Mac OS X 10.4 "Tiger"--How Fast Is It?
While we have been working with versions of Tiger off and on since last summer's Worldwide Developers Conference, we purchased retail copies from Apple's online store and they arrived Thursday. Since then, we have installed Tiger on a 450MHz G4 Cube, a dual 2.5GHz Power Mac G5, a 1.8Ghz iMac G5, a 1.25GHz Mac mini and a 1.67GHz 17" PowerBook G4 and have run our usual suite of speed tests on them. We used the Upgrade installation option on three: G4 Cube, iMac G5 and PowerBook G4. We used the Archive & Install option on the other two: Power Mac G5 and Mac mini. Installation took about 30 minutes and went smoothly with all.
We recommend the Archive & Install saving network settings and preferences, although we did not follow our recommendation on three installations. The reason for our recommendation is that if one has not been dilligent in updating and maintaining Mac OS X, then glitches can show up more readily with an Upgrade Install. However, with an Archive & Install one has to remember to re-install any software that adds things, like Internet Plugins, to the System folder. Examples of this are RealPlayer, MUMenu and TypeIt4Me.
We also recommend that one repair disk and repair permissions from the Tiger installation DVD using Disk Utility before installing Tiger. And, of course, we strongly recommend backing up your important data before installation.
So how fast is it? To determine this we ran our usual suite of speed tests on our two test-bed Macs and three others. The speed tests are comprised of a series of "canned" benchmark applications and several so-called "real world" tests that anyone can repeat on their own Mac for comparison by following the steps that we perform.
Our long-term test Mac (since Mac OS X 10.0) has been a mostly stock, 2000-vintage 450 MHz Power Mac G4 Cube with 704MB RAM. With Mac OS X 10.3, Apple began including improvements that are aimed mainly at the G5 CPU and graphics cards with 32MB of video RAM or more (the Cube's has only 16MB) so we also began using our dual 2.5GHz Power Mac G5 with 1.5GB RAM and an Nvidia GeForce 6800 Ultra DDL graphics card in the speed tests.
For Tiger, we also ran our speed tests on a 20" 1.8GHz iMac G5, a 1.25GHz Mac mini and a 17" 1.67GHz PowerBook G4 to more broadly cover the Mac field. Note that Apple recommends 512MB of RAM for Tiger, so we upgraded our Mac mini to that minimum.
Here are the results for the G4 Cube using Mac OS X 10.4 (Tiger) vs the last update of each "cat" version, i.e. 10.3.9 (Panther) and 10.2.8 (Jaguar) and 10.1.5 (Puma). Other versions back to 10.0 can be found in our previous articles:
| Test | Cube 450 MHz | |||
|---|---|---|---|---|
| Mac OS X Version: |
10.1.5
|
10.2.8
|
10.3.9
|
10.4
|
| Benchmarks: | ||||
| Altivec Fractal (GFLOPS) |
-
|
1.5
|
1.5
|
1.5
|
|
Cinebench 2003 - Rendering
|
-
|
49
|
48
|
48
|
|
- OpenGL Hardware Lighting
|
-
|
92
|
91
|
90
|
| Let1kWindowsBloom 1.0 (sec) |
57
|
61
|
53
|
30
|
| Xbench - CPU |
-
|
52
|
54
|
54
|
| - Threading |
-
|
34
|
42
|
42
|
| - Memory |
-
|
64
|
76
|
77
|
| - Quartz Graphics |
-
|
63
|
74
|
73
|
| - OpenGL Graphics |
-
|
84
|
80
|
88
|
| - User Interface Graphics |
-
|
60
|
102
|
150
|
| - Hard Disk |
-
|
66
|
54
|
56
|
| "Real World" Tests: | ||||
| Startup (sec) |
98
|
54
|
52
|
53
|
| Shutdown (sec) |
5
|
7
|
15
|
16
|
| Launch Classic Mode (sec) |
43
|
-
|
25
|
36
|
| Dup. 700 MB Folder 3600 Files (sec) |
-
|
131
|
152
|
128
|
| Scroll 142p Acrobat Reader File (sec) |
71
|
57
|
55
|
55
|
| Encode MP3 in iTunes (sec) |
-
|
105
|
105
|
104
|
| Export QT FQDV Movie in iMovie (sec) |
-
|
32
|
36
|
38
|
| Quake III Arena 1.32 (frames/sec) |
-
|
22
|
34
|
32
|
There are significant performance improvements in Mac OS X 10.4 for the G4 Cube, the oldest Mac tested. All of the improvements are related to the speed of the Finder. Let1kWindowsBloom measures how fast 1000 Finder windows can be opened and closed and its performance improved over 40 per cent. The Finder speed improvement is also evident in Xbench's User Interface Graphics score (up 47%) and in the time it took to duplicate a folder full of 3600 files (down 16%). OpenGL is also faster in Xbench. For some reason, launching Classic mode in Tiger took much longer than in Pather. Still we are pleasantly surprised to see some speed improvements, particularly in the Finder, in five year old technology. The Cube has an ATI Rage 128 Pro graphics card with only 16MB of video RAM.
