Why (2) two hard drives for recording ?

I have listed a few reasons why two hard drives are better than one as many people continually ask me why I dont recommend partitioning a single hard drive, and instead recommend two separate hard disk drives. If RAID is used then you need 3 hard drives for the same reasons that are outlined below, I strongly recommend against raid unless its really needed in your situation which is very rare. With raid one drive is used for the OS, and another which could be an Array of two or more disks for recording on to. Raid can actually be worse then two disks and this page explains why raid can cause problems when recording.

DATA Corruption

Computers do make mistakes now and then. Have a look at the specs below and you'll see the ERROR RATE of a standard 7200 rpm drive. It has been found that recording audio or streaming large files on to the same hard drive that contains your swap file can cause corruption in your swap file and other important system files...For this reason NEVER record to the same partition as your swap file and windows files. If you only want a few audio tracks (less than 8) then 1 hard drive is all you need... If you want low latency or higher track counts then this is another issue and two disks is probably required.

Faster performance

By having one physical hard drive for your operating system and another hard drive for your Audio files, you achieve many things. The first is faster performance. In the case your only using the one drive for both purposes it takes time for the read/write heads to move from the system area of the drive to where your audio files are kept. Refer to the example specs below to see a real SEEK time.... If the heads have to move from one side of the disk to the other, you have to wait 28 ms depending on the exact hard drive model. You may go "so what ? its only 28ms", well your audio card is probably set for 23ms latency by default unless you have changed the buffer even lower.

Decreases head movement, will lengthen the life of both drives

If you only have one hard drive and this includes a single drive in two partitions, the read/write heads have to move from the system partition all the way to the audio partition. Not only does this take time for the heads to move back wards and forwards all the time, it also really wears out the hard drive due to the excessive movements of the heads. Not to mention the risk of the hard drive not being able to deliver the data in time. If you have two drives then the heads hardly move at all. A reduction in noise will also be noticed. If you partition a single disk in to two partitions then your actually making the drive work harder and perform a lot more unnecessary head movements. The drive is then slower and gives worse performance when compared to leaving the disk as one large partition. Of course leaving the disk as one partition means the drive becomes fragmented much faster...

Prevents both drives from becoming fragmented as fast

The more free space you have, the less the drive becomes fragmented. When a drive becomes fragmented the SPECs of the drive begin to get worse. Use a de-fragmenting utility to defrag your hard drives regularly regardless of how many drives and the size of them. Also if you don't install and uninstall any software, the system disk takes forever to become fragmented and you only need to defrag the audio drive which can be done when your checking emails and the like. You will have a much more responsive system.

Allows for Lower latency

See all the above topics on how it helps give lower latencies. The main one is the SEEK time of the drive. If you have a separate drive for audio and the files are not fragmented then your drive has a seek time of 2ms. With one drive you may still get 2ms most of the time until Windows needs a file to keep running. When this occurs that seek time can jump up to figures over 10ms....Dont forget the drive latency is on top of this seek time as well !

Note on Partitioning drives

Because a hard drive is made up of disks that spin, different areas of the disk can be much slower than others. When data is written to a HDD the drive starts at the outer edge, then as the drive fills up makes its way to the inner tracks... Ever seen the wheels on a car turning ? If you put wheels with a smaller diameter on your call the wheel has the spin much faster to travel the same distance. This is common sense and the same applies to a hard drive. If you partition a drive in half you have just given the fastest part of the drive to your system and the significantly slower part of the drive to your audio files. There are advantages to partitioning, such as the drive does not become fragmented as fast. However the increased head-movement which translates into a larger SEEK means its not as optimal as having a dual hard drive setup.

How to setup your drives for the best performance

If you have read the above topics you can probably work it out for your self...Remember the draw backs of IDE ???

There's two rules to follow when setting up hard disks.

1. Never let the recording disk share an IDE cable with a CD ROM.

2. Don't let the recording disk share with a slower DMA mode hard drive.

There's two commonly recommended ways to set up the drives on an IDE cable.

1. Put the 2 hard drives on the same ide cable and any cd-roms on another cable.

2. Put the OS hard disk on the ide cable with a cd-rom or burner and have the recording drive on its own unshared cable.

Both methods will work and both have their own situations when they are best used. However in a system where There's a dvd rom and a burner with 2 hard disks (4 ide devices) your best to follow the first way. Putting the 2 hard disks together in this situation is by far the best configuration because CACHE means that the OS drive wont be accessed or used much during recording and playback leaving the full bandwidth of the ide channel to the audio drive. If you don't know what cache is then read here. It may seem tempting to install 2 ide controllers so every device has its own channel but this then takes up bandwidth from the pci bus which is worse in most cases then setting it up as per the above advice. Serial ATA will change this when it becomes more common in computers.

