Each time I return to the subject of hard-drive partitioning, the goal-posts have moved significantly. Back in August 2000, when I wrote the 'Divide And Conquer' feature, many of us were still using drives with a capacity of 8GB or less and running Windows 98. By the time I wrote 'The Great Divide', in SOS March 2003, this had jumped to 80GB, and many of us were more interested in splitting our drives to run Windows 98, 2000 and XP alongside each other in a multi-boot configuration. Two years on, in 2005, I still find myself answering lots of questions about partitioning in the SOS Forums, but while most of us are now using Windows XP, there are several new issues to face and splitting drives has almost become mandatory, as some of us install models with a capacity of 200GB or more.
As most PC musicians already know, you don't have to leave each of your hard drives as one huge and rather unmanageable storage area. It's generally far more productive to split each one into several partitions, to keep your data better organised, and therefore safer. This also helps to minimise any reduction in performance due to file fragmentation, by keeping the fragments within a smaller area of the drive so that the drive's read/write heads don't have to dart about so much when accessing files.
Most people also agree that having two drives in a music computer is generally better than one, since you can devote one to Windows and its applications and the other to audio storage, safe in the knowledge that they won't interfere with each other in any way that reduces overall performance. This approach also offers a security benefit: you can back up the data from one drive onto a spare partition on the other, so that if one drive fails your data can still be retrieved from the surviving one (although this isn't a substitute for a proper backup regime onto other media).
There's now a huge number of musicians who run PC laptops on stage, for location recording, or simply for greater convenience. They generally rely on a single drive, yet still want to achieve the maximum audio performance from it. Many musicians are also now using streaming soft-samplers, which brings up issues surrounding hard drives once again. If you're running one of these samplers alongside a multitrack audio sequencer, for example, do you need three drives (one for Windows and its applications, the second for audio tracks, and the third for sample libraries) to achieve maximum soft-sampler polyphony without compromising the maximum number of audio tracks? If not, what's the best way to split your data requirements across one or two drives?
Let's start by considering the implications of partitioning. While it can make our lives a lot easier, it results in a set of storage areas, each performing slightly differently, and with the possibility of interactions between them in real time. Let me explain.
The maximum number of simultaneous audio tracks your hard drive can manage is directly linked to its 'sustained transfer rate', rather than the burst speeds often quoted by manufacturers, because streaming audio is generally a continuous task. This sustained transfer rate is, in turn, directly related to spin speed: the faster a drive spins, the more of its data sectors can be read per second.
So the 4200rpm drives found in some laptops generally achieve the lowest number of audio tracks, the 5400rpm ones found in old desktops and many current laptops are better, and the 7200rpm drives commonly found in more up-market laptops and the vast majority of desktop PCs tend to be the most popular for audio work. There are faster drives available (including 10,000rpm and even 12,000rpm models), but these are significantly more expensive and often more noisy. In any case, 7200rpm models can already run more simultaneous audio tracks than most musicians need.
The sustained transfer rate will also vary from the outside to the inside of any drive, simply because the read/write tracks are arranged in concentric circles. Since the outer tracks are longer, they contain more sectors, and thus at a fixed spin-speed more sectors of the outer tracks can be read in a single revolution. So the fastest area of any hard drive is always on the outside. With most (but not all) drives, the sustained transfer rate falls steadily from the outside to the inside, and may typically drop by half in the process (I have seen exceptions where the rate suddenly jumps up again slightly in the middle, or falls in multiple steps like a sawtooth waveform, but these seem comparatively rare).
Sustained transfer rates can be obtained from manufacturers' drive specifications and are generally quoted as a range, such as "32-58Mb/second", or as an 'average' value, but it obviously is sensible to make best use of the fastest areas. Fortunately, you can easily measure the sustained transfer rate of any partition of your drives using the now-famous DskBench utility (a free download from www.sesa.es/us/dl/dskbench.zip). Simply copy this tiny 36KB file onto the partition or drive you want to measure and then run it from there. Among various other results, DskBench will provide you with figures for Write and Read — the two sustained transfer rates that ultimately determine audio recording and playback performance — as well as approximate CPU overhead. If the latter comes in at under five percent or so, you'll be reassured that DMA data transfers are correctly set up for your drives.
If you want to see how the sustained transfer rate varies across the entire surface of your drive, try downloading the 1Mb HDTach utility (www.simplisoftware.com/Public/index.php?request=HdTach). Now up to version 3, this free hard-drive benchmarking tool provides a useful graphic readout of Sequential Read Speed from outside (left-hand side) to inside (right-hand side), as well as Random Access Time (a measure of how fast the drive can retrieve randomly located sectors) and CPU Utilisation (once again, to check the effectiveness of your DMA transfers). It's also a Windows-based utility, which makes it easier on the eye than DskBench.
