Whether you’re laying down audio tracks, hosting multi-gigabyte sample libraries or working with humungous 4K video files, chances are you’re going to find yourself straining against the limits of your storage space soon enough. These days there are several options for expansion, and this article is designed to give you a quick overview, with an emphasis on what’s most important for the modern music or audio production professional.
If you’d like a quick summary of how best to kit out your new OPUS 101 audio PC, skip to the bottom of this article for our recommendations.
Table of contents
The Workhorse: SATA 2.5″ SSDs (Solid-State Drives)
Solid-State Drives (or SSDs) offer many key advantages over the traditional spinning drives (“hard disks” or HDDs — see below) that used to be the mainstays of our computers, and these advantages are especially keenly felt by those of us working in music production.
The first and most significant advantage is speed. For composers dealing with sample libraries and large DAW templates, streaming samples to memory from a traditional spinning HDD can prove a significant bottleneck. To understand why, it’s important to understand how SSDs physically differ from HDDs.
An HDD consists of a spindle which holds multiple circular disks called “platters”. These platters spin around past a read-and-write head, which stores and retrieves data. To read data from the disk, the read-and-write head needs to move into the correct position and then wait for the disk to spin around to the right sector, where the data it wants to retrieve is located. The speed at which a disk can spin (its RPM) is the main factor that determines how fast the drive is.
These sorts of drives are most efficient when reading data that’s all in the same physical location on the disk. Think of it like a turntable, where the HDD platters are an LP and the read-and-write head is a stylus attached to a tone arm. Your life is going to be a lot easier if you can just drop the needle at one point on the record, sit back and enjoy those sweet analogue sounds. That’s what we’d call a “sequential read”, because all the data you are pulling off the spinning disk is all sitting in the equivalent of a single groove.
But imagine your data is scattered in tiny chunks all over the surface of that record. In order to read this data, the read-and-write head has to move, wait for the disk to spin to the right spot, read a block of data, then move again, wait another revolution, read the next block of data, and so on. It’s hopelessly inefficient.
Unfortunately, this is how samples are organised. Every sample — meaning every round robin of every articulation of every note in every dynamic layer — is a block of data. And because most music isn’t just chromatic scales going up and down, it’s very unlikely your samples are going to be physically located close together when you ask your computer to read them off your disk. This is called a “random read”, and it’s the area where SSDs offer a huge advantage over HDDs.
An SSD has no moving parts. It consists instead of a series of cells that can each be set to different values. In the simplest types of SSDs — which use single-level cell (SLC) memory — each cell is either a 1 or a 0, on or off. Newer technology allows for multi-level cells (MLC) to store two bits, triple-level cells (TLC) to store three bits, and quad-level cells (QLC) to store four bits. (Each of these technologies increases the amount of data a drive can store for the price you pay, but may have speed and reliability trade-offs. All the SSDs we use in our builds are high-quality TLC drives, which we think is the best compromise.)
Because there is no read-and-write head and no spinning platters, an SSD is much more efficient than an HDD at finding random bits of data scattered around its surface. And by “much more efficient”, we’re talking in the region of 50 times faster. (This is all a bit of a simplification, of course, but if you’re curious, there is a useful discussion of the three factors — throughput, IOPS and latency — that affect speed differences between HDD and SSD over at The SSD Review.)
The long-and-the-short of it is that if you have a lot of sample libraries, loading them into memory or (especially) streaming them direct from disk is going to be significantly faster using an SSD than it would be with an HDD.
But that’s not all — there are several other factors that make SSDs a much better choice for audio professionals than HDDs. For a start, because there’s no spinning, there’s no noise. They also tend to be more reliable. And the lack of friction means they consume less power and run much cooler than HDDs. That doesn’t mean there is no place for HDDs in the studio (see below) but it does mean that your bread-and-butter storage solution should be SSDs.
The vast majority of SSDs, like the one pictured, are properly called SATA 2.5″ SSDs. SATA is a type of bus interface that connects your storage device to the rest of your system, while 2.5″ relates to the physical dimensions of the drive. These are the cheapest forms of Solid-State Drive you can get, and also what most people mean when they talk about an “SSD” — even though, technically speaking, an NVMe drive is just another type of SSD. On which note, let’s move on!
The Racehorse: NVMe (Non-Volatile Memory Express) Drives
Non-Volatile Memory Express (NVMe) is a newer technology that has only recently gone mainstream. An NVMe drive is essentially an SSD with the brakes taken off. Whereas traditional SSDs like the ones mentioned above are bottlenecked by using the same relatively slow SATA bus used by old-school HDDs, an NVMe drive transfers data using your motherboard’s PCIe bus. This is the same bus used for very high throughput components such as graphics cards, and is capable of insanely fast read/write speeds in comparison to SATA — we’re talking up to six or 12 times faster depending on the PCIe generation.
The principal way in which NVMe technology differs from SATA technology is through its use of multiple queues. Traditionally, commands to a storage device (such as “read this block” or “write that block”) are held in a single “queue” by your host controller, and commands in this queue are processed one at a time. This is because the spinning disk drives SATA was originally designed to support have only a single read-and-write head capable of carrying out these commands. A SATA drive can hold up to 32 commands in this queue at a time, while SAS drives (used in data centres) can hold up to 254 commands.
