Mac Doom 3 Benchmarks Revisited

With new video cards in hand, we've come back to provide you with more details about how Aspyr's Macintosh conversion of Doom 3 runs. This should give Mac gamers a more general idea of how well the game performs on a high-end Mac system. To that end, we've now tested the game with several graphics cards: ATI's Radeon X800 XT Mac Edition and Radeon 9800 Pro Mac Special Edition cards, Apple's Nvidia 6800 Ultra card and the stock Radeon 9600XT card that came with our test system.

There are certainly a lot of ways to tailor Doom 3's performance to suit your computer, but that's something to examine another time. For comparison's sake, we kept the configuration the same as many of the authoritative benchmarks we've seen on the PC side of the fence, which, for the most part, are the default settings on high-end graphics cards: Graphics settings set to High Quality, and all advanced settings turned on except for Vertical Sync and Full-Scene Anti-Aliasing (FSAA), where noted. Anisotropic filtering was kept to its default setting of 8x for most tests. The one exception to that was the tests performed with the 9600XT -- Doom 3 defaulted graphics quality settings to Medium and set anisotropic filtering to 1, and we maintained those default settings.

High-end results

In our first look at this topic, we provided benchmarks using a Power Mac G5 dual-processor system clocked at 2.5GHz, equipped with 2.5GB of RAM and Mac OS X v10.3.8 and ATI's Radeon X800 XT Mac Edition graphics card. Here's what we discovered:

Radeon X800 XT Mac Edition Results

640 x 480 800 x 600 1024 x 768 1280 x 1024 1600 x 1200
No FSAA 47.1 45.3 41.3 35.6 32.3
FSAA 2x 44.9 42.5 38.2 31.7 27.6
FSAA 4x 40.8 36.7 30.3 23.6 20.8

BEST RESULTS IN BOLD. ALL RESULTS ARE IN FRAMES PER SECOND.

All tests were performed at "High Quality" video settings with all Advanced Options turned on (High Quality Special Effects, Enable Shadows, Enable Specular, Enable Bump Maps) except for Vertical Sync. Anisotropic filtering was set to 8x. Frames per second averages were achieved running the timedemo demo1 command; the demo was run twice, and the higher average was used.

Not to put too fine a point on it, but the results were discouraging for Mac gamers who were hoping to see similar performance to a comparably equipped Windows PC. Athlon-based systems equipped with similar graphics cards turn in much higher results than the Power Mac does.

That brings us to another point: Many readers of our last benchmark article took umbrage with our comment, ".. there's no substitute for CPU cycles when it comes to games like this." We think that a faster PowerPC 970 clock speed would certainly help get those numbers up somewhat. But readers experienced with PC gaming pointed out quite correctly that Windows PCs equipped with AMD Athlon CPUs clocked at similar speeds to our Power Mac turn in dramatically higher frame rate averages than the Mac. Point taken, and thanks to everyone who responded.

Doom 3 turns in higher numbers on the PC when used with Nvidia's GeForce 6800 graphics hardware. It's no surprise -- Nvidia has long worked closely with Doom 3 developer id Software to make its OpenGL drivers and hardware work as well with this game as possible. Do the results hold up the same on the Mac? We tested an Apple Nvidia GeForce 6800 Ultra card to find out.

This is the same $599 card you can buy through the Apple Store that drives two 30-inch Apple Cinema HD displays on the Mac simultaneously. That's not exactly practical for gamers, but certainly useful if you're planning on professionally editing digital video, doing 3D visualization or anything else that requires a lot of desktop space.

Apple Nvidia GeForce 6800 Ultra Results

640 x 480 800 x 600 1024 x 768 1280 x 1024 1600 x 1200
No FSAA 48.3 48.1 46.8 47.7 34.6
FSAA 2x 48.2 45.9 39.7 30.3 22.9
FSAA 4x 47.1 42.8 34.2 25 18.2

BEST RESULTS IN BOLD. ALL RESULTS ARE IN FRAMES PER SECOND.

All tests were performed at "High Quality" video settings with all Advanced Options turned on (High Quality Special Effects, Enable Shadows, Enable Specular, Enable Bump Maps) except for Vertical Sync. Anisotropic filtering was set to 8x. Frames per second averages were achieved running the timedemo demo1 command; the demo was run twice, and the higher average was used.

The GeForce 6800 Ultra fared moderately better at most resolutions, with or without Full Scene Anti-Aliasing turned on. With the exception of 1024 x 768 and 1280 x 1024 pixel resolution without anti-aliasing, however, the results weren't remarkably better. That's a significant departure from the PC benchmarks we've seen, but it's consistent with Aspyr Media's expectations, as noted in our previous benchmark article. The Nvidia 6800 Ultra quickly ran out of gas at 1600 x 1200 with anti-aliasing on -- enough so that the X800 outpaced it at 2x and 4x FSAA.