Suitable Partitioning Tools
You can choose what you want the top and bottom views to be in the right-hand window, but by default these views display a list of Volumes (partitions) at the top, and a Graphical representation of the same beneath, showing primary, extended and logical partitions, plus any free space. What many musicians haven't cottoned onto is that the lower Graphical view also shows the layout of partitions from outside (left) to inside (right) of each drive.
The Disk Management tool is very useful if you know exactly how you want to split and format a new drive, delete existing partitions or change drive letters. However, you can't use it to re-size existing partitions, so unless you get it right first time and never subsequently change your mind as your drives fill up, you'll need a more flexible tool, such as the famous Partition Magic, developed by Powerquest, but now owned by Symantec (www.powerquest.com/partitionmagic).
As you can see from the screenshot, Partition Magic shows exactly the same arrangement of partitions for my two drives as the Disk Management tool, with the bonus that they are appropriately scaled to show what proportion of the total drive they occupy. Partition Magic really comes into its own when you want to reorganise your data: you can re-size, move, copy, or delete existing partitions at will, split them or merge them together, convert them from one format or partition type to another, change their cluster size, or create new partitions, ready to install additional operating systems.
A few alternatives to Partition Magic are now available, including Paragon Software's Partition Manager (www.partition-manager.com) and Acronis Disk Director Suite (www.acronis.com/ homecomputing). Both can be ordered on-line for $49.95 (compared with the $69.95 of Partition Magic), and both offer a comprehensive range of functions. Although the Acronis product doesn't currently offer a couple of options that the other two do, it does support Windows Server Editions (you need to buy the more expensive 'Pro' versions of the competing software to do this). On the other hand, Partition Manager does offer an advanced Defragmentation tool, which looks very useful.
If you have a single drive in your PC and it's formatted as one massive partition (the default in most mainstream systems), it will seem quite nippy when delivered. However, it may gradually become more sluggish as you install more applications and store more data as, inevitably, later storage will be physically further towards the inside of the drive. Not only will files stored in these areas suffer from slower transfer rates, but access times may also become longer, as the hard drive read/write heads have to keep jumping longer distances during normal use to reach all parts of the drive.
Windows does its best to help, by periodically re-organising files so that those accessed most often are moved to the outer portion of the drive, using its 'pre-fetch' optimisations (I discussed this in 'Speeding Up Windows' in SOS April 2003). You can force this re-organisation to occur by running Microsoft's Disk Defragmenter utility, but you can also help a great deal by splitting your drive into several partitions, and placing data that's only needed occasionally (such as large image backup files, or your collection of application update files) on the innermost partition.
In general it's best to make the partition most often accessed the fastest one, and for most general-purpose software this would be the one containing Windows and its applications. Your PC would then boot up in the shortest possible time and launch applications as quickly as possible. However, for musicians streaming audio files from their hard drives, it makes more sense to ensure that the partition containing audio files has the best performance, despite the fact that this will result in Windows and applications launching slightly more slowly.
Here's a practical example, based on the 5400rpm 40GB Seagate Momentus hard drive in my Centrino laptop. As you'll see from the HDTach screenshot above, the performance of my laptop drive falls off from about 36MB/second to about 29MB/second across the first 20GB — a 17 percent reduction. So if, like many people, you have a huge 20GB partition devoted to Windows and its applications, with the rest of the drive devoted to audio duties (Single Drive Scheme two in the diagrams below, simply swapping the positions of the partitions could potentially increase your maximum audio track count by 17 percent.
Rather than re-allocating a large proportion of the outer part of the drive to a single audio partition, it makes more sense to create one partition just big enough to house your current audio project (say, 5GB), and another further in for all your older, finished projects. In this way you can achieve the maximum track count without compromising Windows performance too much (in my graph, drive performance has only fallen by about three percent at the 5GB point).
In this example (Single Drive Scheme three), we've gained a 17 percent improvement for audio tracks and only lost about three percent for Windows, simply by creating an outer 5GB project partition (see 'Windows Activity' box for more details on whether the slight loss of Windows performance will be noticeable). You can extend this idea, if you've got a streaming sample library, by creating another larger partition for the library, between the current project and Windows partitions. In the case of this particular (rather small) 40GB laptop drive, you could perhaps devote 15GB to the sample partition, still leaving plenty of space for Windows and an inner data and backup partition (Single Drive Scheme 4). Although Scheme 4 further compromises Windows loading times, placing the audio and sampling partitions next to each other minimises access times. This arrangement will probably provide the best audio and sample streaming performance on a single-drive laptop.
By the way, don't confuse the requirements of streaming sample libraries with samples that are always loaded into RAM in their entirety. The latter are no different from patch-based libraries. These two categories (RAM-loaded samples and patch libraries) should ideally be separated from your streamable sample libraries and placed into yet another data partition further in on the drive. After all, the only improvement you'll see if you place these libraries in an outer partition is a slight reduction in their initial loading times.
