Universal QuarkXPress gets a speed boost, but only on Intel Macs

The desktop publishing community eagerly awaited QuarkXPress 7.01, one of the first major non-Apple programs to be delivered in Universal format. Now that this free update to the non-Universal XPress 7.0 is available, does it deliver on the company’s “fast as hell” promise ?

Yes and no.

Based on a fresh round of testing, we conclude the following: It really matters which computer you use when running QuarkXPress 7; Intel-based XPress 7 users should definitely update to version 7.01; PowerPC users should definitely stay away from this update; and QuarkXPress 7 is not an undisputed speed demon in any scenario.

We compared the relative speed of three versions of Quark—7.01, 7.0, and 6.5, as well as Adobe InDesign CS2—on four systems, a 2.66GHz (standard configuration) Mac Pro, a 1.66GHz Intel Core Duo Mac mini, a dual 2.7GHz Power Mac G5, and a 500MHz Power Mac G4. Overall, the results were mixed and performance gains you get will vary based on the tasks you perform.

No rating change

We did not change the mouse rating for the program as a whole, as the rating was mostly based on the program’s features, which have not changed with version 7.01. We did not penalize version 7.0 for not being Intel native. The original review (   ) noted the program’s sluggish performance on Intel Macs and the company’s promise of the forthcoming Universal version. The Universal QuarkXPress 7.01, which shipped in August, eliminates the sluggish performance on the Intel platform, but XPress 7.01 is even slower than version 7.0 on Power Macs.

The results

Running XPress 7.01 on Intel Macs clearly makes sense, especially if you have one of the new models. On the Mac Pro, XPress 7.01 handily beats XPress 7.0—nearly doubling its performance on computation-intensive activities such as screen rendering and image manipulation (as seen in the Launch and Open Document, Replace and Reflow, Apply Irregular Wrap, Apply Transparency, and Import Word File tests). It also performs modestly better—about 22 percent—than the non-Intel-native XPress 6.5. But it remains half as fast as the non-native InDesign CS2 for these procedures. For our two disk-oriented tests (Create PDF and Import 200MB TIFF File), XPress 7.01 on the Mac Pro outpaces XPress 7.0 by about 54 percent, XPress 6.5 by about 13 percent, and InDesign CS2 by about 28 percent.

The speedup on the Mac mini was less impressive: In our tests, XPress 7.01 ran about 30 percent faster than XPress 7.0 on computation-intensive tasks and was noticeably slower than both XPress 6.5 and InDesign CS2. For the disk-intensive tests, XPress 7.01 was 30 percent faster than XPress 7.0, but half the speed of XPress 6.5. XPress 7.01 on the Mac mini was 50 percent faster than InDesign CS2 for these tests.

If you work on a Power Mac, version 7.01 will significantly slow down XPress operations. In our computation-intensive tests on the G5 system, there was about a 31 percent performance hit with XPress 7.01, compared to 7.0. XPress 6.5 clocked in at about twice as fast as XPress 7.01, and InDesign CS2 clocked in about 3.5 times as fast. An older G4 system showed a 27 percent hit with XPress 7.01 compared to 7.0, while XPress 6.5 ran 3.3 times as fast as 7.01, and InDesign CS2 ran 4.3 times as fast as 7.01. In the disk-intensive tests on the PowerPC systems, we also saw significant slowdowns in XPress 7.01, though they were not as dramatic as in the computation-intensive tests. On the G5, XPress 7.01 took about a 36 percent hit compared to versions 7.0, and a 65 percent hit compared to version 6.5. Compared to InDesign CS2, it took a 54 percent hit. On the G4, the slowdown was significant but less dramatic: XPress 7.01 took an 18 percent hit versus 7.0, and ran at about 40 percent of the speed of 6.5. Compared to InDesign CS, it was about half as fast.

Quark’s response

Quark acknowledges that XPress 7.01 runs more slowly on PowerPC systems than version 7.0 does. Senior Product Marketing Manager Marc Horne recommends that PowerPC users stick with version 7.0, since the functionality is identical. If you’ve already installed 7.01 on your Power Mac, he recommends that you reinstall version 7.0 (that won’t count against your limit on installed licenses). Horne says that the reason for the PowerPC slowdown is that Quark is using a new compiler tool to generate its software and that its engineers haven’t yet gained the experience to optimize performance using the new tool. He says that as Quark’s developers gain more experience with the compiler tool, they’ll be able to better optimize the code, and they will include that improved code with the updates and bug fixes that Quark issues in the future. Horne says this should narrow the performance gap between the PowerPC and Intel versions over time.

