PowerBook G4/800 2002 Model Tests: For my first impressions on the new 2002 PowerBook G4/800 (w/7455 based CPU, 1MB L3 cache, Radeon 7500 Mobility w/32MB vram), including some performance test comparisons in iMovie2, Photoshop 7 and 5.5, 2D scrolling, Quake3 performance, etc., see this article.
(original Oct. 2001 article comments follow)
This article is just to show the results of a few simple/quick tests of the PowerBook G4/667 w/Mobility Radeon running OS X 10.1 vs a Pentium III 850Mhz PC notebook w/GeForce2GO 16MB/SDR running Windows XP. (I was called over to troubleshoot a wireless networking problem with the PC notebook, so was able to run some tests for comparison to the PowerBook.)
Many readers had asked for a comparison like this so here it is...
System Details: Both systems had all the updates available as of 10/26/2001 applied. (OS X 5L14, plus the OS X 2D window compression trick applied, Windows XP had over 10MB of updates applied including GeForce2GO driver updates for XP.)
- 667MHz G4 w/on-chip 256MB L2 cache
- 512MB PC133 (CL3)
- 133Mhz system bus
- 30GB 4200 rpm drive (Toshiba model w/Fluid Bearings)
- Radeon Mobility w/16MB DDR Video RAM (166Mhz)
- OS X 10.1 (5L14)
- DVD ROM drive*
On 12/17/2001 Apple announced that a 8x8x8x24 Combo DVD/CDRW drive
is now standard on all new PowerBook G4s. (No upgrades for early buyers however.)
- 850MHz Pentium III w/on-chip 256MB L2 cache
- 384MB PC100 (CL3/CL2 mix)
- 100Mhz system bus
- 20GB 4200 rpm drive (Toshiba)
- GeForce2GO (1st gen) w/16MB SDR Video RAM (166mhz)
- Windows XP (home version w/all updates)
- Combo DVD/CDRW drive (4x4x4x24)
Quake3 1.30 Tests:
The latest version of Quake3 (1.30) was used on both machines.
Tests were run at 640x480 and 1024x768 with the standard Quake3 "Fastest", "Normal" and "High Quality" settings. (HQ settings are 32bit mode/32bit textures.) The ejecting brass, marks on walls, etc. game options were all left at their default settings (enabled).
As a FYI, the PB G4/400 delivered 47.4 (fastest), 39.3 (normal) and 18.6 (HQ) scores.
Now Tests at 1024x768 mode:
As a FYI, the PB G4/400 delivered 26.9 (fastest) and 18.7 (normal) - 1024x768/HQ (32bit) tests can't be run with its 8MB Video ram.
You can see where the SDR VRAM of the GeForce2GO and the Radeon Mobility both run out of breath at 1024x768/32bit with the HQ settings. Running 16-Bit mode at 1024x768 helps, but the ATI chips have poor image quality at 16bit modes. One advantage of the Geforce2GO; 16Bit mode doesn't suffer the dithering look. (The previous PB G4's 8MB video ram isn't sufficient to run 32bit Quake3 modes at 1024x768, but it would be too slow at those modes anyway.) GeForce2GO systems with DDR VRAM don't take a severe nosedive like this. And the latest 7500 Radeon Mobility chip is even faster than the GeForce2GO (in any config) according to reports last month. (The 7500 Mobility runs at up to 270MHz clock speeds and includes hardware T&L, which the Radeon Mobility lacks. Hopefully the next PowerBook G4 will use this chip.)
I found it odd that although it was not repeatable, after exiting Quake3 on the PC notebook Windows XP locked solid before completing the desktop refresh. (Cursor froze, desktop partially drawn - had to power off as the system was totally frozen.) This was after applying the Windowsupdate listed Nvidia GeForce2GO drivers for WinXP. (That was one of about 13 updates listed at Microsoft's update page after the initial install of XP.) Rebooting and testing again didn't repeat the lockup (however it did freeze twice during troubleshooting a wireless networking problem.) I still think XP looks to be a much better OS than WindowsME that was originally installed on the Toshiba. However I didn't expect to see a system lockup that quickly (may be due to the fact it was an update, not a clean install).