Some of you may have noticed from the HDTach screenshots that since most drive read speeds fall to approximately 50 percent as you move from outside to inside, buying a larger-capacity drive will help you make the most of the fastest outer area. On a 200GB drive, for example, there may be little fall-off in read speed across the outermost 50GB, making this area ideal for large current projects or sample libraries. So even if you only have one drive in your PC, the larger its capacity, the less you'll compromise audio performance with the sampling split of Single Drive Scheme four.
Swapping the positions of your Windows/applications and audio data partitions initially sounds like a great idea, until you start worrying that perhaps Windows and application performance may, in turn, suffer if they're moved to a slower inside partition. After all, running from a partition with a lower sustained transfer rate is bound to make Windows boot up more slowly, since its system files will take slightly longer to load into RAM. Applications will also launch more slowly, albeit by a tiny amount.
I decided to perform some tests to throw some further light on the matter. I started by running the Filemon utility from Sysinternals (www.sysinternals.com). This monitors and displays all file system activity so that you can see how Windows and your applications access and use files. You can decide which volumes to monitor, and you can stop and start the capture process at any time. Although it's very easy to get swamped with data, with care you can learn a vast amount about which files are being accessed on which drives.
After a run-through playing a Cubase SX song with a clutch of audio tracks, I discovered (as I half expected) that for most of the time the only volume being accessed was the one containing the audio files used by the song. Further monitoring showed that only when opening or closing a file dialogue window (load, save, import, and so on) or exiting Device Setup (when Cubase saved its new Defaults.xml file) was there any obvious activity on the C: volume. Even when Gigastudio was running alongside and accessing my dedicated Samples partition there was still no additional activity on the C: volume. I made similar checks running Cubase on my XP General partition and found similar results.
Sonar was slightly different: I found that it accessed the C: volume once every 2.5 seconds for a tiny amount of time even when songs weren't playing, and during playback it accessed its huge TTSRES13.DLL file about five times per second. However, each access lasted less than a millisecond, leaving the C: volume inactive for at least 99 percent of the time. I think we can safely assume that placing an audio or streaming sample partition on the same drive as the Windows partition won't compromise performance very much.
Other audio applications may differ in their approach, but I suspect that you're reasonably safe in creating an audio/sample partition on the same drive as Windows if there's some reason why you can't devote another drive to this task — just be aware of possible interactions and try to keep them to a minimum. While you may not achieve 100 percent of the performance that you might get using a separate audio/sample drive, it might stay near 99 percent most of the time — but it could possibly plummet very occasionally if the drive is suddenly called on to perform another task. One possible task might be Window's Paging File, created by default on the C: volume for use as virtual RAM, particularly when you're running low on the real thing. I've written about the page file on several occasions in the past, and on my machine it doesn't seem to get accessed very much. However, those running video applications often find that they need a huge page file, which is accessed quite often. If you find yourself in this situation, another option is to create a small extra partition (of several GB in size) near your audio or sample partition, to use as a dedicated page-file partition. That should minimise any disruption of streaming activity.
While some mainstream PCs may arrive with a single massive hard drive, most PCs intended for audio will be better off with two smaller ones, or one small and one large — the first for Windows, applications and other data, and the second for audio duties. This not only removes the possibility of audio accesses being compromised when Windows wants to read or write a system file, but also means that you can back up the data from one drive to a partition on the other.
But even with two drives it pays to consider their partition splits carefully. Although the commonly used split of duties shown in Twin Drive Scheme one (above) generally works well, it still makes sense to split the Windows drive so that you can separate your own data (documents, graphics, and so on). Then, if anything ever happens to Windows, such as corruption or a bad virus attack, your work will be unscathed, even if you have to re-format the Windows partition and install your applications afresh. At the same time, it's a sensible move to create a third partition dedicated to image files of your Windows partition (made using a utility such as Norton's Ghost), so that you can restore it to full working condition within a few minutes rather than having to reformat and start again. The setup just discussed is, in fact, the backup split (Twin Drive Scheme 2).
If your songs contain a lot of audio files, the 'Current Project' audio tweak suggested earlier for the single-drive setup will also benefit a twin-drive system (Twin Drive Scheme three), as it ensures that you always achieve the best performance for the files you're currently using, but you've now got the space to make the Current Project partition much larger, if you're using the 24-bit/96kHz format, without compromising Windows performance at all.