As to why the performance boost on Intel Macs (and the corresponding drop on PowerPC) varies on different Mac models, Horne attributes that to better data buses, graphics engines, and other motherboard components on newer Macs. He points out that such new components aid some operations, especially since the compilers used to generate the shipping software are typically optimized for the latest hardware. Older hardware tends to get less attention by the tool developers, so specific functions don’t get the tweaks needed to run at their best speed.

Testing methodology

The tests we used for QuarkXPress 7.0 are variations of tests that Macworld has used in the publishing arena for about 15 years. The seven tests represent different aspects typical of publishing software.

Five tests are computationally intensive, stressing the Macs’ chip, memory, and graphics subsystems and the code that interacts with them. For example, our Replace and Reflow (a find and replace of both text and formatting throughout a document), Apply Irregular Wrap (wrapping three columns of text around a five-point star), and Import Word File tests focus on changing elements through a document, requiring the program to rebuild the entire layout and then update the onscreen presentation. Our Apply Transparency test requires significant computation over the millions of pixels in our 200MB TIFF image, testing how efficient the software is in handling such calculations. While influenced by the Mac’s underlying disk speed, our Launch and Open test mainly examines the software’s ability to set itself up in system memory, apply global preferences, and build the initial presentation of a layout—it’s a lot like what a car’s engine-check process does when you turn the ignition key.

The other two tests—Create PDF and Import 200MB TIFF—primarily test how the software works with the Mac’s storage systems. While there are some computations involved (to generate the PDF code from the layout and to interpret the TIFF file’s bits for display), the bulk of the effort is to gauge how well the software interacts with the file system.

To calculate our overall performance scores, we average the results of the tests, giving each test identical weight. Thus, in the disk-intensive tests, each of the two tests counts for half the score. In the overall tests, each of the seven tests counts for one-seventh of the score. We average the relative performance, not the number of seconds, for each test. Thus, if one disk-intensive test shows XPress 7.01 is 150 percent as fast as XPress 7.0 and the other shows it to be 110 percent as fast, the result is an overall score that XPress 7.01 is 130 percent as fast as 7.0 on that particular Mac. Because we have two disk-intensive tests and five computation-intensive tests, the computation-intensive tests account for approximately 71 percent of the overall score—reflecting the fact that designers tend to work on their layouts’ design and contents much more than they import and export files from them.

We collaborated with Quark on tweaking our original tests for version 7.01. Quark had opined that some of our tests—importing a 200MB TIFF file, wrapping text around a five-point star overlaid on a three-column text page, and applying transparency to that large TIFF file—are extreme and unrepresentative. However, we don’t believe that to be the case. Our test files are not especially complex, and are similar to a typical newsletter, so that we can accurately reproduce them across a range of products and versions. Moreover, as with all software reviews, it’s important stretch the applications to test their capabilities. Quark says it will publish its own tests, which we’ll link to when they become available.

The bottom line

The Universal QuarkXPress 7.01 makes up for much of the slowdown in version 7.0 on Intel Macs. The speedup on the Mac Pro is much higher than on the Mac mini, so users of the latest Macs will see the best performance. Compared to non-native XPress 6.5, XPress 7.01 is 19 percent faster on the Mac Pro but 38 percent slower on the Mac mini. On the two Intel Macs, XPress 7.01 is slower than InDesign CS2, by about 30 to 40 percent. XPress 7.01 is about 30 percent slower than version 7.0 on PowerPC Macs. As a percentage, the slowdown is slightly higher on a Power Mac G4 (32 percent) compared to a Power Mac G5 (30 percent). In all tests, XPress 7.01 is slower than XPress 6.5 and InDesign CS2—running at just a fifth to a half the speed of the others.

[ Former Macworld editor Galen Gruman has coauthored more than a dozen books on QuarkXPress, PageMaker, and InDesign. ]