• Blogs
• Forums
• Tradeshows
• Partners
• White Papers
• Events

Edit Expertise: Platform Shootout

Aug 1, 2006 12:00 PM, By Jan Ozer

Putting six selected workstations to the test.

Every once in a while, it's fun to bring in a range of computers and see how they compare performance-wise. Certainly, the cross-platform capability of Apple's Boot Camp throws an interesting wrinkle into the picture, as does the general availability of AMD-based Windows computers.

We ran a series of tests on six desktop and notebook computers, primarily with cross-platform applications such as Adobe Photoshop and Sorenson Squeeze. First: a quick description of the hardware and software tested.

We tested two Macs, starting with 17in. MacBook Pro with a 2.16GHz Intel Core Duo processor, 2GB of DDR2 SDRAM, and a Serial ATA (SATA) hard drive. We also tested a dual-processor, dual-core 2.5GHz Power Mac G5 Quad with 4GB of DDR2 SDRAM.

For Windows, we tested an HP xw4300 workstation, running a 3.4GHz Intel Pentium D (for dual-core) processor with 2GB of RAM, and a Dell Precision Workstation 670, running dual 3.6GHz Intel Xeon processors with Hyper-threading Technology and with 2GB of RAM. HP also provided an HP Compaq nw9440 mobile workstation with a Core Duo Processor T2600 running at 2.16GHz with 2GB of RAM, which we tested against the MacBook Pro (with the identical processor) running Windows via Boot Camp.

To assess AMD performance, Boxx Technologies provided a GoBoxx 1400 mobile workstation built around an AMD Athlon 64 X2 dual-core processor with a SATA drive. (Interestingly, I produced most of the DCP NAB 2006 podcasts on the GoBoxx, and it performed like a champ. Check them out at digitalcontentproducer.com/podcasts). At 11lbs., compared to 7lbs. for the nw9440 and MacBook, the 1400 is a serious piece of hardware, which we tested in both the notebook and desktop categories.

Tests included DV and MPEG-2 rendering from Apple Final Cut Pro (universal version 5.1.1) and Adobe Premiere Pro 2.0, using a four-minute project file that included chromakey, slow motion, image stabilization, color correction, still-image pan and zoom, backlight correction, title and logo overlays, and audio noise reduction. For Squeeze (version 4.3), we compressed a three-minute video file to 1Mbps AVC MPEG-4 format, and for Adobe Acrobat (version 6.0), we aggregated 27 8-megapixel images into a single PDF file.

We ran Adobe Photoshop CS2 on all computers except the MacBook running Windows, where the installer failed, forcing us to test Photoshop CS instead, which does not offer noise reduction. The specific Photoshop tests are listed in the various performance tables, and more fully described below.

We tested most applications in their default setups except for Photoshop, where we increased allowable memory usage from the default 55 percent to 75 percent. This boosted performance on all computers except for the G5, which had 2GB more RAM than any other computer. If you find Photoshop churning when working with large images, either increase the amount of allowable memory, buy more RAM, or do both. With this in mind, let's answer some questions.

What's the fastest computer in the land?

This depends upon the function. In video rendering functions, the AMD Athlon is a bear, besting both the 3.4GHz Pentium D and dual 3.6GHz Xeon in all tests by as much as 24 percent — a significant margin. The AMD also proved faster than the G5 in Sorenson tests, but Final Cut vs. Premiere Pro comparisons are inconclusive.

We ran two levels of Photoshop tests — one series on smaller images, and the other on a single, very large (162-megapixel) image. In the smaller-image tests, the Windows computers are bunched together tightly, all significantly faster than the Power Mac G5. Most of that time difference related to the contact sheet test (where we created a contact sheet out of the 27 images), which, for some reason, proved very slow on both Macintosh computers.

In large-image tests, the G5 emerged as the leader, edging the dual-processor Precision and trouncing the AMD — as did all Intel-based computers. Interestingly, the single-processor HP xw4300 was the clear winner in the Acrobat creation tests, with the G5 second and Precision and Boxx computers pretty much grouped in third.

Conclusions? When buying a rendering station, you should strongly consider an AMD-based computer, which appear to be highly optimized for compression-related tasks. On the other hand, for other facets of digital content creation — particularly large-image Photoshop editing — either of the true dual-processor systems appears to offer a significant performance advantage.

How do Intel- and G5-based Macs compare?

As Table 2 shows, when it comes to rendering in Final Cut Pro, the Power Mac outperforms the MacBook about 2:1, meaning you have to wait twice as long when using the MacBook.

The table also shows that dual-platform applications run much faster on Windows than on OSX. Because Adobe has announced it will not retrofit current Adobe applications for Intel-based Macintosh computers, if you have such a computer and get paid for your time, consider the Boot Camp and Windows versions of these applications.

Generalizing the lesson, it appears that Mac developers should be able to achieve dramatic speed improvements with Intel-optimized code. For this reason, for computationally intense functions, you should acquire Intel-optimized versions of your applications as soon as available.

How do Windows desktops compare to Windows notebooks?

This one was a surprise. Sticking with Intel-based computers for a moment, if you buy a notebook with a Core Duo processor and SATA drives, performance is very similar to a single-processor desktop. For example, the HP xw9440 beat the xw4300 in Premiere Pro encoding trials and small-image Photoshop tests, although it lost to the desktop in most other categories — but generally by a small margin.

In contrast, with previous generations of single-core notebooks with IDE or similar drive technologies, the performance difference with desktops was much more profound. Though workstations have greater hard disk capacity, the availability of high-speed external USB2 drives negates this advantage, and the MacBook offers a FireWire 800 port, as well — a feature that other vendors will certainly adapt.

Previously, full-sized workstations were the only alternative for content producers; now, if you're not purchasing a dual-processor workstation, a notebook delivers very similar performance. Note that this conclusion may not hold for 3D design programs such as Autodesk 3ds Max or Parametric Technology Corporation Pro/Engineer, which rely heavily on 3D performance from the graphics card. For bread-and-butter editing and rendering tasks, however, you're good to go.

Also interesting were the performance differences between the MacBook and HP nw9440. Although it uses the same processor, the nw9440 beat the MacBook in nearly every task, often by significant margins. Balanced against this, of course, is the MacBook's unique ability to run Macintosh programs as well as Windows. While those on a perpetual content creation deadline would prefer the HP, buyers considering Windows-based notebooks for casual production should include the MacBook on their shortlist.

Finally, the Boxx notebook, which, at 11lbs., forced us to buy a briefcase with wheels for transportation to and from NAB. In truth, while none of these notebooks, which all feature 17in. screens, could open on the tray table of your average coach seat, you wouldn't even think about it with the GoBoxx 1400. Still, as a desktop replacement — especially for video rendering-type production — the unit is a real winner.

That said, all the notebooks were winners in this trial, as were the dual-processor workstations for certain tasks. Could this signal the demise of the single-processor workstation?

---

To comment on this article, email the Digital Content Producer staff at dcpfeedback@prismb2b.com.

© 2008 Penton Media, Inc.