DAW performance factors
Most stand-alone DAWs have a fixed architecture. They will record a certain number of tracks at a given resolution and with a certain number of effects plug-ins. Their abilities and maximum performance specifications do not change. This is not true of DAWs built around personal computers. Here, there are many factors that influence performance, sometimes drastically. These include such things as operating system configuration, CPU type and speed, audio software, RAM type and speed, audio interface(s) used, system bus, cache type and speed, drivers, plug-in types, drive characteristics and even user preferences and working methods. So let's look at these issues to get a better idea of what to expect from these systems and how to get them working to their fullest potential.
Before getting started it should be noted that it is essential to check compatibility and system recommendations for each of the technological elements mentioned above. Look for the latest information on the audio manufacturers' websites and read the PDF manuals or other documentation, as available, as well as reviews, user forums and recommendations of industry pros. A thorough job of matching known good hardware configurations by respected DAW builders can go a long way towards avoiding serious headaches later.
Also note that computers, operating systems, and recording software change so quickly, some of the information I present here is likely to be outdated nearly as quickly as I can write and post this column so, again, please check this against other current sources….
For many of these reasons just noted, and to help ensure that a computer-based DAW works as consistently as a stand-alone unit, many professional studios are purchasing complete turnkey systems. These are fully configured and tested, and often include preconfigured audio interfaces, operating system, drives, DAW software, and plug-ins. Some companies are much more thorough than others, so be sure to choose a reputable one that offers good support, guaranteed products, and warranty services.
The big OS
Once upon a time, the Mac was the only real choice for professional-level audio on a personal computer. Now, powerful DAWs are often found built around either Mac OS or Windows systems (though powerful UNIX and Linux based systems are also available, as well as specialized iOS and Android ones…). In all cases, it is a good idea to tweak the system to work best for your needs.
For truly reliable, high-performance studio DAW solutions, I generally recommend streamlining systems and applications for use with studio tasks only. No games, no word processing, no e-mail, Twitter, Skype or Facebook… and (for goodness’ sake) no surfing!!! I know that having such a dedicated machine may not be possible in many project studio situations, but it sure is the safest and most stable way of working….
Another important working method is to not allow anything else to run simultaneously with your DAW software. Similarly, do not allow any scheduled tasks to run. (You know… when the message “would you like to update (or register) software x now?” pops up when you’re mixing.) Things such as virus scans, drive optimization, and system backups are all necessary tasks, but should be triggered manually rather than automatically. If they must be run as automatic scheduled tasks, then be sure to schedule them only for times when you know you will not be working!
Malware, such as viruses and spyware, can significantly reduce system performance (reducing track and plug-in counts while increasing crashes) or even induce catastrophic failures (causing the loss of your precious tunes!). Keeping your DAW exposure to the Internet to a minimum is a good start, but good up-to-date virus protection software is also a must. All new media should be scanned when it is added to the system, and a thorough scan should be made once every week or more.
Viruses aren’t only transmitted over the web, however; you can infect your computer from a compromised removable disk or over a local network that has infected computers on it. The likelihood of this can be significantly reduced by combining good anti-virus software and a hardware firewall/router between your local network and the Internet.
You can scan your computer for malware or free using any of the following:
- Norton Security Scan
- Panda ActiveScan
On Mac OS:
CPU type and speed
Though a lot is made of clock speed, it is only one element influencing the actual performance of a CPU. Other factors, such as memory latency, number of processors, multithreading, frontside bus speed and/or HyperTransport speed, bit depth of the recorded audio, hard drive and interface speeds, driver quality, and motherboard quality can be just as important.
Hyperthreading, is one way to get more from a processor by allowing it to perform two threads of code simultaneously. While this feature does not make a single processor work as well as a dual- or multi- processor system, it can moderately increase potential performance. Not all applications can or will take advantage of hyperthreading and, in some instances, it can even hinder performance. To be sure, you should test the speed of your application with it turned off, then on. Leave it in the state it works best.
Another primary factor in actual PC performance is the frontside bus speed (aka memory bus). This determines how fast the CPU can move data to and from system RAM. Basically, if a CPU can perform faster than it can receive or communicate new data, that speed differential is essentially wasted.
Cache memory is smaller, but faster than main RAM. It is used to store most recently accessed data, instructions, and registers, for quicker CPU access. Level 1 or L1 cache is typically very small and located directly inside the CPU. Level 2 (L2) cache offers another level of quick access memory that is larger than the L1, and a third L3 cache may even be included. These caches and their implementation can have a great impact on system performance, especially with multi-core processors.