Twin Drive Scheme four is yet another variant for those who want to stream lots of samples at the same time as playing back many audio tracks. I've also moved the Data and Windows Backup partitions onto the second drive, to leave more space for samples, and to keep it away from Windows itself, for more security. This time it's not so important to move the Streaming Samples partition further to the outside of the drive than the Windows partition, since it's likely, with a twin-drive setup, that each drive is somewhat larger than my laptop example. In my own desktop partition setup, for example (shown in the second screen on page 130), my first drive is a 80GB model with three small Windows partitions, each of 5GB, with the remaining 60GB devoted to Samples. Referring to the HDTach graph for my drive, starting the sample partition at the 15GB point only results in a loss of the top five percent of the drive's performance, which I can live with. If I had a larger first drive, this performance reduction would be even less. On the other hand, placing the Windows partitions on the final (inside) 15GB of the drive would result in at least a 33 percent increase in Windows file-access times.
If you've just bought a PC it's sensible to make the most of its potential performance by organising the drive or drives into partitions. However, if you've already got a mature system with loads of stuff installed on your drives, will re-organising Windows, applications and data result in significant benefits? Well, it depends on how you work, whether or not you've already hit the drive-performance ceiling of your existing setup, and how much the sustained transfer rate of a particular partition could be improved by re-siting it, if it's currently a limiting factor on performance.
If you work mainly with MIDI hardware and software synths and plug-ins, and your MIDI + Audio sequencer application doesn't show a high reading for hard drive activity then there's absolutely no point in reorganising — it may make you feel that your PC is better set up, but you won't notice any improvement in performance. Creating a special outer 'Current Project' partition might shave a little off the initial loading time of your songs, while creating a special outer 'Samples' partition to store large non-streaming sample libraries (like those of Spectrasonics' Atmosphere or Trilogy might instead shave a second off initial loading times when loading a new patch, but you'll see no real-time improvements. Even if you're running loads of audio tracks, if you've never found your drives 'running out of steam', you won't notice any improvement.
However, for musicians running Gigastudio, HALion, Kontakt, or any other streaming soft-sampler, spending a few hours reorganising your drives could immediately offer benefits in polyphony terms, especially if you use instruments such as piano and harp that can easily consume 64 notes or more by themselves. It may even allow you to avoid having to buy and install a third drive just for sample storage.
Reorganisation could similarly improve the lot of the PC musician running a laptop with a single drive, when squeezing every last drop of performance out of the internal drive may make the difference between being able to continue using the laptop by itself, or having to cart around an external drive to beef things up.
I've written on various occasions in the past about partitioning hard drives. Here are links to the main articles:
Backup Strategies For The PC Musician www.soundonsound.com/sos/aug04/articles/pcmusician.htm (covers organising your data, file-naming tips, what to include in backups, data recovery, different media, file compression, dedicated software, sharing data).
Speeding Up Windows www.soundonsound.com/sos/apr03/articles/pcmusician0403.asp (covers chipset drivers, Intel Application Accelerator, Bootvis, boot-up time, device initialisation, Windows XP boot time).
The Great Divide: Partitioning PC Hard Drives For Multi-Boot Systems www.soundonsound.com/sos/mar03/articles/pcmusician0303.asp (covers reasons to split, format & partition types, partition options, Boot.ini name syntax, Microsoft multi-boot options, XP boot errors).
One Box Two PCs www.soundonsound.com/sos/may01/articles/pcmusician.asp (covers swapping drives, multiple operating systems, putting up the partitions, installing a second OS).
You don't have to religiously follow any of the schemes I've outlined here: once you've grasped the reasons for each of my suggestions you can adapt them to your needs. For instance, if you've already created dedicated audio and sample partitions and are currently struggling to achieve a higher sampling polyphony, but you have no problem running lots of audio tracks, it may benefit you to swap their positions on your drive or drives. As long as you bear in mind the falling transfer rate from the outside to the inside of a drive, the relative sizes of partitions are also entirely up to you. The beauty of the partition utilities discussed in the box on page 130 is that you can change partition sizes at any time, if you run out of space in one and find you don't need as much space as expected in another. It's not uncommon to shuffle partition sizes or even their relative positions several times before finding the best arrangement.
My tests on Windows partition activity while running audio applications confirm that laptop owners will generally achieve acceptable performance by splitting a single drive. I've been successfully streaming Gigastudio from a partition on my Windows desktop drive for several years, with no problems to date. But don't run away with the idea that you should necessarily abandon the twin-drive approach: using a single drive is always a slight compromise, so you should be prepared for at least a slight drop in polyphony or maximum track count. You'll also have to be careful to keep the Windows and audio partitions as physically close as possible, to minimise the impact of extra drive activity, as well as avoiding any utilities that regularly access files on the Windows drive.
The central issue is awareness of how placing partitions on a drive affects their performance — and don't forget that it's still important to keep individual partitions regularly defragmented and the data within them regularly backed up. Finally, remember that, however capable the latest partition utilities are, no piece of software is 100 percent foolproof, so make sure you have backup copies of the data in each partition before you attempt to move them elsewhere. Then, if the worst happens during the reshuffle, you won't have lost any data in the process.