SPEED TESTS

FireWire Drives For Music - Part 2 Macs With Built-In FireWire

Published in SOS April 2002
Printer-friendly version

Technique : Computers

We continue our investigation of the effect of drive speed on audio sequencers, switching the focus of the tests to see how FireWire drives can alter sequencer performance on the latest, FireWire-native Macs...

Following on from last month's article on FireWire drives connected to older Macs, this month I will be looking at the performance of a range of FireWire drives on a variety of up-to-date Macs, making some sense of the bewildering array of disk sizes and speeds, and determining whether their performance depends on the computer to which they are connected.

For several reasons, FireWire is fast becoming the connection protocol of choice for computer-based musicians. Although the data-transfer rates are not up to the high-end SCSI systems that are available to the well heeled, the track counts resulting from using FireWire-based storage drives are way above adequate. Based on IDE internal mechanisms, the drives are, for the most part, reasonably priced (there are no 'true' FireWire hard drives -- all the current manufacturers use IDE mechanisms with a so-called bridging chip to the FireWire interface). Connection is a doddle (and it is very much 'connection' as opposed to 'installation'), as is disconnection, giving rise to another major advantage of FireWire -- portability. This has two aspects -- one is the ability of a drive to be carried along with your laptop for the musician on the move (two of the tested drives this month are designed with this specific purpose in mind), and the other is the hot-pluggable nature of these devices. The other two drives on test are designed to be static, but can be plugged and unplugged from laptop to desktop and back at will. In fact, the installation angst that has been a feature of computer-based music-making ever since I can remember was suspiciously absent from every aspect of this test. Not a single compatibility issue, extension conflict or screwdriver was to be seen from test start to test finish. All of which means, of course, that there are no longer any technical excuses to hide behind when you're actually suffering from an acute lack of inspiration...

LaCie sent two drives for test: a 20Gb StudioDrive based on a large-format 7200rpm drive, and a 10Gb PocketDrive based on a small-format 4200rpm drive. Although these capacities seem on the low side, it should be noted that at the time of writing these were the bottom of their respective ranges, and that larger-format drives were continually being announced as the testing proceeded -- indeed, by the time this article went to press, the smallest Studio drive available was 40Gb.

Glyph sent the 40Gb single-drive version of their M-Project, based on an IBM large-format 7200rpm drive (they also do an 80Gb dual-disk version). As its relatively high price suggests, this is just about as serious as t

Test SpecAPPLE IBOOK 500     CPU: G3 500MHz. Buss: 66MHz. RAM: 256Mb. Mac OS 9.1. APPLE IBOOK 600 CPU: G3 600MHz. Buss: 100MHz. RAM: 384Mb. Mac OS 9.2. APPLE POWERBOOK G4 667 CPU: G4 667MHz. Buss: 133MHz. RAM: 512Mb. Mac OS 9.2. APPLE POWERMAC G4 867 CPU: G4 867MHz. Buss: 133MHz. RAM: 512Mb. Mac OS 9.2. APPLE POWERMAC G4 800DP CPUs (dual processors): G4 800MHz (x2). Buss: 133MHz. RAM: 256Mb. Mac OS 9.2

his sort of drive gets. It is rackmounted, shockmounted and comes with a superb warranty that won't leave you stranded. In comparison, the LaCie StudioDrive seems cheaply put together, even down to the FireWire sockets, which feel slightly detached from the chassis. In contrast, the Glyph's feel like they will survive a thousand hot-plugs! You pays your money and you takes your choice... Since I performed my tests, time has also moved on for the M-Project drive, and at the time of going to press, it now comes equipped with a four-socket FireWire hub built in -- perfect for using multiple MOTU 828 FireWire interfaces, which only have a single socket.

Completing the quartet of standard drives is the MCE Transport Pro, based around a 5400rpm small-format drive. It's a beautiful object, which feels slightly delicate for a 'portable' drive; luckily it comes in a padded case with room for cables. It also has an enormous storage capacity (48Gb) considering its tiny physical size. It is the only drive here to use the Oxford bridging chip to connect the drive with the interface (the Oxford chip, as you can tell by any component which gets mentioned by name, is renowned amongst drive connoisseurs as the chip which gives the most impressive performance -- we shall see...). The LaCie PocketDrive is also fairly stylish, and although the mechanism is not internally shockmounted, the flexible rubber sleeve guards against minor bumps in transit.

