MAY 14, 2002
Performance: Apples to Apples
I'm one of those people who always needs to have the latest and best the Macintosh platform has to offer. I need it for my professional work, and I also need it because ... well ... I'm a Mac psychofanatic, and few things get me amped up like a fine piece of hardware or software being added to my increasingly unmanageable collection of fine pieces of hardware and software.
This is why I went out and bought a dual 1 GHz G4 the very first day it was announced. I headed off to an Apple Store in my area and walked out with my two-headed speed demon in tow. Not only was it the fastest Mac in history with two of the fastest G4 processors tucked neatly into one box, but it was also one hell of a deal, coming in just $700 more than its slower, single-headed brother, the G4 933. I can easily say I've never been more impressed with the speed of any Mac that I've owned--this being my seventh in about 14 years. On specs alone, it blows away its nearest competitor on the platform, and, at the personal experience level, it just feels peecee-stompin' fast.
But how much faster is it than its nearest current relation, the G4 933? I mean, what do you really get for your extra $700 aside from double the processors, double the stock RAM and 20 GB more hard disk space? In other words, how does it really perform in terms of the software we use all the time? I decided to put the machines head to head with some of the software creative professionals in the visual arts use in their work. Both the dual 1 GHz and the 933 MHz G4s performed extremely well in the tests. But does the dual 1 GHz really give you double the performance with real-world applications?
As you will see below, it depends heavily on the application. Read on.
In this first installment, I've decided to focus on 2D software, and I'll follow it up with an evaluation of 3D software later. For this installment, I ran series of 30 benchmark tests using six different applications--Adobe Photoshop 7, Illustrator 10 and After Effects 5.5; Discreet Combustion 2; Synthetik Studio Artist 2.0; and Apple Final Cut Pro 3.
The selection of applications was designed to represent a cross section of currently shipping software for Mac OS X in graphics, motion graphics and compositing. Both machines were running Mac OS X 10.1.4, and I swapped RAM (1.5 GB) between the two to make sure all things were equal in the tests. Each test was run multiple times, and, where there was variation, an average was determined. Here are the results.
Adobe Photoshop 7
I began my tests with the poster child for the G4's supercomputing performance--Adobe Photoshop. Nary a convention keynote address goes by without Steve Jobs showing Photoshop on the Mac beating the pants off Intel-based systems. And with good reason: The Mac always wins. Photoshop is highly tuned for the G4 processor, and, in fact, Photoshop 7 is even more tuned than Photoshop 6 and runs drastically faster on OS X than it did in OS 9 (but that's a story for a different time).
This is why I was so shocked to discover that Photoshop takes almost no advantage of the G4's dual processors. The performance of the dual G4 1 GHz and the single 933 with Photoshop was almost identical--identically fast, granted, but it would have been nice to see a big difference between the two. Here's what I tested. (Note that because of a quirk with memory usage in action sequences, I've had to use relatively small files with these benchmarks. Adobe says that the difference between single-processor and multi-processor systems becomes more pronounced with files in the neighborhood of 45 MB.)
For the first series, I began with an image at 640 x 480 (as I did with all the Photoshop tests). The action sequence went as follows: resize image 200 percent, Gaussian Blur, Craquelure, Emboss, Sharpen More, Despeckle and Ink Outlines.
Result: Dual G4 1 GHz: eight seconds; G4 933: nine seconds.
The second test involved layer styles and transformation. Beginning with the same image, I converted the background to a layer, applied bevel and emboss and satin, rotated the canvas 360 degrees in 90 degree increments and then performed a horizontal and vertical flip of the layer.
Result: Dual G4 1 GHz: 2.5 seconds; G4 933: three seconds.
For the third test, I decided to do something memory-intensive: I resized the image 1,000 percent and then rotated the layer 180 degrees twice and flipped it horizontally and vertically. (With less than 1 GB of RAM, this can take 20 times as long as these test results indicate.)
Result: Dual G4 1 GHz: five seconds; G4 933: five seconds.
Action 4 consisted of converting the background to a layer, resizing the canvas 500 percent, rotations and flips as in the previous test, scaling the layer 64.7 percent and then trimming the layer.
Result: Dual G4 1 GHz: 6.5 seconds; 933: seven seconds.
For the fifth test, I went into filter effects. Now, in Photoshop, various filters offer radically differing performance. The Gaussian Blur, for example, can take great advantage of the dual processors, while something like Colored Pencil can't. For this test, resized the image 500 percent and then applied the following filters: Ink Outlines, Glowing Edges, Tiles, Bas Relief, Lighting Effects, Difference Clouds, Mosaic, Pinch, Colored Pencil, Gaussian Blur and Curves.
Result: Dual G4 1 GHz: 75 seconds; G4 933: 88 seconds.
Finally, the last test consisted of a huge string of effects, transformations, text manipulation and adjustments. It went a little something like this: Convert background to layer; resize image 500 percent; apply Inner Glow; apply Gradient Overlay; make four different adjustment layers (Curves, Gradient Map, Selective Color and Color Balance); make text; set antialias; warp text (arc); warp text (flag); rasterize text; set selection; duplicate; scale 411 percent; save alpha; select inverse; and fill with solid color.
Result: Dual G4 1 GHz: nine seconds; G4 933: 10 seconds.
Photoshop test summary: So, in short, there wasn't a huge difference between the performance of Photoshop on the dual 1 GHz machine and the 933 MHz machine. In all, the dual G4 performed all of the actions in 87.3 percent the time it took the 933 MHz machine to complete the same actions under the same circumstances, which accounts for just slightly more than the difference between the 933 and one of the dual G4's processors. (Incidentally, when I removed the multiprocessor extension from Photoshop's Plug-Ins folder, the test results on the dual G4 were about the same as with the extension installed.)
1 2 3 4 5 Next
Related sites: AV Video Creative Mac Digital Animators Digital Media Designer Digital Post Production Digital Producer Digital Video Editing DV Format Film and Video Magazine Film Imaging Hollywood Industry The WWUG