You must have heard of hard drives (HDDs) or solid-state drives (SSDs). Some of you might have even used the term “memory” to call them both, and you’re sort of correct. But things are more complicated than that.
This post will briefly explain digital storage. Among other things, you’ll find the differences between HDDs vs SSDs, and between SSD variants, such as SATA SSD vs NVMe SSD.
When through, you’ll either get super-bored or know how to buy your next SSD (or hard drive) with confidence. Most likely, you’ll be somewhere in between.
Dong’s note: I originally published this post on February 22, 2018, and updated it on August 6, 2022, with additional relevant information.
Digital information in brief
We live in the Information Technology (IT) age, where we store and exchange information digitally. The reason it’s called “digital” is because the whole thing involves digits, two specific ones.
The binary way of information coding
In the computer world, information is stored and processed via strings of zero (0) and one (1) in different patterns. This method of information coding is known as binary.
In a nutshell, you can understand binary as the count limit. Generally, in daily life, we all have taken for granted a system called decimal, in which we count from 0 to 9. We put some of those digits together for any higher value, such as one and zero for 10 or two and one for 21, etc. With binary, we count from 0 to 1. So to convey two and more, we have to put 0 and 1 together in specific patterns.
The more information we need to code, the longer these strings are. And they get long really fast.
For example, “Dong Knows Tech” (no quotes) is a short phrase of text, but to convey it in binary, we have the following string when using a simple text-to-binary software:
01000100 01101111 01101110 01100111 00100000 01001011 01101110 01101111 01110111 01110011 00100000 01010100 01100101 01100011 01101000
The spaces between eight-digit groups are there only to make the whole string less mind-boggling.
In other words, that’s how the computer “sees” that phrase. And if you change the color of the text or use the site’s logo up top, the string will get even longer, and we’ll need different software to code and decode that.
More information requires more zeros and ones of more complex patterns. But there are infinite possibilities. We can use those two digits to describe everything as long as we have the appropriate software for the job.
Using binary to communicate is impossible for humans — we’d fall asleep. But for the machine, that’s the most efficient. It has only two possibilities (zero or one), and an average computer can handle trillions of binary calculations per second without ever getting bored.
Digitalization also makes information extremely compact. An SSD — as big as a little finger physically — can hold the amount of data equal to an entire average-size library’s books.
And it’s much easier to send digitalized information than shipping actual books — we can move information around much faster.
Digitalization means we use computer software to convert information into binary for easy storage and exchange, and have it translated back into texts, images, or videos for us to consume on demand.
But before that, the information has to reside somewhere. Specifically, before you can read this beloved “Dong Knows Tech” phrase, the browser has to know where to pull that crazy string of digits above. And that brings us to “memory” and “storage” and their differences.
Memory (RAM) vs storage (HDD/SSD)
A computer has two hardware parts that hold information: RAM, short for Random Access Memory, and a storage device, often a hard drive or a solid-state drive.
RAM: It’s volatile
RAM holds information that’s being processed. Everything you see on the screen right now is floating within some RAM stick. The more RAM a computer has, the more information it can process simultaneously.
By nature, RAM is speedy compared to general storage, but there is RAM of different speed grades and standards.
The good thing about RAM is that it allows the information to be flexible and malleable. When you scroll a web page, the text and pictures move up and down instantly, and with particular software, you can make changes to a photo, a movie, a document, etc.
In return, everything in RAM is temporary (or volatile) — it needs electricity to keep the information “alive”. If you unplug the computer, information in RAM is lost, which is why you need to save your document before turning off the computer.
And saving a document means you move the information from RAM to a storage device.
Storage: It’s persistent
There are many forms of digital storage — HDDs and SSDs are the most popular — and they all share this attribute: Non-volatility.
Specifically, information stored on an HDD or SSD remains even when you unplug the device and put it away. The next time you turn the computer on, you’ll find the previously saved document intact.
A computer’s storage device holds more than just your documents. It also contains the operating system and all software applications. When you turn the machine on, it loads all that into the memory (RAM) and creates a virtual environment where you can send/receive, consume, or manipulate information.
Digital storage is like books, magazines, or scrolls — the information they contain is always there.
RAM is the part of our brain where temporary ideas, feelings, or imaginations float as we experience the world through our senses, like reading or book or viewing a painting. To keep those ideas alive, we need to write them down.
The faster the storage device, the less time is needed for the software to get ready. And the higher the capacity, the more information a digital storage device can store.
And that brings us to hard drives (HDDs) and solid-state drives (SSDs), the two most popular forms of digital storage.
HDD vs SSD: Similarities
A solid-state drive and a hard drive behave the same from a user’s perspective: they store information and make it available when a user needs it.
