We’ve lauded the recently introduced Mac Pro desktops for their performance prowess, whether it’s the standard eight-core 2.8GHz configuration or the build-to-order 3.2GHz model. But there’s been one consistent disappointment with the Mac Pros introduced in January regardless of which version we’re testing—the performance of the ATI Radeon HD 2600XT graphics cards in our 3-D tests, particularly Quake 4 frame rate tests.
In fact, our testing has found the previous generation of Mac Pros beat the new systems in Quake. More troubling, the new models were even outperformed by a 2.5GHz quad-core Power Mac G5 in this frame-rate test.
We theorized that the slow performance by the new graphics cards could be due to a lack of driver optimization and hoped that an OS update might put things right. Well guess what? It happened. Kind of.
You see, Apple shipped a later version of OS X 10.5.1 on the new Mac Pros, and that version included later graphics drivers than the version available from Software Update. As a rule, we reformat the drives before testing and we’ve been restoring our systems from a master Speedmark image and then running all of the updates on top of that. So while our initial testing was with OS X 10.5.1, it wasn’t the specific version of 10.5.1 that Apple shipped only in the boxes of new Mac Pros. Because of this, if you’re the owner of a new eight-core Mac Pro, you may not see the profound difference we saw after applying the graphics update.
What graphics update? Well, on Monday, Apple released OS X 10.5.2. Once you’ve installed that version of OS X 10.5, another update shows up in your Software Updates window—Leopard Graphics Update 1.0. And though the note accompanying the download says only that the update “improves the stability and compatibility of your Mac,” I’m happy to report that it improved the new Mac Pro’s frame rates performance in Quake—and pretty dramatically at that.
10.5.2 Graphics Tests
| 10.5.1 | 10.5.2 | 10.5.1 | 10.5.2 | |
|---|---|---|---|---|
| Unreal Tournament 2004 | Unreal Tournament 2004 | Quake 4 | Quake 4 | |
| FRAME RATE | FRAME RATE | FRAME RATE | FRAME RATE | |
| Mac Pro Xeon/2.8GHz (8 cores) | 100.8 | 117.9 | 33.3 | 63.7 |
| Mac Pro Xeon/2.66GHz (4 cores) | 87.8 | 92.7 | 51.4 | 54.7 |
| Power Mac G5/2.5GHz (4 cores) | 32.6 | 32.2 | 37.7 | 41.2 |
| >Better | >Better | >Better | >Better |
* Best results in red. Reference system in italics.
How we tested: Each system was running with 2GB of RAM and either Mac OS X 10.5.1 or 10.5.2 with the graphics update applied. We used Unreal Tournament 2004’s Antalus Botmatch average-frames-per-second score; we tested at a resolution of 1,024 by 768 pixels at the Maximum setting with both audio and graphics enabled. We ran Quake 4 net time demo at 1,024 by 768 pixels, full screen, high quality settings.—MACWORLD LAB TESTING BY JAMES GALBRAITH, JERRY JUNG, AND BRIAN CHEN
The standard configuration Mac Pro powered by an eight-core 2.8GHz Xeon chip, jumped from 33.3 frames per second in our 1,024-by-768, high-quality test up to 63.7 frames per second after switching to OS X 10.5.2 and the Leopard Graphics Update. The older systems we retested with the graphics update—a four-core 2.66GHz Mac Pro and that 2.5GHz Quad Core Power Mac G5—both gained about 3 frames per second in Quake performance, but now they both trail behind the new Mac Pro and its standard issue ATI Radeon HD 2600XT graphics processor.
In our Unreal Tournament 2004 test—an older benchmark we continue to run due to its compatibility with a great number of Mac systems—our G5 test system’s UT frame rates were virtually unchanged. The older Mac Pro gained a few frames per second with the graphics update applied. However, the new Mac Pro saw a more substantial performance gain, boosting its 100.8 frames-per-second performance to nearly 118 frames per second.
Back to the OS X 10.5.1 issue we mentioned up top: If you’re wondering whether this affects the new Mac Pros’ Speedmark scores, it shouldn’t. The biggest impact was in Quake test, which is not part of the Speedmark testing. Nevertheless, we’re doing some retesting with the other version of 10.5.1 to see if there’s any noticeable impact; we’re also looking at how the 10.5.2 update might affect other tests.
We’re still waiting for some of the optional Nvidia graphics cards to make their way to the Macworld Lab, and when they arrive, we’ll test them and report the results as soon as possible.
Note: After publishing this story initially, I decided that the OS X 10.5.1 graphics driver issue should be mentioned higher up in the article, so I republished the article after making that change. The testing data in this article was not affected by my edit.
[James Galbraith is Macworld Lab director.]