Apple Power Mac G5 Quad

Product Review: Page (3) of 4 - 12/19/05

Apple Power Mac G5 Quad Apple's first quad-processor Mac is a performance demon Motion graphics and video encoding
For the motion graphics category, I'm going to start off with Adobe After Effects 6.5--not because the G5 Quad is particularly impressive with that software, but because it bears the most explanation. I dismiss After Effects out of turn as a viable application for testing the performance of the Mac platform against the Windows platform(s). After Effects simply is not written efficiently for the Mac. As I've demonstrated in past articles, it leaves so much potential processing power idle (up to 75 percent at times while rendering) that you can literally run two instances of it and come out with the exact same results--or double the render speeds by running After Effects' render engine off a second drive (or partition) on the same Macintosh.

However, After Effects is such an important core application in and of itself that comparisons between systems are relevant. If you use After Effects, you want the best performance you can get, and that's not necessarily the G5 Quad.

Here are the raw numbers.

Adobe After Effects 6.5 Benchmarks (Results in minutes:seconds. Lower results are better.)
	Apple G5 Quad (4 x 2.5 GHz G5)	Apple G5 Dual (2 x 2.0 GHz G5)	Dell Precision Workstation 470 (2 x 3.6 GHz Xeon)	Alienware MJ-12 7550a Workstation (4 x 2.21 GHz Opteron 275)	HP xw9300 (4 x 2.4 GHz Opteron 280)
Test 1	0:05	0:08	0:03	0:03	0:02
Test 2	0:34	0:53	0:37	0:45	0:40
Test 3	0:59	1:48	1:16	1:00	0:50
Test 4	0:17	0:26	0:19	0:20	0:17
Test 5	1:53	3:16	2:46	2:09	1:46
Test 6	1:58	3:34	1:58	1:47	1:24
Test 7	13:07	31:05	NA	37:49	35:01

Now, as you can see, the G5 Quad beat the fastest PC system in the universe (at the moment) in two tests (the last one quite handily, I might add). But it was beaten in four others (with one tie).

Test 1 was a simple cel-style animation that involves a PICT file and tracing paths. Test 2 was a composite using a variety of effects. Test 3 involved the animation of layers from Photoshop and Illustrator documents with 3D effects and random sequencing of numbers across the screen. Test 4 is a 2D compositew originating in Adobe Illustrator and rendered out at 720 x 486. Test 5 involved moving shapes around in 3D space. Test 6 involved rendering out a 3D environment created entirely in After Effects from 2D images. And Test 7 was a composition called Nighflight, one that involved a 3D layer, moving cameras and lights and an adjustment layer with a Levels filter applied to it.

Take your pick.

Now on to more efficiently written motion graphics programs, the poster child for which is clearly Apple Motion 2. Motion, of course, is a Mac-only program, so we can't compare performance across platforms (at least not until the Intel Macs are out). What you do see in Motion 2 is just stellar performance on G5 hardware. For instance, to compare it with After Effects, I ran two tests on 720p footage involving various effects and adjustments. On the first test, what took After Effects 3:28 to render took Motion 2 only 0:33. On the second test, what took After Effects 6:38 to render took Motion 2 only 0:47. (More on this in the future.) So, Motion being such an efficient program, what kinds of speed bumps do we see when moving from two 2.0 GHz processors to four 2.5 GHz processors? Pretty good, but not as great as I'd expected in some cases. Here are the raw render times for four projects.

Apple Motion 2 Benchmarks (Results in minutes:seconds. Lower results are better.)
	Apple G5 Quad (4 x 2.5 GHz G5)	Apple G5 Dual (2 x 2.0 GHz G5)
Test 1	0:23	0:44
Test 2	0:32	0:53
Test 3	0:21	0:35
Test 4	2:59	4:45

The first test was 12 different blur filters applied to DV-resolution footage in a 10-second (300-frame) project. Certainly nothing you'd do in the real world, but I wanted to try to tax the CPUs with this test. The G5 Quad was almost twice the speed of the dual 2.0 GHz machine.

The second test, again, involved numerous filters--all of the Stylize filters that ship with Motion 2, to be precise. That's 24 filters all being processed at the same time. The G5 Quad rendered all of that in a 10-second (300-frame) project in 16 seconds, while the dual 2.0 GHz G5 took 38 seconds. Both are pretty impressive, but the Quad is about 2.4 times as impressive as the dual 2.0 GHz model. (Keep in mind that the best you'd expect to see as indicated by gigaHertz difference between the two systems is 2.5x performance.)