HDDs first became available in the late 50s and have undergone many changes. SSDs — the flash-based type we use — have been widely known to consumers since only about 2010.
The popular SATA connection type
In the beginning, an SSD is the faster alternative to an HDD.
And that’s because they initially are based on HDDs by sharing the same interface called serial AT attachment (SATA). This allows an SSD to replace an HDD seamlessly — without extra hardware or software requirements.
The SATA interface has two prominent designs: 3.5-inch (for desktop computers) and 2.5-inch (laptop). You’ll find SSDs only in the latter form, but they work in all applications where a hard drive fits.
SATA is currently in its third revision (SATA 3), with a top speed of 6 gigabits per second (Gbps) or 750 megabytes per second (MB/s).
The future of SATA
There will not be a newer and faster version of SATA in the future. That’s because SSDs have been increasingly popular and more affordable, and they have unique designs and interfaces that deliver significantly higher speeds — more below.
As SSDs’ speeds get faster, the SATA connection standard is the bottleneck, and the growing popularity of SSDs will eventually render SATA obsolete.
Since 2018 there have been more and more notebooks without SATA support, opting for much faster interfaces explicitly designed for SSDs and other high-bandwidth applications.
HDD vs SSD: Differences
First and foremost, SSDs are significantly faster than HDDs. It’s safe to say that even the fastest hard drive is slower than the slowest SSD.
On average, you can expect a SATA SSD to have at least twice the sequential (copy) speed of a SATA hard drive. But SSDs’ strong point is in the random access performance, which contributes more to the general performance of a computer.
And on the inside, an SSD and an HDD are different in principles. Let’s start with HDDs.
Hard drives: Mechanical machines
Open up a hard drive; you’ll find a few platters stacked on top of one another like a spindle of compact discs. The more platter a hard drive has, the more information it can store.
Each platter contains multiple circular paths in concentric rings, called “tracks”, on its top and bottom surfaces. These paths can be magnetized by the drive’s read/write head — which hovers on top of each platter — into binary patterns to store information.
This process is the same as writing on a portion — called sector — of the platter for the first time or any subsequent times (overwriting). In either case, the drive magnetizes the sector directly into the patterns required at the time of writing, regardless of their state.
When you “wipe” a hard drive, you instruct the computer to write “zeros” to all its platters’ surfaces.
A hard drive’s reading/writing speed depends mainly on how fast it can spin its platters. Most consumer-grade HDDs spin at either 5,400 or 7,200 rounds per minute (RPM). High-end hard drives can spin at 10,000 RPM.
Even though hard drives are on the way out among consumer-grade applications, each of them is a mechanical marvel on the inside.
Specifically, the read/write head hovers within a few nanometers above the platter without ever touching it.
If you enlarge a hard drive to the size of a football field, the platter would be the game area, and the head would be the size of a golf cart. Now imagine that little cart is moving at about a million miles per hour, just a hair above the ground.
It’s amazing that it doesn’t crash every few minutes.
Extra on HDDs: SMR vs CMR
As mentioned above, the more platters a hard drive has, the more information it can store.
Generally, storage makers achieve higher storage capacity of hard drives by increasing the number of platters and the density of each platter’s track.
This type of data recording is called “Conventional Magnetic Recording” (CMR) — some companies might call it “Perpendicular Magnetic Recording” (PMR).
With CMR, each data track on a platter doesn’t overlap the adjacent ones — there’s a gap between them.
CMR allows for straightforward recording. It generally fills up each platter the drive track by track, which is simple and fast. But we lose some space for the gaps between tracks.
Since 2015, there’s been a new way of recording on a hard drive called “Shingled Magnetic Recording” (or SMR).
The name comes from the fact with SMR, each data track overlaps its previous one, similar to shingles on a roof, eliminating the gaps between tracks and increasing the density of each platter a great deal. As a result, SMR hard drives are affordable.
In return, when the computer writes to an SMR hard drive, the drive picks where it wants to write randomly. After that, when idle, it reorganizes the data.
An SMR drive is much slower than a CMR counterpart when you write a lot of data simultaneously. It also works all the time — even when the system is idle — meaning it won’t last as long.
Solid-state drives: It’s a complicated world
An SSD has no moving parts. It’s an integrated circuit that uses less energy yet can provide much faster access to the information it stores than an HDD.
It also can come in much smaller physical sizes and doesn’t necessarily need to conform to a standard design. Still, most do anyway for ease of use, though many also come in a proprietary form that fits only in a particular application — like a smartphone.
The world of SSDs can be quite complicated due to the nature of the technology. The first thing you should know about them is that you can only write so much to an SSD.
