As some of you may know, I’m relatively paranoid about backups—you can’t have too many, and you can’t make them often enough :). The macosxhints.com site is backed up twice a day via a set of scripts that use ssh and scp (and are scheduled via cron). For my personal machine, I use two external hard drives. The smaller of these two (an older version of this 250GB Maxtor drive) is used throughout the day to make backups of my key files. It also holds secondary copies of key things such as my iTunes music collection, iPhoto library, and digital video snippets. The larger of the two drives is a LaCie 500GB Triple Disk Extreme. At the end of each day, I run a full backup of the machine to the LaCie disk, and then power it down. But this article isn’t really about my backup strategy; it’s about the three interfaces on the Triple Disk Extreme (TDE), and a simplistic comparison of their performance on my machine (Dual 2.0GHz G5, first gen).
The TDE is so named due to its FireWire2, FireWire, and USB2 interfaces. A recent conversation with Chris Breen about FireWire vs. USB2 on the iPods led me to run a few tests on my hard drive, just to see how each interface performed. What got me started down this road is some stuff that Chris wrote in a couple of different iPod reviews:
In my tests, a dual-processor 2GHz Power Mac G5 filled a 6GB mini in 15 minutes and 17 seconds over USB 2.0. Using a FireWire connection shaved a scant 18 seconds off that time.
…
The nano is also quicker to sync than other iPods. I synced the same 903-track playlist on a 4GB nano and a 4GB iPod mini. It took 9 minutes and 15 seconds to sync the nano. The mini took nearly 7 minutes longer to sync, finally finishing the job in 16 minutes and 13 seconds.
So while USB 2.0 may not fare so well with other devices, as far as iPods go, syncing performance doesn’t appear to be a problem.
I thought I’d use my TDE to run a few tests in the Finder, just to see how things compared there. Read on for my results…
I didn’t have a ton of time for these tests, so the following results are based on only two runs with each interface. However, the results in each case were quite similar, so I don’t think there’s much to be gained from additional repititon. For all tests, the Finder was left as the foreground task, and no significant background tasks were running. There were, however, a number of open applications, but I kept these the same through all the testing.
My sample data was a 4.55GB (4,661.1796MB) folder with four sub-folders, holding about 80 different video clips in total of varying lengths—movie trailers, commercials, a bit of home recorded stuff, and a couple of music videos. I timed from the time I released the mouse button on the “drag” until the Finder’s progress dialog vanished. Since my timer was just the menubar clock (showing seconds), the following results may be off by a second or so. But as you’ll see, timing errors don’t drastically affect the outcome of my testing.
So without further ado, here are the results of my tests:
Interface | Time (min:sec) | MB per min | % of max |
---|---|---|---|
FireWire2 | 1:47 | 2,613.8 | 100% |
FireWire | 2:39 | 1,758.9 | 67.3% |
USB2 | 5:21 | 871.3 | 33.3% |
As you can see, the hands-down winner (no surprise) is FireWire2. Though nowhere near twice as fast as FireWire, the copy operation was still nearly a minute quicker. FireWire came in a respectable second, still moving over 1.7GB of data per minute. What surprised me, though (especially given Chris’ test results) was how poorly USB2 faired in my tests—less than one third the speed of FireWire2, and half again as slow as FireWire.
Note that the ‘% of max’ column above isn’t related to performance against the max speed of a given interface’s specifications; it’s just a measure of how the slower interfaces (FireWire and USB2) faired against the fastest (FireWire2).
Next I thought I’d see if the Finder was to blame for the poor USB2 performance. I hooked up our iPod nano, and then used the Finder to copy 3.57GB worth of data to it, which took 6:37. I then switched to iTunes and made a 3.57GB playlist, and copied that to the nano. In iTunes, this took 7:10. Given the difference in the number of files (there were more in the iTunes playlist than in the Finder selection), this is basically a ‘no difference’ result. So there’s no “Finder effect” that I can see to explain my slow USB2 performance; it’s just slow everywhere.
Nov 14th Update: Here’s the bit I took the story offline to modify, and sadly, it’s clearly not worth the multi-week delay it required!
Then I plugged in my third-gen iPod via FireWire, and copied the same 3.57GB playlist to it…and got my first real surprise of these tests. It took 7:04 to copy the songs to my iPod via FireWire—nearly the same amount of time as it did to copy them to the nano! Based on the test resuls from the Finder above, and even allowing for some iTunes overhead, I would have expected 3.57GB worth of data to copy in about three or four minutes, not seven!
In the original version of this article, I wrote that something was ‘clearly amiss’ with these slow results on the iPod. But that’s not the case; I simply overlooked the (very) obvious answer: the reason that Apple’s moved from FireWire to USB2 on the iPods has little to do with the superiority of one interface over the other. Instead, it’s because the hard drive in the iPods is slow enough that it’s the limiting factor, not the speed of the interface. Combine that with the fact that (a) PCs tend to have USB2, not FireWire, and (b) manufacturing costs drop if you only need support one interface, then Apple’s move to USB2 seems like an obvious and smart decision. With that said, I’m still curious has to how an iPod nano would do with a FireWire interface to its flash memory, instead of USB2.
