Hands-On Speed Report--Mac OS X 10.4.7, How Fast Is It?
Last Tuesday, Apple released Mac OS X 10.4.7. An updater was released for the client and server editions of Mac OS X 10.4 "Tiger" and for PowerPC-based and Intel-based Macs. We provided an initial report and later noted that Apple had released a revised 10.4.7 (Intel) delta updater. Since then, we have used 10.4.7 on a variety of Intel-based and PowerPC-based Macs (MacBook Pro, iMac Core Duo, iMac G5, Power Mac G4 Cube, PowerBook G4s) without encountering any of the problems noted on other Mac web sites. Of course, we do not have all of the peripheral hardware and software cited as problematic but we have tried all of the Web sites and many of the applications without finding a problem. Just to be sure, we re-updated our MacBook Pro with the revised Intel delta updater, yielding a build number of 8J2135a.
Mac OS X 10.4.7 has been trouble-free for us but has it provided any speed benefits? It seems like it, so we ran our usual series of speed tests on our 2.16GHz 15" MacBook Pro with 2GB RAM and our 2GHz 20" iMac G5 with 512MB RAM. Due to variability in the results between replicates of the tests, we only consider a difference of over 10 percent to be worth noting.
Here are the results for our PowerPC-based iMac G5:
| Test+ | Mac OS X | ||
|---|---|---|---|
|
10.4.7
|
10.4.6
|
Perf. Chg. %
|
|
| Benchmarks: | |||
| Altivec Fractal (GFLOPS) |
5.1
|
5.1
|
0
|
|
Cinebench 2003 (9.5.1) - Rendering
|
272
|
272
|
0
|
|
- OpenGL Hardware Lighting
|
1182
|
1197
|
-2
|
| Let1kWindowsBloom 1.0 (sec) |
9
|
9
|
0
|
| Xbench 1.2 - CPU |
80
|
88
|
-10
|
| - Threading |
60
|
60
|
0
|
| - Memory |
64
|
64
|
0
|
| - Quartz Graphics |
99
|
88
|
+12
|
| - OpenGL Graphics |
121
|
119
|
+2
|
| - User Interface Graphics |
65
|
65
|
0
|
| - Hard Disk |
88
|
90
|
-2
|
| "Real World" Tests: | |||
| Startup (sec) |
40
|
42
|
+5
|
| Shutdown (sec) |
8
|
8
|
0
|
| Launch Classic Mode (sec) |
13
|
13
|
0
|
| Dup. 500 MB Folder of Files (sec) |
65
|
61
|
-7
|
| Scroll 144p Adobe 7.0.7 Reader File (sec) |
47
|
39
|
-21
|
| Export QT FQDV in iMovie HD 5.0.2 (sec) |
19
|
22
|
+14
|
| Quake III Arena 1.32 (frames/sec) |
141
|
141
|
0
|
|
Halo 1.5.2 (frames/sec)
|
20
|
20
|
0
|
|
Call of Duty 1.5 (frames/sec)
|
57
|
56
|
+2
|
| Doom 3 1.3A MQ (frames/sec) |
32
|
32
|
0
|
|
Quake 4 MQ (frames/sec)
|
29
|
28
|
+4
|
Only the Quartz Graphics score from Xbench of all of the benchmark applications shows any improvement above 10 percent in speed--the others show mostly no change. In the real world tests, scrolling is much slower (which conflicts with the higher Quartz Graphics score) and exporting an iMovie as Full Quality DV is faster. Our take is that 10.4.7 did little or nothing for speed in a single CPU G5 PowerPC-based Mac.
Here are the results for our Intel-based MacBook Pro:
| Test | Mac OS X | ||
|---|---|---|---|
|
10.4.7
|
10.4.6
|
Perf. Chg. %
|
|
| Benchmarks: | |||
| Altivec Fractal (GFLOPS) |
4.1
|
4.1
|
0
|
|
Cinebench 2003 (9.5.1) - Rendering (2X)
|
628
|
622
|
+1
|
|
- OpenGL Hardware Lighting
|
2888
|
2475
|
+17
|
| Let1kWindowsBloom 1.0 (sec) |
15
|
15
|
0
|
| Xbench 1.2 - CPU |
80
|
76
|
+5
|
| - Threading |
212
|
211
|
+0
|
| - Memory |
122
|
111
|
+10
|
| - Quartz Graphics |
61
|
62
|
-2
|
| - OpenGL Graphics |
154
|
142
|
+8
|
| - User Interface Graphics |
nv
|
nv
|
-
|
| - Hard Disk |
36
|
36
|
0
|
| "Real World" Tests: | |||
| Startup (sec) |
28
|
22
|
-27
|
| Shutdown (sec) |
10
|
8
|
-25
|
| Dup. 500 MB Folder of Files (sec) |
33
|
34
|
+3
|
| Scroll 144p Adobe 7.0.7 Reader File (sec) |
68
|
69
|
+1
|
| Export QT FQDV in iMovie HD 6.0.2 (sec) |
47
|
46
|
-2
|
| Quake III Arena 1.32 PPC/Intel(frames/sec) |
135/426+
|
134/nr
|
+1
|
|
Halo 1.5.2 (frames/sec)
|
25
|
22
|
+14
|
| Call of Duty 1.5 (frames/sec) |
35
|
32
|
+9
|
| Doom 3 1.3A MQ (frames/sec) |
62
|
59
|
+5
|
| Quake 4 MQ (frames/sec) |
66
|
56
|
+18
|
There is a very significant improvement in the speed of the OpenGL Hardware Lighting test of Cinebench 9.5 but the OpenGL test in Xbench, while positive, does not exceed our 10 percent threshold. All of the 3D games use OpenGL drivers and all show some improvement with Halo (running on Rosetta) and Quake 4 (a Universal application) showing gains above the 10 percent threshold. For some reason, startup and shutdown are much slower than previously under 10.4.6 but startup is still much faster than any other PowerPC-based Mac that we've tested. We'll be checking our Mac mini Core Duo to see if there is a corresponding decrease in startup and shutdown performance.
In summary, our tests show that Intel-based Macs may show significant performance changes while PowerPC-based Macs may not. Intel-based Macs with ATI graphics may get better OpenGL performance. They may also suffer a bit in startup performance. We'll revisit these conclusions when we've tested our Mac mini Core Duo and iMac Core Duo and have received some reader feedback [Revision: See Supplement posted 7/6/06]. [Bill Fox]
Saw something? Send a tip
The archive ran on reader tips. What did you see, where, and do you want the credit?