Test System/Drive Info:
- G4/450 Cube, 384MB RAM, with OS X 10.0.4/OS 9.1 installed. (G4 Firmware update 4.1.8 had been applied.) Internal drive is the original equipment Western Digital 20GB 5400 RPM/2MB Cache
Drive was 35% full (12GB free) and not optimized. (I later ran speed disk to optimize the drive and noted the difference it made below the graphs.)
(WD drive specs here)
- Other World Computing "Elite" (Oxford 911 IDE/Firewire bridge) Case kit
with 60GB IBM 75GXP 7200 RPM/2MB cache IDE drive installed.
(IBM drive specs here)
- TransIntl Portable (Oxford 911 IDE/Firewire bridge) case kit
with IBM Travelstar 48GB 5411 RPM/2MB cache IDE drive installed.
(IBM drive specs here)
Intro: The purpose of this article is to compare Firewire performance in OS 9.1 vs. OS X (10.0.4) with both a portable Firewire drive and AC powered drive. (Both cases use the fast Oxford 911 bridge board and high performance drives of each type.) In tests of times to duplicate files, the original 20GB Cube's internal drive performance is also shown. In addition to Finder duplicate/file copy tests, some pure benchmarks were also used (tested under OS 9.1 since as no OS X versions of the benchmarks were available). I also show results of MacBench's Disk mix tests since they show random read and write performance, not just sequential I/O from file copy/duplicate tests.
Real world test results are shown first since they matter most. Pure benchmarks like ATTO (which I've used since 1997 in articles here) show much higher rates than typically occur in normal use. In any tests that use the internal system's hard drive (file copies to/from it), remember the internal drive can be a factor depending on its speed and state (how full it is/how fragmented it is). The main point was to show how each of the drives performed when run from the exact same system, since copying files from/to the firewire drive is something everyone would be doing in actual use. As you will see from the results, often the differences in real world tests between the drives are much smaller than what you'd expect if you only looked at the pure benchmark results. (This fact is also shown in many past drive articles here. The same is true of any drive - IDE, Firewire or SCSI.)
The drives were bootable on G4/AGP and later Macs (including PB G3 firewire and PB G4). Comments on this, movie/DV performance and a strange Firewire problem I saw near the end of testing is listed at the bottom of this page.
G4/450 Cube Tests:
(Other systems will be tested in OS X and OS 9.1 with these drives as time allows, after I return from MWNY.)
Real World Tests:
Write 102.3MB File TO FW Drive:
Time to copy a single 102.3MB file to the Firewire drive.
The effective MB/sec rate in real world tests like this are far lower than the pure benchmark results such as ATTO's Benchmark shown below. (Copying lots of files reduces speeds even more, due to more file I/O overhead.)
Read 102.3MB File FROM FW Drive:
Time to copy a single 102.3MB file from the Firewire drive to the G4 Cube's internal IDE drive.
Duplicate 102.3MB File:
Note: After optimizing the Cube's 20GB HD with SpeedDisk, the time to duplicate this file was 13.15 seconds in OS 9.1 and 16.3 seconds in OS X.
Duplicate Folder (Thousands of files):
Note: After optimizing the Cube's 20GB HD with SpeedDisk, the time to duplicate this file was 60.5 seconds in OS 9.1 and 68.1 seconds in OS X.
Notice although Firewire to internal HD copy performance was faster in OS X, when duplicating a single file it was slower than OS 9.1. (Duplicating thousands of files in a folder was faster in OS X than OS 9.1 however.) OS X's copy dialog by default shows the file-by-file status, which adds some additional work basically for the system to display/update compared to the default progress bar only display in OS 9.1. If that were the reason for the lower single file duplicate performance however, then OS X should have been much slower when duplicating a folder with thousands of files - but it wasn't as shown below. (You can expand OS 9's dialog to show file-by-file copy status, but in OS X if you collapse the details you lose the progress bar also, which I needed for the stopwatch timing.)
Copy Folder with Thousands of Files FROM Firewire Drive:
Real world read peformance test. Copy the folder from the Firewire drive to the internal hard drive. (Unlike single file copy rates, there's more overhead with thousands of files so the net rates are far lower.)
OS X again shows much better performance than OS 9.1. (I still don't understand why single file duplicate times were slower in OS X than OS 9.1.)