Both the portable drives have two major advantages over the large drives. Firstly, the FireWire buss powers the portables, so they require no kettle leads or mains sockets, and secondly, they have no fan. Both the Glyph and the StudioDrive make a noise similar to that emitted by the latest PowerBook, and to some minds (like those of some of the frequent visitors to the SOS web forum) this is far too loud to be in the same room as a mic. However, if you put an iBook with a Transport Pro, the result is quieter than a public library on a Bank Holiday.

Following on from Paul Wiffen's Apple Notes piece on the ability of the iPod to play back AIFF (and WAV) files, Apple lent me one so I could assess its ability as a hard drive. Yes, it is an MP3 player, but a thinly disguised 5Gb FireWire drive at that price ought to have other uses...

The majority of testing was carried out on a 600MHz G3-based iBook, with some comparative testing on a 500MHz iBook, a 667MHz G4 PowerBook, an 867MHz G4 Mac and an 800MHz dual-processor G4 (see the 'Test Spec' box below for more detailed info on the test machines). It should be noted that these machines weren't all available at the same time, and since the nature of my tests evolved over time, not all the tests were carried out on every computer and drive combination.

iBook 500 G4 867 iBook 600 G4 733 Glyph long stereo 30 49 34 n/r LaCie StudioDrive long stereo 20 39 35 31 LaCie PocketDrive long stereo 16 30 26 n/r Internal drive long stereo 18 52 23 35 MCE long stereo n/r n/r 24 n/r Table 1 -- Long stereo audio files on the Power Mac G4 867, Power Mac G4 733, iBook 500 and iBook 600.

As explained last month, there are quite a few parameters used to measure the performance of a hard drive, with no one overall figure available for direct comparison of absolute performance. Seek time an

Moving Audio Files     All the audio files created in the tests performed in this article were initially saved to the internal drive of the test computer (which was also one of the drives tested). When the time came to test an external drive with the same audio material, I found several methods for moving the audio in a way such that Logic could still find it. The slow way of moving them is by using the 'Copy/Convert' or 'Move File' functions from Logic's Audio menu. A much faster method is to copy the files across from drive to drive using the Finder, and then delete the originals that are referenced by the Song. This forces Logic to look for the files somewhere else, so when you reopen the Song, Logic asks you to locate the first file in the audio window list, and provides you with a browsing window. When you tell it where to look, the sequencer then looks in this location for all the other files. If any of the required files are not in this location, you're again asked to find them. The point here is that copying audio files in the Mac Finder is much faster than copying files from within Logic, and these test songs (and your recordings, if you fully exploit the ability of these drives) referenced enough data to make this a significant time-waster.

d sustained data transfer rate are regularly quoted by hard-drive manufacturers. Less commonly quoted are the head settle time (or 'latency') and the burst transfer rate. Total access time to reach a given piece of data should be the sum of seek time and latency. Typically, these might be 9mS and 1.5mS, the total of which is greater than the magic 10mS figure usually quoted by music-software people as an absolute maximum for successful recording and a happy life. However, there are many kinds of work being carried out on computer-based music systems, all of which stress the hard drive in different ways.

To show you what I mean, consider this example. In all the tests I performed this month, every audio track was made to start at the same point in time. For Emagic's Logic (the sequencer used for most of the tests this month), this was always the moment of greatest stress, and the point at which the program would stop playing if too many tracks were unmuted. Once the sequencer's 'virtual playback head' had

iBook 600 LPB On/LDB Off LPB Off/LDB On Glyph long stereo 34 47 LaCie StudioDrive long stereo 35 43 LaCie PocketDrive long stereo 26 37 Internal drive long stereo 23 33 MCE long stereo 24 35 Table 3 -- Long stereo file tests on the iBook 600.

passed this point, the reading on the hard drive activity meter fell to a lower level, especially in the short file tests (more on these in a moment). Drives with a higher so-called 'burst transfer rate' would be better equipped to cope with this moment in the Song. So although prospective drive buyers for musical applications are normally told to steer clear of the burst transfer rate and look at the sustained data transfer rate instead (the burst rate can be up to four times the sustained rate, and can therefore be misleading when used as a yardstick for overall performance), the burst rate does nevertheless give an idea of how well the drive might stand up to periods of short-term stress.

In order to come up with some meaningful comparisons, I decided to structure the tests around two basic type of audio files, 'long' ones and 'short' ones. The 'long' audio file is the type that you mig

iBook 600 LPB On/LDB Off LPB Off/LDB On Glyph short stereo 24 30 LaCie StudioDrive short stereo 18 23 LaCie PocketDrive short stereo 14 5 Internal drive short stereo 18 29 Table 4 -- Short stereo file tests on the iBook 600.

ht be dealing with if your hard disk recording setup is functioning as if it were a tape-based system (ie. files which begin at the start of the song and end at the end with no breaks in between). The 'short' audio file is of the type used in sample-based compositions requiring the playback of many short drum hits or loops, for example.