The slowest type of cache is actually not RAM memory, but a small amount of hard disk space used by the operating system, called the swap file (or page file) in Windows. It is a much slower type of virtual memory, and its use should be minimized on a DAW system by utilizing plenty of real RAM and limiting the number of applications running simultaneously.
When it comes to RAM, more is better (up to the max your operating system and applications can support). When it comes to getting the best performance from your computer, however, RAM type, speed and the depth at which data is communicated between the RAM and CPU (and so the true speed of processing) are also major factors. As previously mentioned, performance is effected by the memory bus speed, bit-depth, and queing/scheduling. The memory itself should be appropriate to this bus and take advantage of its full capabilities.
Current faster RAM types include DDR (Double Data Rate), DDR2, DDR3, each allowing for faster and faster speeds, up to 1333MHz. These are based on DIMM (Dual Inline Memory Module) configurations.
SODIMM (Small Outline Dual Inline Memory Module) is made specifically for laptops, small-footprint PC’s and some printers. It is currently offered in SDRAM, DDR, DDR2, and DDR3 types.
To get more memory bandwidth, some systems use dual-channel memory. This allows two memory pathways to work in parallel, theoretically doubling the possible bandwidth. This technique can be done with any type of RAM, depending on your motherboard/memory controller. Some types of RAM are specifically designed for this type of function and sold in pairs to do so.
Some RAM is designed to protect against errors, crashes, and loss of data. Most of the memory in personal computers is non-parity memory, which offers no real protection in the case of any errors. Parity memory, the next step up, can detect errors and communicate with the operating system regarding what has happened. ECC (Error Correction Code) is more expensive, but can actually correct minor errors on the spot rather than just report that they have occurred. To use ECC memory, your motherboard and OS must support it.
The size (capacity) of a drive is simply how much data it can hold. By itself, it really has nothing to do with determining the drive’s performance. Don’t let those big Gigabyte, or even Terabyte, numbers make you blind to other factors. That being said… with drive capacity, like the amount of RAM, more is better.
The following numbers are supplied to give an idea about the amount of drive space needed when recording audio under various conditions.
This is how much room is needed for 1-minute, mono audio tracks at the following specs (sample rate/bit-depth):
- 44.1 kHz/16-bit = 5.3 MB
- 88.2 kHz/16-bit = 10.6 MB
- 44.1 kHz/24-bit = 7.9 MB
- 88.2 kHz/24-bit = 15.9 MB
- 192 kHz/32-bit = 46.1 MB
Four-minute 24-track tunes (all mono with no overdubs or other virtual tracks, and before mixdown) recorded in these formats would take up approximately the following space:
- 44.1 kHz/16-bit = 508 MB
- 88.2 kHz/16-bit = 1 GB
- 44.1 kHz/24-bit = 758.4 MB
- 88.2 kHz/24-bit = 1.5 GB
- 192 kHz/32-bit = 4.4 GB
It is not uncommon for album length “project studio” productions to take up as much as 150GB at the higher (not highest) resolutions. Having at least 250GB of record space is a good minimum these days. It will let you work on an entire project without having to back-up and retrieve data in the middle. This does not mean that you shouldn’t back up your work; back it up for safety, not because you must make room to work!
If video is involved, especially in HD, drive space become an even greater concern and a Terabyte should be considered a good minimum working size….
Drive performance factors
Again, size is only about storage capacity. There are a number of other factors which determine a drive’s performance (and, therefore, the number of simultaneous tracks it can play and/or record). A new wrinkle in the drive market is the emergence of Solid State drives (SSDs). Solid State drives offer significant performance improvements over traditional hard disks (HDDs), as well as improved reliability, ruggedness, weight and life expectancy. The major downside to these is their capacity for the price. In addition, the cost per MB rises with capacity in SSD drives, while it greatly decreases in HDD models. For example, at the 250MB mark, solid state drives are around 4.5 times the cost of hard disks: and while HDDs can now be found topping 1TB for just over $100, a 1TB SSD would cost closer to $3,000. That’s 30 times more expensive!
Due to the market pressures of such a large cost differential, a third option is quickly becoming popular, the hybrid drive. While retaining the higher performance and speed of the SSD, hybrid drives cost closer to around twice as much as an HDD and come in the large capacities typical of current HDDs.