Copy Folder with Thousands of Files TO Firewire Drive:
Real world write peformance test. Same test folder, but copied to the FW drive from the internal hard drive.
Benchmark Tests: Although MacBench 5 is no longer available and its CPU benchmark never supported dual processors, or Altivec and was very influenced by L2 cache, its disk test is one of the few on the Mac that do Random Reads and Writes, which is useful information in my opinion. (Most disk tests are simply sequential I/O.)
Remember the 48GB FW drive is a portable/2.5" drive, the 60GB drive is a 3.5"/AC powered drive. (Both cases have the Oxford911 bridge.)
MacBench5 Disk Tests: Graph compares the Cube's internal 20GB drive to the portable and AC powered Oxford911 bridge firewire cases.
ATTO Tools 2.5 Benchmark:
Rates in MegaBytes per Second (longer bars are faster).
One thing not shown in a graph of just the sustained/peak numbers is the graph across I/O sizes. I'm including that to show how the AC powered drive shows less spikes. (see below)
VST 2.3 Benchmark:
The VST Format utility benchmark requires the VST extension to be enabled. This is the only test ran where the VST extension was used. In the graph below, longer bars are faster for all tests *except* the Access times (in milliseconds), where lower scores are better. Other rates are in MegaBytes/second. (This utility will only test firewire drives.)
Note: When running this benchmark on the OWC Elite case the VST software popped up a dialog that the drive needed an update. Since the OWC drive was formatted with Apple drivers and has a different bridge board than the VST drives, I cancelled the update dialog.
Booting from Firewire Drives: As noted in my past articles, the drives tested here were bootable with G4/AGP and later Macs. (Including the PowerBook G3 firewire series and PowerBook G4.) Older Macs and Macs using PCI Firewire cards may not be able to boot from Firewire drives. As mentioned in the FAQ here (and at Apple's site), the B&W G3 and Yikes G4/PCI are the only Macs with onboard Firewire that are not bootable from Firewire drives. (And to prove this I was not able to boot from a Firewire drive with my B&W G3 (a rev 2) onboard FW ports.) The internal drive with OS 9.1/Startup Disk Control Panel 9.2.1 showed the Firewire drive grayed out. The Option key at startup tip that works on G4/AGP and later Macs does not work with the Yikes G4/PCI, B&W G3 (or earlier) machines.
Movie Playback from Firewire Drives: Every one of these Firewire drives (and previously tested models) had no dropped frames playing back Quicktime movies or DV movies. Some codecs like Sorenson are very CPU intensive, so slower CPU speed or older Macs might have dropped frames due to that issue and not the drive however. (Scaling the high-res QT Star Wars Phantom Menace movie to full screen at 1024x768/millions colors played with no dropped frames on all the modern Macs here from Firewire drives.)
DV Capture: Capture using iMovie had no apparent dropped frames (using my Panasonic 910 camcorder). I don't currently have access to Final Cut Pro to test. (Some past reports in the main site news noted there were dropped frame issues with FCP related to Quicktime 5.01. Initial reports seem to indicate QT 5.02 may have solved that.) As noted in a past article, although DV is under 4MB/sec rates, Digital Origin (makers of EditDV, MotoDV, etc.) that notes a drive with an avg. sustained rate of 5MB/sec or faster is required. (This provides some margin over the DV rate.) Even Firewire drives with older IDE/Firewire bridges can usually sustain those rates, given there is no other system or software/driver issues. As noted previously, I captured DV from iMovie to my older VST 12GB drive from a PB G4/400 with the camera daisy-chained to the VST drive (since the PB G4 has only one FW port) with no dropped frames - even with IE 5 in the background. That PB has 512MB of RAM and VM off which is a plus.
I also wanted to mention a strange Firewire problem I saw near the end of testing. I had used both these drives with several Macs for days before starting the review tests and didn't see any problems. (I like to use products for several days rather than just run simple tests on them, just to see if any problems occur in normal use.) However near the end of the tests on the Cube I saw a very strange (and disturbing) problem.