THE LONG FILE TEST

My first Logic test Song (see above) was created to test 'long' stereo file playback. Using Logic Audio Platinum v4.8.1, I imported 64 files into the Audio window from an audio sample CD which I had previously converted to AIFF format using the freeware batch converter SoundApp (available from http://www-cs-students.stanford.edu/~franke/SoundApp/). I used the sample CD as the source material because it contained 99 reasonably long ready-made tracks. Each file in turn was dragged from the Audio window onto a separate audio track in the Arrange window, making a song which would require the hard drive to play back up to 64 continuous audio tracks simultaneously, depending on the drive's ability.

This test is of course the same as the one used in last month's article on FireWire drives connected to older Macs, so it should give some useful comparisons with those systems. There was one major difference from last month's test, however. Classic multitrack tape recording is based on the concept of mono tracks; that is, a stereo track would take up two tracks. Partly to emulate this situation, and partly to confirm that a stereo audio file causes a hard drive less stress than two mono files (the disk only has to find one file), the long files test was constructed this month using mono audio files. The means of construction also differed slightly. I imported one mono file of suitable length into the audio window, then dragged it onto the first audio track on Logic's Arrange page. I then copied this by holding down the Control key and dragging the first instance to create a new region per audio track. To create actual audio files, I used Logic's 'Convert Regions to Individual Audio Files' function from the Audio menu, having sele

Differing Results From Logic & DP3     The maximum track count for Logic Audio using a single audio device is 64 tracks (although as of the upgrade to v5, this will increase to 192 mono tracks). Until recently, this restriction has never been much of an issue, as drive performance tended to confine the maximum possible track count to below this limit. However, due to the efficient performance of the drives I was testing, it seemed that I might have exceeded the 64-track maximum in several of my tests, if Logic had been capable of it. I noticed this when running the fourth (DP3-based) test. Logic managed 47 tracks with the Larger Disk Buffer on, but DP3 ran 84 tracks! Emagic were unable to explain this or replicate the problem at the time of going to press. It's possible that something in DP3 allows it to run more efficiently under these conditions, but it's impossible to be certain.

cted all the new regions on the arrange page. The process of testing with each song consisted of unmuting tracks one at a time until Logic stopped playing, ensuring that the song pointer passed the beginning of the files each time a track was unmuted, as this was the point of highest disk stress in nearly all cases.

THE SHORT FILES TEST

My second test Song (shown on the next page) was based around 'short' audio files. I dragged one 'long' audio file onto Track 1 in Logic, trimmed it to eight bars in length, and then divided it into 16th-note regions. Since these all refer to one audio file, it was again necessary to use Logic's 'Convert Regions to Individual Audio Files' function. I then copied the contents of this track to the other 31 tracks used in this test. Since the copies are merely regions of the audio files created on Track 1, I also had to use the 'Convert Regions to Individual Regions' function, followed by 'Convert Regions to Individual Audio Files' to create the multitude of sho

Memory Bandwidth     A recent thread drew my attention on the SOS forum on the subject of memory bandwidth. It was pointed out that Motorola's G4 processor design has changed during its existence (I can't imagine why -- perhaps to benefit the customer...?) with the effect that older processors such as that featured in the Titanium PowerBook G4 500 can run more PlatinumVerb plug-ins in Logic than its 'faster' successors, the supposed reason being memory bandwidth. For more on this, see: www.xlr8yourmac.com/systems/PowerBookG4 _fall2001/powerbook_g4_667_quake3.html#logic Memory bandwidth is the speed at which data can move through the memory, and a utility such as Gauge Pro (available from www.macupdate.com/info.php/id/4512) can give you a benchmark to compare your processor with others. The xlr8yourmac.com site, incidentally, is a great source of such useful information for owners of any kind of Mac, particularly if you're intent on avoiding any proper work! I ran Gauge Pro on some of the computers used for testing this month, with the following results (see the first table below). Perhaps we should include a measurement like this along with processor and buss speed when considering a Mac's potential? Although I (and others) have reservations about using the PlatinumVerb plug-in as a means of benchmarking, it is widely used as such, and I therefore include some PlatinumVerb results of my own (see the second table). The PlatinumVerbs were mono, and the Larger Process Buffer was on in these tests. As with hard drive parameters, it should be obvious from these results that no one figure can be used to describe a machine's ability.

rt files I wanted. As with the long files test, this test was also recreated using mono files.