Here are the results for the Power Mac G5 comparing 10.4 with several previous versions of Mac OS X 10.3.x:
| Test | Dual 2.5 GHz Power Mac G5 | |||
|---|---|---|---|---|
| Mac OS X Version: |
10.3.5
|
10.3.7
|
10.3.9
|
10.4
|
| Benchmarks: | ||||
| Altivec Fractal (GFLOPS) |
12.7
|
12.7
|
12.9
|
12.9
|
|
Cinebench 2003 - Rendering
|
647
|
644
|
647
|
649
|
|
- OpenGL Hardware Lighting
|
1569
|
1579
|
1601
|
1608
|
| Let1kWindowsBloom 1.0 (sec) |
11
|
11
|
11
|
6
|
| Xbench - CPU |
243
|
243
|
241
|
242
|
| - Threading |
241
|
260
|
262
|
268
|
| - Memory |
340
|
335
|
333
|
385
|
| - Quartz Graphics |
300
|
342
|
348
|
295
|
| - OpenGL Graphics |
209
|
178
|
202
|
233
|
| - User Interface Graphics |
374
|
410
|
415
|
454
|
| - Hard Disk |
110
|
108
|
106
|
105
|
| "Real World" Tests: | ||||
| Startup (sec) |
45
|
38
|
40
|
35
|
| Shutdown (sec) |
9
|
9
|
10
|
8
|
| Launch Classic Mode (sec) |
9
|
9
|
10
|
11
|
| Dup. 700 MB Folder 3600 Files (sec) |
62
|
56
|
66
|
58
|
| Scroll 142p Acrobat Reader File (sec) |
13
|
12
|
11
|
12
|
| Encode MP3 in iTunes (sec) |
54
|
54
|
54
|
52
|
| Export QT FQDV Movie in iMovie (sec) |
20
|
19
|
20
|
22
|
| Quake III Arena 1.32 (frames/sec) |
417
|
407
|
411
|
405
|
|
Halo 1.0.5.3/1.5.1 (frames/sec)
|
57**
|
61***
|
63***
|
67***
|
|
Call of Duty 1.4 (frames/sec)
|
n/a
|
84
|
85
|
86
|
For the now next-to-latest Power Mac G5, there are the same performance improvements related to the Finder and OpenGL in Xbench. In addition, memory is faster and startup is faster. Curiously, Quartz Graphics is measured as slower by Xbench, about the level in 10.3.5. Of the 3D games tested, only Halo shows that it may be benefitting from the faster OpenGL graphics. Halo was consistently faster but not by much. Perhaps, the updated Nvidia graphics drivers (v1.40) in 10.4 need more tweaking.