Photoshop 5.5 21 Filter (PSBench) Tests:
The recommended settings (1024x768/32bit, etc.) settings were used for both machines. Since there's no OS X native version of Photoshop yet for the Mac, PB G4 was booted into OS 9.2.1 for this test. Each of the 21 filters is run 3 times and the average recorded.
As you can see, some Photoshop filters support the SIMD extensions of the Pentium III chip as well as the G4's Altivec instruction set. (As a FYI - the PB G4/400 took 119.3 seconds for this test.)
You can download the latest PSBench action script at http://www.geocities.com/Paris/Cafe/4363/index.html.
Although Windows XP (and ME) come with a Movie Maker application, it pales in comparison to iMovie in my experience. Even my PB G4/400 works very well with iMovie, but I saw dropped frames with a much faster (clock speed) PC notebook using Movie Maker capturing video from my DV camcorder in ME. iMovie also had a much superior interface - totally intuitive and very productive. (I have not tested WinXP's movie maker, but comments I've seen from some media sources note it's very limited also.)
With no registry, much fewer security and virus threats, I'd choose the PowerBook G4 over the Toshiba, despite it's better 3D game performance. The Toshiba also has very poor battery life - about 2 hours max and weights almost 8 LBs. However it does have the most impressive speakers of any notebook I have ever seen. (No other notebook built-in speakers are even close to its sound quality.) It's baffling however that the Toshiba has no way to adjust the screen brightness other than via a Control Panel applet (there's a myriad of settings you can adjust for different power profiles, even varying the LCD brightness depending on battery levels). However the applet is not WinXP compatible - so there's no way to adjust the brightness in WinXP at least until Toshiba updates its software. Toshiba should have included keyboard function keys for brightness.
Summary: (Note: the list below was created before 12/17/01, when
Apple announced that all PB G4s would have a Combo DVD/CDRW drive standard.)
PowerBook G4 Advantages:
- Larger and Superior Display (higher res/better quality/wider)
- Powered Firewire ports
(on PC notebooks, even with portable FW drives you have to
carry around an AC adapter for drives)
- Far longer battery life (about twice as long)
(easier to change and cheaper batteries also)
- Gigabit Ethernet (vs. 10/100 on the Toshiba)
- Easier to use interface
- Much Better for DV/Movie use
- Lighter and thinner
- Looks like a million bucks
- No registry and less OS security/virus threats
Toshiba 2805-S402 Advantages:
- Better 3D game performance
- Best Audio (speakers) of any notebook
(including tiny "subwoofer")
- Quieter fan (moves a huge amount of air quietly)
- Combo 4x4 DVD/CDRW drive standard
- More rugged than PowerBook G4
- Better keyboard feel w/full size arrow keys
- Two PCCard slots (although if you want wireless networking one is used
by a PCcard)
- Smartmedia card slot (granted most will never use this)
Neither the PB or the Toshiba have any expansion bays. Both support external monitor (VGA) used simultaneously with internal LCD.
As another PB G4/667 (2001) user noted, the fan seems to come on very frequently. Also the DVD ROM drive doesn't feed CDs as well as my PowerBook G4/400. (Repeatedly I have had to literally push the CD completely in the drive rather than part way, as it stalls trying to draw the CD into the drive.)
Other PowerBook G4/667 Benchmarks/Comparisons:
I ran my standard iMovie2 test (6 clips from the Tutorial, no transitions, time to export with "CDROM Medium" settings.)
iMovie2 OS X Export:
PB G4/667 (2001) = 107.7 seconds
PB G4/400 = 157 seconds
iMovie2 OS 9 Export:
PB G4/667 (2001) = 112.4 seconds
PB G4/400 = 186 seconds
Here's the results of ATTO's benchmark on the 30GB/4200 RPM Toshiba OEM drive. It appears that there's literally no write caching seen with the Toshiba (fluid bearing) 30GB drive in this PB G4/667 (2001). (Note the peak write rates vs. the PB G4/400), although sustained rates are higher than previous PB G4 models (about 3MB/sec faster).