Finite P/E cycles
That’s because there’s no “overwriting” with SSDs.
Except when the SSD is brand-new, writing to an SSD always means erasing existing information from memory cells and programming them with new binary patterns.
For this reason, writing to an SSD is often referred to as a program/erase (or P/E) cycle. It’s pretty similar to writing on a whiteboard with a Sharpie — unless the board is brand-new, you’d always need to clean it before you can draw something new.
And just like a whiteboard that won’t last forever — eventually, you’ll wear out its surface — an SSD also has a finite number of P/E cycles. You can only program a memory cell so many times before it wears out and becomes unreliable.
And programming information on an SSD is a complicated and inefficient process.
Blocks and pages
An SSD organizes its memory cells in pages and blocks, with one block containing many pages. Here’s the problem: An SSD writes page by page but erases block by block.
As you can imagine, if you want to write more pages to a half-used block, the SSD will first need to copy the good pages (those with valid information) to a different place, erase the entire block, then write back the good pages together with the new pages onto that block.
As a result, an SSD has to write more, even a lot more, than the amount of information you need to store, which brings us to half a dozen interesting technical terms.
To understand SSD well, you must have all of these terminologies in your vocabulary.
As mentioned above, the term refers to the phenomenon that an SSD has to write more than the amount of information the user needs. The higher the write amplification, the shorter the drive’s life span.
A process where an SSD needs to reallocate pages of a block before erasing the entire block so it can write on that block, as described above.
The more efficient this process is, the faster a drive can write.
An SSD can dedicate a part of its storage (typically 10%) to make garbage collection more efficient.
That’s like having an extra room in your house to store stuff temporarily when you need to do a major cleanup.
As a result, many SSDs don’t have full capacity. For example, 256GB or 512GB drives only deliver 240GB or 480GB of actual storage space.
TRIM is an actual command of an operating system and not an acronym.
The command notifies the SSD when a page of old data is no longer valid — the garbage collection will skip it during the reallocation. When enabled, TRIM helps reduce Write amplification a great deal.
Endurance and wear-leveling
Endurance is the amount of data you can write to an SSD before it becomes unreliable.
To increase usability, SSDs use wear-leveling, algorithms that make an SSD use up all of its memory chips, cell by cell, before the first cell is erased and written on again.
Consequently, the entire drive “wears” evenly. For this, SSDs with larger capacities generally have higher endurance than smaller ones.
Of all the items above, endurance is likely the most important since it decides the value of a drive. Let’s find out more about it.
SSD endurance is generally presented in two ways: Terabytes Written (TBW) or drive writes per day (DWPD).
TBW is the total amount of data you can write to an SSD over its life span before it becomes unreliable. The higher the TBW value, the better the endurance and the longer the drive will last.
It’s important to note that TBW can be misleading since larger capacities generally mean a higher TBW rating. For example, if you stack a 1TB low-endurance SSD against a 250GB high-endurance SSD, the former will have a higher TBW value.
For this reason, there’s another more consistent measurement for endurance.
Drive writes per day
DWPD is the percentage of a full drive writes per day over the SSD’s warranty period, which tends to be between three and five years.
For example, if a 250GB SSD has a 1 DWPD rating and a warranty of three years, you can expect to write up to 250GB daily and every day for three years. If the same drive has the DWPD value of .5, you can write 125GB to it per day, and so on.
The higher the DWPD value, the better the endurance, but remember that DWPD needs to be weighed against the warranty period. For example, a drive with .5 DWPD over three years has lower endurance than one with .4 DWPD over five years.
The endurance anxiety
Since the write on SSDs is finite, we tend to worry a lot about their longevity. In reality, though, there’s no need to worry at all.
Even though you can’t write to an SSD forever, with normal usage, it’d take many years to deplete even a low-capacity SSD’s P/E cycles. Most of us don’t write more than 10GB per day to a drive, and many days we don’t write anything.
That said, if you still want to extend the life span of an SSD even more, then reduce any unnecessary amount of writing you do to it. Considering SSDs have become more affordable, endurance is no longer a huge issue.
SSD types: NVMe SSD vs. SATA SSD
Over the years, there have been many standard designs (form factors) of SSDs. That’s not to mention most Apple computers use proprietary SSDs — the pool of SSD shapes and sizes so large that hardly anyone can remember them all.
The classification of SSDs is also confusing because the form factors (designs, physical shapes), the interfaces (how a drive connects to a host), and the speed standards overlap between different types.
For example, SATA is both an interface and a speed standard. And then, you have standard SATA and mSATA form factors. mSATA itself is also another interface variant of SATA. And higher-end NVMe SSDs come in different types, too. That’s not to mention there are proprietary SSDs, like those you find inside most Apple computers.