THE MIXED FILES TEST

After creating the first sets of tests, I decided to make a test song which, although still standardised and quantifiable in terms of files and tracks, reflected more faithfully the demands of most projects, which might have roughly equal numbers of tracks based on long and short files. Test Song 3, then, features Track 1 playing a 'long' file, Track 2 playing 'short' files, and repeats this pattern up to 32 tracks (see screen grab, below left).

MOTU DP3 RECORDING TEST

The fourth test song was devised to check the simultaneous recording ability of the drives on test, and necessitated a move to a different sequencer. As it turned out, all the drives I was using were very capable, which posed a problem for Logic. With my test setup, using Mac AV and a MOTU 828, the maximum number of inputs available for recording in Logic was 18 (eight analogue, 10 digital) so I needed an alternative way of simulating a really large multiple-input recording situation. Unlike Logic, MOTU's Digital Performer v3 can record its buss outputs, the number of which can be specified in its user-definable Studio Configuration. For example, 64 busses combined with 18 inputs gives a total of 82 possible audio tracks that might be simultaneously recorded, so I switched to DP3 for this test. This led in turn to another surprising result (see the 'Differing Results' box later in this article).

OTHER TESTS & SETTINGS

Along the way, a number of other test options presented themselves, but since these were one-off tests to confirm a suspicion, I shall describe them as I discuss the results.

For the purposes of fair play, each computer was restored to its factory default using the machine-specific System disks. All extensions were left on, and disk cache settings were set to default. Also, each drive (except internal System drives) was initialised using Mac OS rather than manufacturers' proprietary disk-formatting software (more on this later) before each test song folder was copied onto it.

iBook 600 PowerBook G4 667 G4 PowerMac 800DP LPB On/LDB Off LPB Off/LDB On LPB On/LDB Off LPB Off/LDB On LPB On/LDB Off LPB Off/LDB On Glyph long mono 44 64 49 64 50 64 LaCie StudioDrive long mono 48 64 52 64 61 64 LaCie PocketDrive long mono 28 47 28 52 29 51 Internal drive long mono 36 50 30 49 64 64 MCE long mono 31 49 29 48 n/r n/r iPod long mono 21 32 23 32 n/r n/r Glyph short mono 18 25 19 25 20 31 LaCie StudioDrive short mono 17 23 17 24 18 27 LaCie PocketDrive short mono 9 12 9 14 9 17 Internal drive short mono 10 18 10 16 20 32 MCE short mono 5 11 9 13 n/r n/r iPod short mono 5 7 5 7 n/r n/r Table 5 -- Mono file tests on the iBook 600, PowerBook 667 and dual-processor Power Mac G4 800.

All the tests which resulted in Tables 1 and 2 in this comparison (see below) were carried out with Logic's 'Larger Process Buffer' (or LPB) switched on and 'Larger Disk Buffer' (or LDB) switched off from the Audio Hardware and Drivers window. This is not necessarily the best setting for disk performance, but was used for parity with earlier tests. The audio device used was Mac AV. 'n/r', incidentally, means 'no reading' (ie. I didn't take one).

On the Power Mac G4 867, the Glyph drive beats off the LaCie StudioDrive by 10 tracks in the long files test and five tracks in the short files test. However, both were bettered on long files by the 867's internal drive, proving a better sustained data-transfer rate. The PocketDrive fares reasonably well considering its lower specification.

A combination of lower buss speed (66MHz instead of 133MHz) and processor speed (500MHz instead of 867MHz) results in a reduction in all counts on the iBook 500. Its close relation, the iBook 600, shows the benefits of a faster system. The short file tests show much greater improvement than a 100MHz hike in processor speed should allow, thus indicating that the faster buss speed (100MHz in the iBook 600) is perhaps a more significant factor.

One anomalous figure is that for the Glyph drive running long stereo files on the PowerMac G4 867. This is the only figure in the whole test which doesn't fit the general conclusions at the end of the article. There's always one!

Tables 3 and 4 (shown above) show the benefit of paying attention to Logic's audio preferences -- just look how many extra tracks can be obtained by altering the LPD and LDB settings! It should be remembered, however, that if you were operating at the upper edge of a disk drive's ability, favouring the disk at the expense of

iBook 600/Glyph/ iBook 600/Glyph/Glyph driver Mac OS driver Long Stereo Tracks (LDB On) 47 47 iBook 600/Studio Drive/ iBook 600/Studio Drive/Lacie driver Mac OS driver Long Stereo Tracks (LDB On) 45 45 Table 7 -- Tests on manufacturers' own disk drivers.

the processor would limit the number of plug-ins you could use.