Atto Tools HD Benchmark
- Peak Read:
PB G4/667 (2001) w/Toshiba 30GB = 60.21 MB/sec
PB G4/400 w/IBM 20GB = 59.31 MB/sec
- Peak Write:
PB G4/667 (2001) w/Toshiba 30GB = 20.27 MB/sec
PB G4/400 w/IBM 20GB = 56.10 MB/sec
- Sustained Read:
PB G4/667 (2001) w/Toshiba 30GB = 20.15 MB/sec
PB G4/400 w/IBM 20GB = 16.48 MB/sec
- Sustained Write:
PB G4/667 (2001) w/Toshiba 30GB = 19.75 MB/sec
PB G4/400 w/IBM 20GB = 15.94 MB/sec
As a FYI, I ran ATTO tests on my portable Oxford911 Firewire case with 48GB (5411 rpm) drive inside (using the latest firmware update for the Oxford911 bridge from OWC). Peak rates were much higher than the previous PB G4, but sustained rates were not that much better. Remember notebook 2.5" drives can only *sustain* about 20MB/sec or less (often 16MB/sec) rates even in pure benchmarks like ATTO's. Real world file copy tests showed under 10MB/sec effective rates with a 102MB .sit file (the same file used in real world copy rates in previous FW drive articles here). AC Powered (3.5" desktop drive based) Firewire drives perform better of course. They have sustained rates in ATTO's pure benchmark in the 30MB/sec range often times. (Although most real world file copy, etc. rates are far lower - shown in past articles here.)
Cinebench's 3D benchmark results; includes a reader's PB G4/500 comparison scores. (For an explanation of this benchmark, see this page.)
- Software Shading:
PB G4/667 (2001) = 5.22
PB G4/500 = 5.42
PB G4/400 = 4.51
- OpenGL Shading:
PB G4/667 (2001) = 6.98
PB G4/500 = 6.50
PB G4/400 = 5.43
PB G4/667 (2001) = 7.99
PB G4/500 = 6.64
PB G4/400 = 5.35
Throughput benchmark results (includes a reader's PB G4/500 comparison scores). As shown in Throughput results in previous articles like the G4/733 vs Dual G4/500 comparison, the later UniN chipset, faster CPUs/graphics cards can show much better rates in this "pure" benchmark. (And even driver versions can affect scores, esp. the "CopyBits" score.)
Throughput 1.5 Scores
PB G4/667 (2001) = 166.3 MB/sec
PB G4/500 = 50.1 MB/sec
PB G4/400 = 50.1 MB/sec
PB G4/667 (2001) = 177.3 MB/sec
PB G4/500 = 83.3 MB/sec
PB G4/400 = 83.2 MB/sec
PB G4/667 (2001) = 196.4 MB/sec
PB G4/500 = 125.7 MB/sec
PB G4/400 = 125.6 MB/sec
PB G4/667 (2001) = 178.2 MB/sec
PB G4/500 = 170 MB/sec
PB G4/400 = 169.1 MB/sec
(Note: The 4.2.9 firmware update for the PowerBook G4/667 did not affect these scores; literally the same rates were reported running this test under OS 9.2.2 with the firmware update applied.)
Although it's a hotter CPU, I wish the new Powerbooks had used a 7410 G4 (as before) instead of the 7440/7450, which hasn't really delivered impressive performance compared to the 7400/7410 G4 CPUs in previous owner tests. (Even the G4/533 systems did well compared to a G4/733 system that had the 1MB L3 cache). Even with the 133mhz bus, the PB G4/667 (2001) has much lower reported memory bandwidth than my PowerBook G4/400 with 100Mhz system bus. (Appx. 240MB/sec with my PB G4/400, only 145MB/sec reported with the PB G4/667 (2001) - if you wait long enough rates will creep up to 158MB/sec or so however, but still far below the 7410 based PB G4/400 even with its 33mhz slower system/memory bus.) [Update - see below for another report on this and how it affects Logic audio performance, where the PB G4/500 significantly outperforms the PB G4/667 (2001).] This was also seen with the desktop 7450 G4 systems. (All the arguements on compiler optimizations, etc. have been hashed over and over in the forums here months ago.) The rev 2.1 CPU in the PB G4/667 (2001) doesn't show any better bandwidth than the 2.0 original 7450 chips. Granted tests like GaugePro vary a lot at times (I once saw 205MB/sec with VM on, but it was not repeatable and GaugePro doesn't recognize the 7450/7440 CPU types or clock speeds correctly.)
