Giga Designs M5D-1213Q G-Celerator DUAL 1.33 GHz in a Digital Audio G4

I. Intro & Installation | II. Temps and Cooling | III. System Tests | IV. Graphics Tests | V. Apps Tests

V. Applications Tests

I set aside the graphics cards because they didn't impact the following applications tests results.

Links to apps tests:  iTunes | QuickTime | Final Cut Pro | InDesign | Photoshop | Filemaker Pro

iTunes:
I used a stopwatch to record the time to convert an audio CD outer track (4:06 min) to MP3 (with default settings) -- average of 3 runs:

iTunes v6.4	733 MHz	2 x 1.33 GHz
Time to convert 4:06 min audio CD track to MP3 (secs)	22	15

QuickTime Pro:
I used a stopwatch to record the time to convert a 20:44 min, 219.1MB QT movie to MPEG4 (with default settings) -- average of 3 runs:

QuickTime v6.5.1	733 MHz	2 x 1.33 GHz
Time to convert 20:44 min, 219.1MB QT movie to MPEG4 (secs)	458	276

Final Cut Pro:
For Final Cut Pro v3.0.4 testing, I set the scratch disk to a Granite Digital FireVue FW800/USB2 enclosure with a Hitachi 60GB 180GXP. Drive connected to the DA's motherboard FW400 port using the included FW800 to FW400 cable. (I was testing my Cube around the same time, hence the Cube-friendly setup.) FCP project files were stored in the home directory, media copied to the FW drive.

In initial testing, I used 3 projects and recorded render times:

1. Final Cut Pro 3 Tutorial 1 (36.13 sec. project)
2. Final Cut Pro 3 Tutorial 2 (8 sec. project)
3. Final edit from Jerry Hofmann's Jerry Hofmann on Final Cut Pro 4 (14:49.26 min. project)

Jerry had kindly sent me an FCP3 compatible version of the final edit for the DVD included with his FCP4 book. It's a large project full of edits, so I figured it would be a good 'real world' test. As I explain below, it didn't bring out any dual CPU advantage. I later created longer versions of the FCP3 tutorials by copying and pasting the project:

4. Final Cut Pro 3 Tutorial 1 - 10 iterations (6:04.22 min project)
5. Final Cut Pro 3 Tutorial 2 - 10 iterations (1:20.22 min project)

In further testing, I also used TinkerTool System to shut down one of the CPUs to get render times with only one 1.33GHz CPU.

Here are the full results of render times, using a stopwatch:

Final Cut Pro v3.0.4 render times (min:sec)	733 MHz	GigaDesigns 1 x 1.33 GHz	GigaDesigns 2 x 1.33 GHz
FCP3 Tutorial 1	2:35	1:40	1:04
FCP3 Tutorial 2	2:22	1:26	0:53
FCP3 Tutorial 1 - 10 iterations		18:56	13:17
FCP3 Tutorial 2 - 10 iterations		17:18	12:44
Final edit from Jerry Hofmann on Final Cut Pro 4	21:18	13:30	13:20

Commentary:
With Jerry Hofmann's project, the Giga dual improves times significantly over the stock 733MHz, but there's negligible difference when comparing one versus two 1.33GHz CPUs. In fact, my Cube with a Giga 1.3GHz M7-1213N renders it in the low 11:00 min range; the first sign that was something wasn't right with this project as a dual CPU benchmark.

After putting aside initial disappointment, I re-ran FCP3 Tutorials with a single 1.33GHz. As the above table shows, with these projects, there's a clear difference between the 733MHz, a single 1.33GHz, and dual 1.33GHz. (And the Giga dual 1.33GHz easily beats my Cube with those tutorials as expected.) Because they're so short, I was afraid the FCP tutorials might skew the dual advantage. So I created longer versions composed of 10 iterations of originals; the dual processors still render them significantly faster. Had I known about this earlier, I would have timed the longer versions on the stock 733MHz as well.

I suspect that the different single v. dual results with Jerry's project and FCP tutorials may have something to do with the original source material/media. The FCP3 tutorials have high-res, Jerry's low-res. It could be the low-res material didn't elicit the benefits of the dual processors.

I'll mention again that with FCP3 dual processors also help with the GUI, which is much snappier and responsive compared to a single 1.33 GHz CPU. Menus, disclosure triangles, windows opening, etc. are noticeably quicker with the dual.