There's a lot of other things to consider when setting up hard drives which I wont go into as it will become outdated in a few months time with SATA. As its always changing its yet another reason why to contact someone like myself to setup your computer for audio or better yet have a custom made DAW made for you with all this pre done.

Which drive do I install my programs to ?

The operating system and all other programs MUST go on the one drive leaving the second drive free 100% for files that are recorded and played by the software. So in most computer c: will have the windows installation and all the programs like cubase, logic, protools ect and the only thing on the second drive will be the WAV files and the projects saved file that you open in your software. Other files which are created on the second drive when recording are meant to be there as well. So a good rule to follow is if the file is created when you record it should be on the 2nd disk.

General background info

TRANSFER RATES

First of all a ATA100 hard drive cannot sustain the theoretical transfer rate of 100 MB a second....When buying a hard drive look at the AVERAGE SUSTAINED / LOW PEAK transfer rate. This value will be a more realistic value to go by. Manufacturers very rarely print this spec on the product. Visiting the web site can usually find the figure or sometimes a phone call to tech support is needed. To have a DAW that can reliably do 16 tracks of 16bit / 44.1Khz audio your hard drive will need to be able to sustain 2MB per second of throughput or more. Modern Drives can supply 12MB / sec or more which can deliver approximately 60 - 80 tracks of audio at 24bit 44.1Khz. Most cd burning software can test a hard drive and give you the sustained average figure for your drive. The difference between a ATA33 and ATA133 is not as big as what you may think by looking at the name, in fact they may very well have the exact same sustained transfer rates in real life situations like recording.

SEEK TIMES

The other important spec on hard drives is the SEEK time... Once again this can be miss quoted...For an example a hard may have a seek time of 2ms between tracks that are next to each other but it has seek time of 28ms from one side of the disk to the other (full stroke seek). You will want to look at the AVERAGE SEEK...even here they can miss quote this figure. Look at the specs for my old Western digital 30 GIG drive below and you will see they quote 10.9ms typical and a maximum of 20ms !!!! I should point out that the Seek time is directly related to the SPINDLE SPEED so SCSI drives that have spindle speeds of 10,000 RPM can out pace an IDE drive. IDE 10,000 RPM drives are just beginning to be made now and will take ages before they are available at computer shops. SCSI is now at 15,000 and the high end SCSI drives are 20,000 RPM ! This is what you pay extra money for with SCSI. Lower seek times and a much higher sustained data transfer rate. These two things are the two factors that determine how many tracks you can record and the latency you can achieve (more on this below)...An IDE drive (with the same specs) can match a SCSI drive for short bursts but if you want to stream multiple tracks of audio this is irrelevant. You will also notice that read seek times are faster than write seek times, IDE can handle audio playback quite easily but when trying to get low latency recordings IDE can struggle on recordings .... Hence the drive buffer then comes into play.

Drive Buffers and Drive latency

When ever you add a buffer to a data path, the buffer will cause latency...Hard drives, even if the head is in the exactly the right place will still take some time before the data can be provided to the computer for use ! Seek times do not take drive latency into consideration and are not quoted with this figure included. Hard drives with larger buffers are not a bad thing as they are also much faster and use the buffer in a much shorter period of time.

Performance Specifications (my old recording drive which is a ATA100 7200 RPM 30 GIG WD drive)

Average Seek (Read) 8.9 ms typical, 18 ms maximum
Average Seek (Write) 10.9 ms typical, 20ms maximum
Track to Track Seek 2.0 ms typical, 5 ms maximum
Full Stroke Seek 21 ms typical, 28 ms maximum
Index Pulse Period 8.3 ms (nominal)
Average Latency 4.2 ms (nominal)
Rotational Speed 7200 RPM (nominal)
Controller Overhead 0.3 ms average

Transfer Rate (Buffer to Host)

100 MB/s (Mode 5 Ultra ATA)
66.6 MB/s (Mode 4 Ultra ATA)
33.3 MB/s (Mode 2 Ultra ATA)
16.6 MB/s (Mode 4 PIO)
16.6 MB/s (Mode 2 multi-word DMA)

Transfer Rate (Buffer to Disk) 244 Mbits/s minimum
400 Mbits/s maximum
Interleave 1:1
Buffer Size 2 MB
Error Rate (Non-Recoverable) <1 in 1014 bits read
Spindle Start Time
- From Power-on to Drive Ready1 5.5 s typical, 9s maximum
Spindle Start Time
- From Power-on to Rotational Speed2 4.5 s typical, 7s maximum
Contact Start/Stop Cycles (CSS) 40,000 minimum