The G4/667 is faster at applications than my G4/400 however, but not dramatically noticeable in use ("seat of the pants" tests). Benchmarks often show more difference than is noticeable in actual use. But then my PB G4/400 has 1GB of RAM and a 5400 rpm hard drive so it's well equipped. The Radeon Mobility is a plus over the Rage128 Mobility for gaming, but as you can see above, for 3D gaming you're much better off with a desktop if you want to run 1024x768/32bit modes (or higher). Granted less demanding games than Q3 and UT run fine on the PowerBooks. Although the 16MB vram allows 32bit modes at 1024x768, it's not the best mode for survival in actual gameplay. (30FPS sounds good to the inexperienced, but in actual play framerates drop below that often times and there's response lag that makes 1024x768/32bit mode not responsive enough for best play.) Even a B&W G3 w/PCI Radeon card delivers better performance and G4/AGP desktops have much better performance than that. (Some readers wrote to say they bought a PS2 console for gaming since it had a much wider selection of games than the Mac and had better controllers, etc.)
The new Powerbooks do have the later UniN chipset revision (same as the 2001 iBook rev) which delivers desktop level Firewire drive performance, but sustained rates with portable drives are limited by the actual drive itself. (As I've noted and shown in past Firewire drive tests however, even the 16MB/sec (appx) sustained write rates with the original PB G4s was as fast as the internal hard drives and more than adequate for most needs including DV capture, etc. Real world file copy rates are usually well below the 16MB/sec limit - see real world effective rate tests (vs pure benchmarks) shown in this article.)
Very nice Powerbook overall, although the fan seems to run very frequently, at least when used on AC power.
Memory Bandwidth and Logic Audio Tests: (from the 12/14/2001 www.xlr8yourmac.com news page.)
Regular readers here know that since the first 7450 G4 based Macs were released there's been comments on much much lower than expected memory bandwidth compared to the G4 7400/7410 CPUs. (The 7410/7400 G4s had dramatically higher bandwidth - nearly 100MB/sec more in some tests.) A first report on the 2.1 stepping 7450 seemed to show this was fixed, but tests I've run since then show the rates are still much lower even with 2.1 rev 7450s compared to external L2 cache/shorter pipeline 7400/7410s.
(There have been discussions on this many times since the beginning of the year about compiler optimizations, etc. which is too long to discuss here). My comments above noted my PB G4/400 had much higher memory bandwidth than the PB G4/667 (2001).
A reader sent an email tonight on why the much lower memory bandwidth of the new PowerBook G4/550 and G4/667 models (compared to the PB G4/400 and /500 models which use the 7410 CPU) make the earlier Powerbooks a better choice for Logic Audio.
(Update: I've added a reader's results with the 2002 PB G4/800 to the listing below. For other PB G4/800 2002 model tests, see this article.)
Hi - though you might like to have this info - I and a few others
have been doing some testing on the old Powerbook G4/500 vs the new
550 & 667.
Memory bandwidth on the 550/667 is a *lot* lower than on the 500 -
the new 7440 CPU is to blame, not the amount of cache. The older 7410
CPU is better - we've measured memory bandwidth at 230MB/sec on the
400/500, vs 145MB/sec on the 550/667.
[Again this is not news to regular readers here and was noted in my earlier comments above comparing the PB G4/667 (2001) to my PB G4/400. It's good to see an example of where this has an
actual applications impact.-Mike]
What this means is that for audio use, where many simultaneous
realtime DSP processes are required, the old PBG4/500 will perform
significantly *better* in many operations than either the 550 or the
A simple test using simultaneous stereo Platinumverbs in Logic Audio 4.7.3:
- PB G4/667 (2001) - 9 platinumverbs
- PB G4/550 - 7 platinumverbs
- PB G4/500 - 13(!) platinumverbs
- PB G4/800 (2002) - 18 platinumverbs
(He later said his Umax S900 w/G4@473Mhz managed 6 Platinumverbs on 67MB/sec memory bandwidth.-Mike)
We didn't test the PBG4/400, but it's likely that will also
outperform the 550 and perhaps even the 667... Memory bandwidth is
While end-of-stock, secondhand, or refurbished PBG4/500 and PBG4/400s
are available, they are a *much* better choice for audio work than
the new 550 or 667... The 400/500 also have the advantage of allowing
the processor speed to be reduced which increases battery life &
eliminates fan noise - the 550/667 do not allow this I believe.