Here is a table comparing Tiger (10.4) vs Panther (10.3.9) using the 20" iMac G5, Mac mini and 17" PowerBook G4:
| Test |
20" iMac G5**
|
Mac mini
|
PowerBook G4
|
|||
|---|---|---|---|---|---|---|
| Mac OS X Version: |
10.3.9
|
10.4
|
10.3.9
|
10.4
|
10.3.9
|
10.4
|
| Benchmarks: | ||||||
| Altivec Fractal (GFLOPS) |
2.6**
|
2.4**
|
4.5
|
4.5
|
5.7
|
5.7
|
|
Cinebench 2003 - Rendering
|
242
|
246
|
116
|
115
|
145
|
150
|
|
- OpenGL Hardware Lighting
|
680
|
936
|
444
|
442
|
854
|
857
|
| Let1kWindowsBloom 1.0 (sec) |
18
|
12
|
33
|
12
|
36
|
15
|
| Xbench - CPU |
88
|
114
|
150
|
152
|
197
|
194
|
| - Threading |
70
|
97
|
110
|
110
|
142
|
140
|
| - Memory |
154
|
240
|
136
|
132
|
128
|
128
|
| - Quartz Graphics |
171
|
202
|
160
|
132
|
190
|
160
|
| - OpenGL Graphics |
169
|
173
|
108
|
123
|
134
|
147
|
| - User Interface Graphics |
204
|
314
|
217
|
233
|
232
|
263
|
| - Hard Disk |
83
|
85
|
56
|
55
|
76
|
77
|
| "Real World" Tests: | ||||||
| Startup (sec) |
49
|
38
|
53
|
48
|
57
|
49
|
| Shutdown (sec) |
9
|
7
|
15
|
11
|
-
|
-
|
| Launch Classic Mode (sec) |
-
|
-
|
13
|
15
|
13
|
16
|
| Dup. 700 MB Folder 3600 Files (sec) |
84**
|
80**
|
102
|
103
|
100
|
74
|
| Scroll 142p Acrobat Reader File (sec) |
99**
|
105**
|
26
|
30
|
22
|
23
|
| Encode MP3 in iTunes (sec) |
-
|
-
|
100
|
98
|
66
|
64
|
| Export QT FQDV Movie in iMovie (sec) |
-
|
-
|
44
|
37
|
37
|
32
|
| Quake III Arena 1.32 (frames/sec) |
-
|
-
|
89
|
88
|
112
|
104
|
|
Halo 1.0.5.3/1.5.1 (frames/sec)
|
-
|
-
|
-
|
-
|
-
|
-
|
|
Call of Duty 1.4 (frames/sec)
|
-
|
-
|
26
|
26
|
41
|
40
|
The results from the additional computers are very similar to our usual test-bed computers. The G5s are faster in more categories than the G4s. The key Finder-related tests show significant improvements in all computers. Startup is significantly faster in all computers tested.
We didn't expect a down side but Quartz graphics scores are significantly lower for three computers. Perhaps, the Xbench test needs a revision for changes in Tiger's Core Graphics.
[Update 5/4/05--We discovered that Tiger's Quartz 2D Extreme technology which employs the graphics card's processor is disabled by default. Hard to believe but true. We enabled it using Quartz Debug and re-ran the Xbench Quartz and User Interface tests for the three slowed Macs. The Mac mini is not supported but enabling Tiger's Quartz 2D Extreme on the Power Mac G5 and PowerBook G4 increased the scores drastically. See our note with results on our 5/4/05 news page.]
After improving over the previous versions, Classic mode in Tiger also takes longer to launch in all computers testing this function and significantly so in most.
Curiously, Quake III Arena runs much slower on the PowerBook with Tiger than Panther. We ran Quake many times, checked our settings, tried another drive using 10.3.9 and 10.4 but the results were the same. Quake runs on OpenGL which tested faster using Xbench so the results are confusing.
In summary, Mac OS X 10.4 Tiger is noticeably faster when using Finder operations, particularly with G5 Macs. Fortunately, the Finder speed improvements also extend to our older G4 Cube. We have encountered some wireless networking glitches, both with AirPort and Bluetooth as carried in our notes posted Saturday and Sunday, but otherwise Tiger is a terrific new Mac OS X. [Bill Fox]
Saw something? Send a tip
The archive ran on reader tips. What did you see, where, and do you want the credit?