As previously mentioned, running Logic with its 'Larger Process Buffer' on and 'Larger Disk Buffer' off will not achieve the best performance from a hard drive. Nevertheless, it remains useful to see the figures from Table 5 (above) which illustrate expected track counts if a large number of plug-ins and/or virtual instruments are required.

The Glyph only just squeezes out the LaCie StudioDrive on all three machines, and in both

Pricing & Contacts     APPLE 5Gb iPod: £349. Apple Store +44 (0)800 039 1010. www.apple.com/uk GLYPH 40Gb M-Project (single drive): £499.38. Global Distribution +44 (0)870 464 0600. www.glyphtech.com LACIE 60Gb StudioDrive: £198.58 (20Gb not sold). 20Gb PocketDrive: £186.83 (10Gb not sold). LaCie +44 (0)20 7872 8000. www.lacie.com MCE 48Gb Transport Pro: £539.33. AM Micro +44 (0)1392 426473. www.mcetech.com MOTU 828 interface: £795. Musictrack +44 (0)1767 313447. www.motu.com All prices include VAT.

tests. On the laptops, the internal drives level-peg the portable drives with long files (indicating an equivalent data-transfer rate), but win easily with short files (as they offer a better seek time), and the two portable drives are impossible to separate. The Power Mac's internal drive, on the other hand, comes in with a performance equal or better than the best of the external drives. This ought not to be a surprise, since the FireWire drives are based on internal mechanisms, but with a bridging chip in the way!

Comparisons drawn between the figures for the iBook and PowerBook are surprising -- average performance across the range of drives is virtually identical. Given the advantages that the PowerBook has in both processor and buss speed. I can only suggest that the limiting factor might be a similar level of FireWire implementation on both the small motherboards.

Interestingly, the iPod gives reasonable results here (considering its size) if using long audio files, but really falls down when its seek time is challenged by short files.

The 'Combination tests' (ie. long and short files) are shown in Table 6 above. Just as with the stereo files (but bizarrely not the mono files), the Glyph drive manages significantly better than the LaCie StudioDrive, and is matched in performance by the Power Mac's internal drive. Both are ahead of the portable drives, and the iPod, not surprisingly, trails a little further back. Again the PowerBook and iBook score almost identically, with only small improvements in some areas when using the Power Mac.

I also carried out some supplementary tests (see Table 7 above) to compare performance of the drives following both a standard installation and one using the manufacturers' 'bespoke' disk drivers, which are installed when you format a drive with its supplied disk utilities. Despite manufacturers' claims that their drivers are 'optimised' for audio use, these tests go some way toward proving that there is no advantage to be gained from using the bespoke drivers. What's more, it's much more convenient to use Mac OS to format drives; simply single-click on the mounted drive icon on the desktop, and select Erase from the Finder's 'Special' menu!

Additionally, I tested performance with the Glyph drive on pairs of long and short files via both the Mac AV audio driver and the ASIO driver supplied with MOTU's 828 interface (see Table 8 below). This test shows that the audio driver selected does not seem to influence track count on FireWire drives.

Finally, the memory control panel in Mac OS allows you to specify a disk's cache size. Although programs like Final Cut Pro specify a setting as low as possible (128K), the results of doing this (see Table 9, below) don't really offer sufficient evidence that tweaking this control panel will give any significant improvement in disk performance.

Looking firstly at the computers used for testing, it seems that, as far as hard disks are concerned the iBook and PowerBook offer pretty much the same level of performance, whether you're using an internal or external FireWire drive. External drives only seem to offer a slight improvement when connected to a Power Mac (except in a few isolated instances), but the internal drive of this machine is probably the best-performing drive on test here.

The Glyph M-Project is the best FireWire drive on test, but then it should be at the asking price, and on pure performance, it isn't a great deal better than the LaCie StudioDrive, which has to win the tracks-per-pound prize. Both portable drives are quite useable in an audio context, as long as the project doesn't involve too many 'short' files, and come with the added bonus of being almost silent in operation. Finally, the iPod could be used for very small projects (in terms of track count) and also for emergency backup. This may seem measly compared to the other test drives, but then you have to admit that none of the others are portable stand-alone sound file players!

Overall, the figures speak for themselves. If you have a requirement for a second hard drive (and if you are doing any serious computer-based recording, leaving your System disk unsullied by audio chores is, by and large, a necessity) FireWire combines the double benefit of portability and connectivity with superb performance, whichever format you choose.

Home | Search | News | Current Issue | Digital Editions | Articles | Forum | Subscribe | Shop | Readers Ads

Advertise | Information | Links | Privacy Policy | Support