But when there is no spinning disk and no physical read-and-write head involved, as is the case with SSDs, there is no longer the need for all these commands to be executed one after another. NVMe takes advantage of this fact by offering not just one queue, but a staggering sixty-four thousand queues, each of which can hold sixty-four thousand commands. Because each queue can be processed independently of the others, reading and writing data is orders of magnitude more efficient.
So does this translate into a real-world performance improvement for an audio PC? Yes and no. Not all software is written in such a way that it can fully take advantage of NVMe technology, and this is partly true of samplers like Kontakt. Nevertheless, even with Kontakt’s somewhat inefficient design, an NVMe drive can load samples into memory noticeably faster than SATA SSD — and if you run a purged template you should also see a substantial boost to direct-from-disk (DFD) streaming.
You may sometimes see NVMe drives referred to as m.2 drives, but this is a bit of a misnomer: m.2 in fact refers to the form factor of the drive — a thin, flat rectangle with some pins at one end and a semi-circular mounting hole at the other. Good old-fashioned SATA SSDs can come in this form factor too, so make sure you specify “m.2 NVMe” when searching for a new drive, as there’s little point in using one of the two or three precious m.2 slots on your motherboard for a drive that’s still bottlenecked by SATA limitations. (All the m.2 drives we stock at OPUS 101 are NVMe drives and vice versa.)
To complicate matters, there are multiple types of NVMe drive — because, as PCIe bus technology has advanced in recent years, so too have NVMe drives. PCIe Gen 3 drives are capable of sequential read speeds of up to ~3,500MB/s (vs ~560MB/s for SATA SSD — while PCIe Gen 4 drives are double that speed again, topping out at a whopping ~7,000MB/s. As explained above, sequential read speeds aren’t massively relevant to streaming sample libraries, but depending on how many cores your CPU has, you can see dramatic increases in random read speeds with these drives as well. All OPUS 101 machines can take either PCIe Gen 3 or PCIe Gen 4 drives, but the latter do of course carry a higher price tag.
Good for the Knacker’s Yard? HDDs (Hard Disk Drives)
While we don’t recommend old-fashioned hard disk drives (HDDs) for either system disks or sample libraries, the one thing they have in their favour is price: you can buy a multi-terabyte HDD for a fraction of the cost of a multi-terabyte SSD. This can make them useful for backing up your important files, e.g. project files or your system disk. You don’t want to be reading from them or writing to them constantly because they’re slow and noisy by comparison to SSD drives, but running daily backups is always a sensible idea and this can be done while you sleep.
One word of caution when selecting an HDD, however. In order to maximise space on the spinning platters of an HDD, many manufacturers now use a technology called Shingled Magnetic Recording or SMR. Unfortunately these drives are more prone to error than the more traditional Perpendicular Magnetic Recording (PMR) drives, also known as Conventional Magnetic Recording (CMR) drives, so if you’re mainly using an HDD for backups we’d steer clear of them. For a while it wasn’t obvious which drives used which technology, but consumers have got wise to the distinction recently and it is now fairly trivial to find out which manufacturers uses what for their different drives. You will tend to find SMR more commonly in large-capacity 2.5″ SSD drives, such as the ones you find in laptops, while the bulkier 3.5″ drives are usually (but not always) PMR/CMR. Google the model you’re interested in and the term “CMR” or “PMR” to be sure.
Which Drives to Use for What: Our Recommendation
Depending on which motherboard you opt for, your OPUS 101 system will have six or eight SATA connectors, which can take either SSDs or HDDs, and two or three m.2 slots, which can take NVMe drives. All our systems come with two drives as standard: a system disk (for Windows and your applications) plus a storage drive (for sample libraries, projects, or both). So what type of storage should you use for which purpose?
System Disk: A standard 2.5″ SATA SSD makes an excellent system disk. They’re fast, silent and reliable. A 500GB drive will be good enough for most people’s Windows installation, programmes and plugins. While an NVMe drive will be faster still, in our opinion the difference isn’t so huge in terms of productivity gains that it’s worth sacrificing one of your m.2 slots for.
Sample Libraries: The faster you can stream samples from your disk, the better performance you’re going to get. A fast SSD means you do not need to allocate as much RAM for preload buffers in Kontakt, for instance. Once again, a SATA 2.5″ SSD makes a great choice here — but for very big libraries or libraries that otherwise take an age to load, you may want to consider hosting them on one or two m.2 NVMe drives. Whatever you choose, make these drives as big as you can afford — because sample libraries, as we all know, are an addiction for which the only remedy is more storage space!
Projects and Documents: Once again, good old SATA 2.5″ SSDs are ideal for hosting project files and documents. You can store documents and projects on the same drive as your system disk, but in an ideal world you’ll want to have them on a separate disk just in case something goes wrong and you need to reformat your system disk. (This risk can also be mitigated with regular backups.)
Backup Drive: It’s never a bad idea to keep backups of your important files. While an HDD can be useful here, our preference would still be to use SSDs wherever possible, because they are less at risk of corruption or degradation. Better yet: use a cloud backup service like Backblaze or Dropbox and let them manage the risk for you.
So in summary: unless you have more m.2 slots than you know what to do with, use a standard SSD for your system disk, reserving your m.2 slots to host your biggest, slowest sample libraries on fast NVMe drives. All your other sample libraries can live on high capacity SSDs — as can projects, documents and (if you make them, which you absolutely should) backups.