In both hybrid and HDD designs, some of the important specs to consider include buffer size, seek time, access time, rotational speed, and interface. A hard drive includes a small amount of RAM to use as a buffer, to keep the flow of information in and out as smooth as possible while the drive’s read and write heads do their work. Buffer size varies by make and model. Most drives will show increased data transfer speeds with larger buffers, especially when reading multitrack audio or other non-contiguous (or fragmented) files.
The seek time is the average amount of time it takes the drive head to move to the location on the platter where a requested file is located (a process called seeking). It is determined, in large part, by the speed and inertia of the head, as well as the rotational speed of the platter (more on this in a minute).
Appropriate seek times for audio production run in the range of 3 (super fast) to 8-10 (decent) milliseconds (ms). While the difference between 5.3 ms and 8 ms may not seem that great, it can make a big difference in actual performance, as the effect is cumulative. An 8 ms seek time is 50% greater than 5.3 ms and may take up to 50% longer to accomplish the same task. This can be especially problematic with multitrack productions, when using lots of edits, or when a drive becomes fragmented (where a single data file is stored across non-adjacent sectors of the drive platter; when fragmentation occurs, you should run software to defragment the drive, but it is recommended you backup all of your data first.)
The term access time is often, erroneously, used interchangeably with seek time. Access time, the time to actually obtain data from the disk, includes seek time plus rotational latency – how long you must wait once the head’s in place for the disk to rotate to the right point to read the data you need.
Rotational speed, the speed at which the hard disk platter spins (given in rpm, rotations per minute) determines how quickly a particular point on the platter will pass underneath the path of the head that reads and writes data. As on a CD or record player, the device that reads the stored information moves only across the radius of the disk along a fixed line. This is why the media must itself move, so that all data is accessible to the heads. Once the proper point on the drive has been cued up, the amount of data accessed or recorded per second is also greatly influenced by rotational speed. So, once again, faster is better.
Though 5400 rpm drives are popular in inexpensive machines and laptop computers, that speed is not recommended for serious multitrack audio production. 7200rpm should be considered the minimum. Some drives are available that run at 10,000 or even 15,000 rpm, but they’re quite expensive and may be restricted to a certain interface type your computer might not have. Speaking of which….
The interface type (the connection and data protocol for transferring data to and from the drive itself) is important in determining the maximum amount of data that can be transferred from one or more drives over a single bus.
Here’s a list of some of the current faster interfaces used for multitrack audio recording and production work. For working at 24 or more tracks in 24 bits (and especially at higher sample rates) interfaces slower than these should be avoided.
ATA (Advanced Technology Attachment) – drives under this standard are also referred to as EIDE (Enhanced Integrated Drive Electronics):
- IDE/ATA 66 – 66 MB/sec
- IDE/ATA 100 – 100 MB/sec
- IDE/ATA 133 – 133 MB/sec
SATA (Serial Advanced Technology Attachment):
- SATA 1 – 150 MB/sec
- SATA 2 – 300 MB/sec
- SATA 3 – 600 MB/sec
- eSATA – 300 MB/sec
- eSATAp – 300 MB/sec
SCSI (Small Computer Systems Interface):
- Ultra 160 – 160 MB/sec
- Ultra 320 – 320 MB/sec
- Ultra 640 – 640 MB/sec
Universal Serial Bus:
- Firewire 400 (IEEE-1394a) – 50 MB/sec
- Firewire 800 (IEEE-1394b) – 100 MB/sec Firewire
- S1600 – 200 MB/sec Firewire
- S3200 – 400 MB/sec
- Fibre Channel (optical) – 1 GB/sec
- Fibre Channel (copper) – 400 MB/sec
Thunderbolt – 1.2 GB MB/sec
These figures are theoretical maximum data rates; in the real world your numbers will be significantly smaller. To achieve these levels of data throughput generally requires multiple drives (maybe even in a RAID array… more on this below).
Maximum bandwidths, however, can be misleading… especially as regards FireWire, Thunderbolt and both USB 2 and 3. While these busses are capable of very fast data transfer rates, that does not mean that individual drives using these interfaces will perform at anywhere close to these rates. The performance of each drive is still subject to all of the previously mentioned factors. Putting a USB 3.0 connector on a 5400 rpm drive does not turn it into a fast drive. It just means that its slow performance is not the fault of the interface!