I didn't spot the lack of lower speed CPU options on the new PowerBooks - but checking the Energy Saver control panel shows that feature isn't there. (The old "processor cycling" option is, but that's not the same option as the lower speed CPU one seen on the original PB G4s.)
The new PowerBooks do have a faster graphics chip/more Video RAM (DDR/16MB) and other subtle improvements and do perform faster in many apps than previous models, but the lower memory bandwidth can't be argued as only a "benchmark" issue anymore based on the above tests. (If any other audio apps users have comments or test results please let me know.)
List of Changes/Improvements in new PowerBooks:
- Radeon Mobility w/16MB
(much better 3d gaming performance than 8MB Rage128 Mobility, but still has limits)
- UniN chipset rev 17 (same as 2001 iBook rev) improves Firewire max rates to that of desktop levels. [Note - For portable firewire drives, even Oxford911 bridge based ones remember the notebook drives used in those cases can't *sustain* rates over 20MB/sec, regardless of the Mac model used. With a 1/3 full Travelstar 48GB portable drive (5411 RPM) in a Oct. 13th firmware updated Oxford911 case I saw _sustained_ rates of only appx 16MB/sec - although _peak_ rates were much higher (cache effect.) Again I think for most users the firewire speed issue is a moot point, since the rates even with the original PB G4 were as good (sustained) as the internal drives and more than fast enough for DV capture, etc. I captured DV camcorder movies in iMovie2 with my PB G4/400 fine even with IE 5 in the background.]
- AC adapter smaller, with lit ring at the plug which shows if the battery is charging (amber) or fully charged (green). The adapter does get warmer than the previous Yo-Yo model however.
- 133Mhz bus (on 667mhz model only)
(although memory bandwidth tests show much lower rates than
even the PB G4/400, due to the 7440/7450 CPU chip change.)
- Slightly faster internal drive benchmark results (although peak writes were much lower as noted above with the 30GB drive.)
Not sure if this is due to the revised UniN chip or the later IBM travelstar drive model.
- The rear port cover had slots in it for the fan (the G4/667 seems to run the fan very frequently, much more frequently than my PB G4/400.)
- A Command Key on the right side of the space bar,
instead of the Option key on the original PB G4 models.
- A reader with a PB G4/500 noted his 512MB SODIMMs were only recognized as 256MB each in his new PB G4/550. (Apple's developer docs note that even the 100mhz bus 550 model requires PC133 SODIMMs.)
Although the antenna lens is thicker, I did not see significant Airport range improvements over the original PB G4 models. (It's still noticeably less range than the PB G3 series, due to the metal case I suspect.)
PowerBook G4/550 and G4/667 Firmware Update: (from the 12/21/2001 www.xlr8yourmac.com news page)
Owners of PowerBook G4/550 and 667MHz models will see a new firmware update (4.2.9) if you run Software Updates (in OS 9). If you'd rather download the file separately, see this Apple page which notes the following changes:
"About the PowerBook Firmware Update 4.2.9
The PowerBook Firmware Update 4.2.9 will only run on 550MHz or 667MHz PowerBook G4 computers running Mac OS 9.1 or later from a local drive. If you are using Mac OS X you must boot from a local Mac OS 9.1 writeable partition (not a CD, or network disk) prior to following the update instructions.
Firmware Update 4.2.9 includes improvements to Firewire target disk mode when running Mac OS X, and improves starting up on systems with 1gigabyte (GB) RAM and enables Netbooting over gigabit Ethernet.
Before you install PowerBook Firmware Update 4.2.9, read the instructions below. You may want to print them for reference.
See the linked page for the installation instructions. Remember firmware updates are not reversable so if you're not having problems now you may want to wait for reader reports on any issues seen. (If there are any - but in the past with some firmware updates there have been some.)
Other PowerBook Related Articles:
See the PowerBook section of the Systems page for other PowerBook performance tests, reviews and guides on installing hard drives. Rob Art also has posted tests with PB G4/550 vs a PB G4/500.