The third test involves a D1 clip replicated 100 times on the canvas and animated for rotation, scale, color, opacity and position, with separate behaviors controlling each one of those parameters. Here I expected the Quad to stand out a bit more dramatically than it did, but a 1.5x speed boost isn't too bad.

The fourth and final test was a composition made up entirely of particle systems--eight in total, all running on top of one another. The particle systems were made up of both still images and QuickTime movies, and each one was animated with combinations of the behaviors Scale over Life and Random Motion. As you'd expect, these renders weren't particularly quick. The dual 2.0 GHz system rendered the 10-second sequence in 3:06, while the G5 Quad took 1:55 to render the same scene.

So it seems that for Motion 2, the best performance improvements on the G5 Quad were in the area of filters.

I also had a chance to test out Shake 4 on both of these systems. Now, I'm not going to pretend to know a whole lot about Shake. I just touched it for the first time last week, and I'm not quite ready to start teaching courses on it just yet. (Give me at least two more weeks for that.) But I am undumb enough to set up a few scenes, render them and write down how much time those renders took.

Here are the numbers on the render times in Shake. Explanations follow.

Apple Shake 4 Benchmarks (Results in minutes:seconds. Lower results are better.)
	Apple G5 Quad (4 x 2.5 GHz G5)	Apple G5 Dual (2 x 2.0 GHz G5)
Test 1	0:57	1:37
Test 2	0:16	0:38
Test 3	2:12	3:28
Test 4	1:55	3:06

The first test was the first project I ever set up in Shake: 122 frames of 720p video with two filters applied, Gaussian Blur and Grain. The second test was applied to the same footage and involved rather random 3D rotation and scaling over the course of the sequence. The third test involved two HD layers (1,920 x 1,080) with a Primatte keyer applied, for whatever reason, three times. The sequence was 188 frames in length. And, finally, the fourth test begins to show my growing expertise with this program, as I applied a Warp node* to a Matchmover node* tracking a subject over the course of 168 frames of 720p footage. For three of the tests, I experienced about a 1.6x performance gain on the Quad, though in the case of the second test, the improvement was much more pronounced, at about 2.4x.

* Note: We Shake masters refer to them as "nodes," not "filters."

And, finally, I also wanted to test the relative perfomance of the G5 Quad's video encoding. I used two applications for this: Compressor 2 and QuickTime Pro 7.

Compressor first.

Compressor 2 Benchmarks (Results in minutes:seconds. Lower results are better.)
	Apple G5 Quad (4 x 2.5 GHz G5)	Apple G5 Dual (2 x 2.0 GHz G5)
Test 1	1:34	3:14
Test 2	0:32	0:42
Test 3	0:18	0:31

The first test involved converting a 720p Final Cut Reference Movie to a self-contained H.264 movie with a 10.3 Mbps data rate. Obviously Apple has a lot invested in this format, and the result is that, on Apple hardware, you're going to see some pretty efficient encoding. In this case, the G5 Quad was a little more than twice as fast as the dual 2.0 GHz G5.

In our second test, we used Compressor to scale 1080i footage down to SD and encode it as MPEG-2 with a two-pass variable bit-rate encode with deinterlacing at an average data rate of 5 Mbps and a maximum data rate of 8.2 Mbps. The Quad did it in about 76 percent the time it took for the dual 2.0 GHz system to do the same thing. Not an amazing speed bump by any measure.

Then, for the final test, I used Compressor to convert some 720p footage to uncompressed 1080i60. The Quad came in at about 1.7 times the speed of the dual 2.0 GHz system.

With QuickTime Pro, the results for the Quad were a little more impressive.

Apple QuickTime Pro 7 Benchmarks (Results in minutes:seconds. Lower results are better.)
	Apple G5 Quad (4 x 2.5 GHz G5)	Apple G5 Dual (2 x 2.0 GHz G5)
Test 1	0:13	0:26
Test 2	0:04	0:07
Test 3	0:07	0:11

All of the tests here involved exporting from Final Cut Reference Movies to various other self-contained formats, including (in order of the tests): MPEG-4, DV Stream and a QuickTime movie using the Animation codec. The MPEG-4 encode was the most impressive for the Quad, coming in at twice the speed of the dualie.

Page: Prev Page 1 2 3 4 Next Page

Related Sites: Creative Mac , Digital Producer , Media Workstation , Digital Animators , Animation Artist , Siggraph News , Audio Video Producer , Digital Post Production , Oceania , MacAnimationPro , MacDesignPro , CEN - Desktops
Related Newsletter: DMN Newsletter , Timeline Newsletter , Pixels Newsletter , Mac Alert Newsletter , Renders , Digital Media Net , DMNForums , Review Seeker