So where does that leave us? Here’s what I think I know as facts (at least relative to my G5)…
- A FireWire2 hard drive is the fastest external storage option, at least using the built-in parts on the G5.
- Using a FireWire drive in the Finder is roughly 2x faster than is using a USB2 drive.
- Copying in iTunes via USB2 takes roughly the same amount of time as does copying in the Finder via USB2.
- Copying in iTunes via FireWire takes roughly the same amount of times as does copying in iTunes via USB2.
- Copying in iTunes via FireWire is roughly twice as slow as copying in the Finder via FireWire. This is due to the slow speed of the iPod’s hard drive, as noted above.
It seems Chris’ thoughts were correct—there’s no real speed penalty for using USB2 in iTunes. However, there’s a definite performance hit for using a USB2 hard drive in place of a FireWire hard drive in the Finder…and if speed is what you’re really after, FireWire2 looks to be worth the investment.
Sorry this was offline so long for some relatively minor tweaks…
Hmmmm - so I ran an i/o test on my Dual 2.5GHz PowerMac (June 2004 model). It has 4,5 gig of RAM and runs Tiger 10.4.2. I have a Maxtor FW400 120gig, Maxtor USB1.1 250gig, MacPower USB2 300gig and a MacPower FW800 300gig. I attached these in turn directly to my G5 and ran tests of writing a 6 GB file to each of them immediately after a system restart to remove any advanatge the kernel buffer cache might provide. I chose a large file that exceeded my system's RAM size to ensure buffering wasn't skewing results.
I created the 6 gig file using Apple's provided 'mkfile 6g 6gig_file'
I then timed the i/o using 'time cp 6gig_file /Volumes/device' where device was the Volume name of each device.
Here are my results of averaging 3 runs for each device.
FW400 took 464.6 secs on average which gave i/o rate of 13.54 MB/sec
FW800 took 141.2 secs on average which gave i/o rate of 44.55 MB/sec
USB1.1 took 857.9 secs on average which gave i/o rate of 7.33 MB/sec
USB2.0 took (had trouble with this one so will post back on this later)
This is definitely not my field of expertise, but it strikes me that the iTunes overhead for copying individual files and putting them in the hidden multi-folder structure may be the key to solving the time anomaly. I've noticed a major difference between the amount of time it takes me to do a Firewire 400 backup on my drive versus copying music to my second- and third-generation iPods.
I guess I just reason that whatever overhead iTunes adds to copying files is done on the iPod end, slowing the rate with which FW can copy, but just meeting the speed that USB 2.0 sends information. Just a theory.
Rob, I don't think the fourth bullet point is correct. If I'm reading the rest of your tests correctly, you only tested FW writing to the iPod through iTunes. So, your sentence should be "Copying in iTunes via FireWire to an iPod is roughly twice as slow as copying in the Finder via FireWire to an external HDD.* If you copy a file through Finder or Terminal to the iPod through FW, the time should not be drastically different from the iTunes time.
What your tests show is interesting, though. Right now, since the HDD in iPods is so slow, there is no performance hit to using USB2 versus FW, even though when the external device is fast enough to handle the quicker interface, FW would outperform USB2. The same thing is slightly evident in your first experiment. The reason FW2 is not 2x FW is not because of the transfer protocol, but because the HDD can't take data any faster.
It also means that as the small HDDs in iPods get faster, Apple is not going to be able to take advantage of it, because of the limitations of USB2. Had they allowed FW to stick around, then future iPods could have faster sync times.
Rob,
There is an 1.5 year old report on the Bare Feats page (http://www.barefeats.com/usb2.html) which might be of interest for you. It states that Apple's USB 2 I/O "management" seems to have some serious problems compared to PC computers. If one would want to be nasty, this person could say that maybe iTunes was programmed to slow the transfer rate of FireWire to match the poorer performance of USB 2 on a Mac.
But still, there are some inconsistencies....
Cheers,
arne
Wait... Did you try a direct firewire copy from the Finder to the iPod?
Seems like the problem could be the firewire interface on the iPod itself, rather than firewire transfers from iTunes. i.e. If you copy files from the Finder to the iPod and the transfer speeds up, then the problem is iTunes; if it doesn't speed up, then the problem is the firewire interface on the iPod.
Yes? No? Did I mis-read something?
You might want to compare with smaller files (MP3-sized) as well, because of the differences between FireWire and USB. I think USB is better for quick bursts of smaller files, but FireWire is able to sustain high transfer speeds for a longer time. FireWire would favor a few larger files while USB would favor lots of smaller ones.
Comments are closed.