Card*
ATI X800 XT
NV 6800 Ultra
Cinebench 2003 Rendering
639
646
Cinebench 2003 OpenGL HW
1844
1801
Let1kWindowsBloom
8
7
Xbench Quartz Graphics
471
521
Xbench OpenGL Graphics
297
234
Xbench User Interface
514
517
Scroll 142p File (sec)
14
12
Startup (sec)
53
37
*Quartz 2D Extreme Enabled
**Blue is better in each card comparison

Cinebench 2003 is a wash. Even though the numbers are different for Cinebench 2003, they are not significantly different given the variability in individual runs and that the results are for an average of three runs. In every other test except the Xbench OpenGL Graphics test, the Nvidia GeForce 6800 Ultra DDL card is faster and in many cases much faster. Curiously, starting up (chime to a complete desktop) is affected by the ATI card. Virtually all of the difference is during the black screen phase, i.e. before the gray screen with the dark gray Apple logo.

In Summary

We are surprised that the Nvidia GeForce 6800 Ultra DDL performed better overall than the ATI Radeon X800 XT. The Nvidia will also drive two 30" Apple Cinema Displays where the ATI will only drive one. It is also a bit quieter to our ears than the ATI card. The downside is that the Nvidia costs $100 more than the ATI offering and takes an extra PCI slot. The other ATI positive is that the card has an ADC port while the Nvidia has 2 DVI ports so if one has an older Apple display or non-Apple monitor that's a significant benefit.

We are disappointed in the problems that we have experienced with the ATI Radeon X800 XT, a first for us. Our previous ATI's cards have always been top-performers and generally problem free. And, according to postings on other web sites and forums, it's not just our card that has been having the problems. ATI claims that only a small set of system configurations have experienced the problems, although it is still a fairly vocal set of owners if not a large number. Since we continue to have firmware issues as noted above, the remaining problems may lie with the multi-link port and its associated firmware/software needed to drive the 30" Cinema Display.

We recommend those with 30" Apple Cinema Displays fork over the additional $100 for the better all around and problem-free performance of the Apple-made Nvidia GeForce 6800 Ultra DDL. However, others have many reasons to go with the ATI Radeon X800 XT which is less expensive and has somewhat different features. [Note: Later in the day, June 13, ATI released firmware 109 that fixed the problem running 3D games at full screen. We reran the benchmarks for DOOM 3 but got the same frame rate as noted above. Later we re-ran all of the benchmarks and got significant differences for Quake III Arena and Xbench so we revised the tables.--Bill Fox]



Saw something? Send a tip

The archive ran on reader tips. What did you see, where, and do you want the credit?

Read by the editor. Never published without your say.

More in Reviews · This month in the archive