Drive data transfer rates
The term data transfer rate is often used in published drive specs to mean the maximum interface speed. Again, don’t be fooled into thinking this has anything to do with the actual (real-world) data throughput performance of the drive. A slightly less misleading term sometimes used for this is external data transfer rate.
A more meaningful number can be found in the internal data transfer rate. This is a measure of how fast single-channel contiguous data is read, once reading has begun. It is an absolutely best case scenario that no audio session will realistically see. Multitrack audio production entails a lot of seeking, which constantly interrupts this type of data transfer. This makes the actual average data rate way way slower. What these numbers do give is a relative scale for expected real-world transfer performance. In other words: a larger number is faster (is better). Common rates range from 60-80 MB/sec for decent drives.
Oh no… it’s a RAID!
Multiple hard drives are a good idea when working with computer audio. Think about it: you have to access your system software, run your DAW software, and record audio while playing back other audio, all at once. Spreading this work over multiple drives greatly increases your efficiency.
(Note that it's not vital to do this - many laptop users can accomplish a fair bit with a single drive. But it's a definite help in almost any system to have multiple hard drives.)
The usual recommendation is that a studio computer should always include at least two drives. One drive should be dedicated solely to the operating system and software applications, like your DAW, while the rest are for audio files. In addition, extra inexpensive large capacity USB or FireWire drives are great to have around for backups.
In addition, if you plan to use virtual instruments that rely on large sample sets or frequently use sound libraries, it’s a good idea to have a third hard drive especially for those. If you can't, you'll probably have to store the libraries on the same disc as the applications, which will not be as bad as storing your audio tracks there, but still isn't as efficient.
RAID (short for Redundant Array of Independent Disks) uses multiple disks working in tandem. The idea here is that two or more disks working together on a single task will outperform a single disk. Different RAID types incorporate the two disks in various ways: striping, which writes part of a data stream to one disk and part to another for faster read/write access, and mirroring, which writes a data stream simultaneously to two or more disks for redundancy (and faster reading once it’s written).
RAID comes in various types. RAIDs 0 and 1 are the most commonly found in desktop PC and some project recording studios. Though there are software-based RAID solutions, performance and reliability are both significantly increased by use of a dedicated hardware RAID controller because the task of controlling the array isn’t forced on your CPU. Several of the major PC manufacturers include RAID options in some of their basic system configurations using the internal drives.
RAID-0 is both fast and efficient because it uses striping to spread data across two or more drives. This significantly reduces the load on each drive of the system, but offers no built-in protection to data loss from drive failure. This is a particularly important point: since data is written across multiple drives, the failure of any one drive can be catastrophic. When using this type of array, it is recommended that all data be backed up, off of the array, frequently.
RAID-1 offers protection against data loss by mirroring: simultaneously writing the same information to two or more disks in case one fails. Compared to using a single disk, performance is slower when writing information, but faster when reading.
RAID-10, or 1+0, combines mirroring and striping, striping data across two pairs of mirrored devices. You have the speed of striping and the redundancy of mirroring, but require (at a minimum) four hard drives for only two drives’ worth of storage, so it's the most expensive RAID type.
There are lots of other RAID types that are either obsolete or irrelevant to studio applications. You'll occasionally see RAID-5, which works best with multi-user situations – though it can work with three drives, more drives are often used.
Hot and Noisy…
A couple final factors when choosing drives for the studio are noise and heat. Fans, heads, and spindle motors can all create noise. If the drives must be located in the control room, they (or the whole computer) should be chosen for quietness and enclosed in a sound isolating space. Do not place drives in the actual recording space or their noise will be picked up by the mics and end up in your recording.
At the same time, the heat from the computer and drives must be dissipated, so do not be tempted to unhook the fans! If you need to, you may be able to replace the fans with quieter models. Wherever the computer and drives are located, be sure they do get air circulation to help them stay cool.
Well, that covers quite a bit about basic computer hardware. Next time we’ll look further into getting the software tuned up to work its best and talk about actually using these darned DAWs.
- John Shirley is a recording engineer, composer, programmer and producer. He holds a PhD in music composition from the University of Chicago and is a Professor in the Sound Recording Technology program at the University of Massachusetts Lowell where he serves as chairman of their music department. You can check out some of his more wacky tunes on his Sonic Ninjutsu CD at http://www.cdbaby.com/cd/jshirley.