While copying a file on the portable drive the Finder popped up an alert that a Firewire drive had been disconnected. (The drive was still connected of course and the bus power LED was still lit.) I had to force a restart at that point. After this no Firewire drive would mount on the Cube - not even the AC powered drive. Both drives mounted fine on other Macs here. Disk First Aid said they were OK. I ran Disk First Aid on the Cube - no problems found. Apple System Profiler said no Firewire devices were detected even when they were attached. I booted from an OS 9.1 CD (which has Firewire extensions enabled) and still no Firewire drives would mount. Booting into OS X with a FW drive connected resulted a system error (command line text error message overwrote the GUI screen). I had never seen anything like this happen before.
Remembering a friend's recent comments on 4 new Macs that had Firewire ports suddenly die after using portable drives (twice for each system, even after having their logic boards replaced) made me wonder if this was a hardware failure.
Zapping the PRAM did not help. Apple's Hardware Test CD showed all Cube components passed (RAM, Logic board, Video, etc.). At this point things got even worse, as any attempt to connect a Firewire drive would hang the system. Restarts then without any Firewire drives attached would hang at the start of the "Welcome to MacOS" screen (even when booting from CD and booting with extensions disabled). Holding the reset key in for several seconds solved the hang at boot problem but still Firewire was not working, and any attempts to use them caused a repeat of the hang at boot problem again even on restarts after the drives were disconnected. Disk First Aid said the drive was OK, and the Apple Hardware Test CD again showed all tests passed. I was almost certain I had a hardware problem now. I removed the Cube's internals from the housing and checked for any obvious problems, but only saw that the IDE cable was not completely seated to the motherboard (by just a hair, not enough to cause any problems). I then put it back together, pressed the reset button again for a few seconds and magically Firewire started working again. I still don't know what caused the problem and wonder if it will return.
Although the portable drive mounted fine on another Mac, after the repeated attempts to connect the portable drive to the Cube during this episode, Norton Utilities found "serious" data errors on the drive (not just the usual bundle bit/creation dates NU often reports). I've used the portable drive again with the Cube and it seems fine so far, but it's something I wanted to mention. I'm still not sure what the culprit of the problem was - the Cube or the drive.
OS 9.1 vs OS X Firewire Performance:
- OS X was consistently and significantly faster at Firewire file copies to and
from a Firewire drive than OS 9.1
- For duplicating a single file, OS 9.1 was faster, at least with this Cube system. However OS X was faster at duplicating a folder with thousands of files.
AC Powered/3.5" FW Drive vs Portable/2.5" Drive Performance:
- With both FW cases using the same Oxford 911 bridge and high performace
drives, real world performance was closer than you might expect, despite
the fact the 3.5" desktop drives have better performance specs, access times
and much higher sustained rates in benchmarks.
- Keep in mind the cost of the notebook drives are higher per GB than desktop drives (often you can get twice the GB in a Desktop/3.5in drive as a 2.5in/Notebook drive for the same cost.) But if portability and small size are major factors for you, then the price difference may be worth it. Check sponsors Transintl.com and OWC for current pricing and models.
- Although not shown in this review, previous articles covered the issue of lower Firewire
performance on early Powerbooks G4s. (The fall 2001 Powerbook G4 has a later revision (rev 11) of the Uni-N chip that
improves Firewire performance to desktop levels however. (The dual USB 2001 and later iBooks don't use Uni-N but have improved FW port performance also.) Real world impact however is
not as pronounced as the benchmarks would lead you to believe, since many file I/O operations have lower net rates than the appx. 20MB/sec limits seen with benchmarks. And even the OEM internal drives of those Powerbook models can't sustain much more than the Firewire interface. Although the fast FW cases can't reach their full potential with the early Powerbooks, they're still more than adequate for most needs, if there are no other system/software issues present (such as some past reports on QT 5.01 and FCP 2).
For tests with Firewire drives with a PowerBook G4/400 and PowerBook G3/500 (real world and benchmarks) - see this previous review.
I've posted a step-by-step guides with photos on installing drives in the OWC Elite Case and the Transintl portable case as well as some other brands. There's also a previous performance comparison on the OWC Elite case to a Clubmac case and VST portable drive tested with 4 Macs (G4/500 DP, PowerBook G4/400, PowerBook G3/500 and 9600/350 w/PCI Firewire card).
For other Firewire reviews of drives and controllers, see the Firewire articles page.