Mid-range performance

We also put the game through its paces on an ATI Radeon 9800 Pro Mac Special Edition -- a 256MB graphics card designed especially for Power Mac G5 systems. Up until the X800's release this past January, the Radeon 9800 Pro was the fastest ATI-based graphics card you could get for your Power Mac G5. ATI still makes the card available for Mac users whose pockets aren't deep enough for the X800 card.

ATI Radeon 9800 Pro Mac Special Edition Results

640 x 480 800 x 600 1024 x 768 1280 x 1024 1600 x 1200
No FSAA 40.2 36.1 28.2 20.6 15.6
FSAA 2x 36.3 30.2 23.6 17.2 12
FSAA 4x 25.6 20.3 15.3 10.9 8

BEST RESULTS IN BOLD. ALL RESULTS ARE IN FRAMES PER SECOND.

All tests were performed at "High Quality" video settings with all Advanced Options turned on (High Quality Special Effects, Enable Shadows, Enable Specular, Enable Bump Maps) except for Vertical Sync. Anisotropic filtering was set to 8x. Frames per second averages were achieved running the timedemo demo1 command; the demo was run twice, and the higher average was used.

As you can see, the Radeon 9800 has some significant limitations. While performance was still adequate at lower resolutions, the 9800 couldn't keep up with the X800 or the 6800 at the resolutions many gamers prefer (1024 x 768 and higher). In fact, the 9800 is only turning in playable speeds (we'll loosely define "playable" as 30 frames per second or higher) at 640 x 480 and 800 x 600 pixel resolutions. At higher resolutions with anti-aliasing on, the game slows to an unplayable crawl.

Last and arguably least, we put Doom 3 through its paces using the 128MB Radeon 9600XT graphics card that was included as standard equipment on the 2.5GHz Power Mac G5 -- also available as $50 upgrade for other Power Mac G5s that include an Nvidia GeForce FX 5200 Ultra card.

ATI Radeon 9600XT Results

640 x 480 800 x 600 1024 x 768 1280 x 1024 1600 x 1200
No FSAA 34.9 27.3 19.8 13.4 9.1
FSAA 2x 29.4 22.7 16.5 10.6 6.4
FSAA 4x 21 16 11.5 6.8 3.9

BEST RESULTS IN BOLD. REFERENCE SYSTEMS IN ITALICS.

All tests were performed at "Medium Quality" video settings with all Advanced Options turned on (High Quality Special Effects, Enable Shadows, Enable Specular, Enable Bump Maps) except for Vertical Sync. Anisotropic filtering was set to 1x. Frames per second averages were achieved running the timedemo demo1 command; the demo was run twice, and the higher average was used.

Although the Radeon 9600XT card is standard equipment on Apple's highest-end Mac configuration, it's clearly not up to the task of running Mac OS X's highest-end game unless you throttle down the graphics quality and resolution significantly. It lacks the gobs of VRAM and fast processor you'll find on better, more expensive cards, but may yield playable results if gamers are willing to compromise.

Conclusion

What's clear from all these tests is that there are very significant differences running Doom 3 on high-end Mac and PC hardware. It's hard to say exactly where the problems lie, but there are some fundamental differences between the Mac and Windows worlds that may go at least part of the way to explaining these differences.

ATI and Nvidia have both worked extensively with id Software to maximize the performance of their OpenGL drivers with Doom 3, and we expect some of that work will have to carry over to Apple's drivers if we're to see the numbers head upwards. What's more, there are differences in the way compilers work, the ways that PowerPC chips crunch numbers compared to AMD and Pentium hardware, and other variances that affect performance. Some of these factors may be able to be compensated for in the future, but it's tough to say for sure.

Doom 3 timedemo benchmarks are only one measure of how well a computer runs this game. While they help to get an overall understanding of how well our test system ran the game, it's also worth noting that in actual gameplay, general performance was often better than these numbers might indicate -- on high end cards, sometimes pegging at Doom 3's cap of 60 frames per second.

Doom 3 uses advanced lighting and shadowing effects the likes of which we haven't seen on the Macintosh before, so Aspyr is on the bleeding edge with this game. The vast majority of games on the Macintosh aren't nearly this demanding, and Mac gamers with well-configured rigs will see perfectly fine performance for just about anything they throw at their machines.

While the Macintosh numbers may disappoint gamers looking to keep pace with their PC counterparts, there's a silver lining, because there's hope that these numbers will improve over time.

When id Software's last game for the Macintosh -- Quake III Arena -- came to market, there was a Mac to PC performance disparity there too. That eroded, though, as id Software updated the game's code and as Apple's own OpenGL drivers improved. We're optimistic that if Apple, id, Aspyr, ATI and Nvidia work together in the same direction, we'll likewise see Doom 3's performance on the Mac improve.

Of course, there's always the promise of faster video cards and better Mac hardware just around the corner. That's certainly something to look forward to, but we hope that all parties involved can make Doom 3 work even better on today's hardware than it does right now -- not just for gamers who buy this title now, but for games that use Doom 3's core engine which are likely to find their way to the Mac in the months to come.

Subscribe to the Best of Macworld Newsletter

Comments