Adobe InDesign CS:
With a stopwatch, I timed the export to pdf using the default Press settings. I used two Lessons from InDesign2: Classroom in Book, but increased their sizes:

   1. Lesson 3 - 40 pages, 20 spreads. (22.4 MB file; 1.6 MB pdf created)
   2. Lesson 7 - 10 pages, 10 spreads. (25 MB file; 16 MB pdf created)

Lesson 3 originally had a book layout, 2 page spread -- light on graphics. Lesson 7 was a single page design, heavier on the graphics with layers that must be flattened during the export. The following times are an average of 3 runs:

InDesign CS v3.0.1 - Time to export as pdf (secs)	733 MHz	2 x 1.33 GHz
Lesson 3 as 40 pages	77	46
Lesson 7 as 10 pages	207	126

Adobe Photoshop:
I ran the PS7Bench v1.1.1 Advanced test (50GB image) with Photoshop v7.0.1 and Photoshop CS v8.0.

Scratch disc: boot volume
Cache levels: 4
RAM: 350MB (25%)
Resolution: 1024 x 768
Colors: Millions

RAM was set low because of Cube testing. (I forgot about the cache levels setting but had already run the tests several times.) Times below are averages of the 3 runs for each of the 21 actions in secs.

PS7 Bench v1.1.1 (secs)	Photoshop v7.0.1			Photoshop CS v8.0
	733 MHz	2 x 1.33 GHz		733 MHz	2 x 1.33 GHz
90° Clockwise Rotation	4.3	4.4		3.1	3.0
9° Clockwise Rotation	13.0	9.8		8.4	6.4
.9° Clockwise Rotation	12.3	9.7		7.3	5.4
1 px Gaussian Blur	5.4	5.1		3.2	3.0
3.7 px Gaussian Blur	9.9	7.4		7.6	5.2
8.5 px Gaussian Blur	12.7	9.4		9.7	6.1
50%, 1 px, 0 level Unsharp Mask	8.3	7.9		3.8	3.3
50%, 3.7 px, 0 level Unsharp Mask	11.3	7.5		8.8	5.7
50%, 10 px, 0 level Unsharp Mask	12.5	8.1		10.2	6.3
Despeckle Filter	6.6	5.8		3.5	3.4
RGM to CMYK	16.3	9.9		10.9	5.0
60% Reduction	3.6	3.2		2.8	2.2
Lens Flare Filter	16.5	10.7		11.0	6.8
Color Halftone	13.4	9.8		27.2	16.9
NTSC Colors Filter	13.6	8.4		10.3	7.2
Accented Edges Brush Strokes	34.3	19.1		41.4	24.4
Pointillize Filter	52.5	21.2		66.7	28.2
Watercolor Filter	74.7	40.4		72.0	46.0
Polor Coordinates	15.0	8.1		12.7	6.2
Radial Blur	97.6	37.3		85.6	34.4
Lighting Effects	10.6	6.7		8.8	5.6
Total Time 21 actions:	445.4	250.3		415.1	230.6

With PSBench, the additional time to render the action on the screen is ignored, most likely because it's dependent on the graphic card. With some actions, the Radeon 9600 Pro rendered up to 3 secs faster; while with others there wasn't any noticeable difference between the 9600 Pro and GeForce 2MX.


Filemaker Pro Client:
I used FMBench v2 benchmarks for Filemaker Pro v6.0.4 and v7.0.3. Database was stored in the home directory; 10,000 records used.

FM Bench v2 - 10,000 records (secs)	Filemaker v6.0.4			Filemaker v7.0.3
	733 MHz	2 x 1.33 GHz		733 MHz	2 x 1.33 GHz
Simple Find	0	0		0	0
Complex Find	1	0		1	1
Related Find	2	1		0	0
Generate Records	62	35		126	82
Sort	2	1		2	2
Export	3	1		7	4
Delete	18	10		25	16
Import	48	26		111	71
Summarize	2	1		3	2
Replace	54	29		105	69
Relookup	9	5		26	16
Loop Script	13	7		13	9
Total Test Time (reported by FM Bench):	208	113		416	268

I. Intro & Installation | II. Temps and Cooling | III. System Tests | IV. Graphics Tests | V. Apps Tests

The Gurus Certify This Review Rules