That said, strictly from the interface point of view, you only need to know two popular types: SATA and M.2. I’ll explain the form factors and their speeds.
The SATA standard has been in use for the past few decades and is currently at the third revision — SATA 3 — which has the cap speed of 6Gbps (or 750MB/s). In real-world usage, the fastest SATA SSD, after overheads, has the top sustained copy speed of around 550MB/s
SATA SSDs come in two main designs:
Standard SATA: Standard SATA SSDs share the same design as laptop (2.5-inch) hard drives though most are slightly thinner than 7mm, compared to 9.5mm of the HDD. Here are the current best standard SATA SSDs you can buy.
mSATA SSD: Short for mini-SATA, mSATA is a variant of standard SATA with a much smaller form factor and uses a different interface (called mSATA) to connect to a host. mSATA is a transitional form factor and has largely become obsolete since 2017.
M.2 is the latest interface with the highest number of design variants via different lengths and “module keys.”
- Length: Typically, an M.2 drive is always 22mm wide. Its length, however, varies from 42mm to 110mm. Most M.2 SSDs, however, use the 2280 (22mm wide and 80mm long) design.
- Module keys: This determines how an M.2 device connects to a host (like a computer motherboard). There are A-key, E-key, B-key, and M-key. The first two are used mainly in Wi-Fi, Bluetooth, and cellular cards; SSDs only use B-key and M-key.
|Module key||Measurements||Bus speed/interfaces||Usage|
|B||3042 (30mm wide + 42mm long), |
2230, 2242, 2260, 2280, 22110
|PCIe x 2 (up to 16Gbps),|
SATA (up to 6Gbps),
PCIe x2 SSDs
|M||2242, 2260, 2280, 22110||PCIe x4 (up to 32Gb/s), |
SATA (up to 6Gbps)
PCIe x4 (NVMe) SSDs
All variants of M.2 SSD have even smaller physical sizes than mSATA. So small that people tend to call M.2 devices “sticks” or “cards.”
M.2 SSD: B-key vs. M-key (SATA M.2 vs. NVMe M.2)
The M.2 keys are where it gets confusing. M.2 SSDs are always super compact but not always necessarily fast.
In the early days of M.2 SSDs, most drives use the SATA speed standard for compatibility reasons. This type is also transitional — it’s not faster than standard SATA SSD — and has slowly faded into obscurity in the past couple of years.
In any case, here’s the break-down of M.2 SSDs:
- Original versions of M.2 SSDs use the B-key. They tend to use SATA speed standards and are not faster than regular SATA SSDs. Some B-key M.2 drives use the 2x PCIe speed standard and are slightly faster. These drives only fit in the B-key socket on a host (like a motherboard).
- Some M.2 SSDs use B & M keys to fit in both B-key and M-key sockets. These drives also tend to use SATA or x2 PCIe speed standards.
- The latest M.2 SSDs always use the M-key. These drives use x4 PCIe (or faster) speed standards and are the fastest SSDs on the market. They are known as NVMe drives — more below.
If you get an M.2 SSD today, it’s likely an NVMe drive with an M-key.
M.2 SSD speed
M.2 SSDs have a few (and growing) speed variations.
Speeds of SATA-based M.2
SATA M.2 drives have the same speed as regular SATA SSDs (6Gbps). Sometimes, you can find the same drive in two different form factors, like the WD Red SA500 SSD case.
Speeds of PCIe-based M.2
M.2 drives that use PCI Express (PCIe) lanes to communicate with a host computer are labeled as NVMe (Non-Volatile Memory Express) M.2 drives.
NVMe comes in several flavors, including NVMe 1.3, 1.3c, 1.4, and even more in the future. These revisions decide how the information is organized on the drive via the blocks and pages mentioned above. They have little to do with speeds and nothing to do with physical design.
As a result, they have much faster speed than SATA SSDs and will get even faster in the future with new generations of PCIe, as shown in the table below.
Specifically, currently, PCIe Gen 3 has a top speed of 8Gbps (985MB/s) per lane. That said, a PCIe Gen 3 x2 (two lanes) M.2 drive has a cap speed of up to 16Gbps. Similarly, a PCIe Gen 3 x4 SSD has a ceiling speed of up to 32Gbps.
|Commercially Available||Rate per lane|
|1||2003||2 Gbps||250 MB/s||0.5 GB/s||1.0 GB/s||2 GB/s||4.0 GB/s|
|2||2007||4 Gbps||500 MB/s||1 GB/s||2.0 GB/s||4 GB/s||8.0 GB/s|
|3||2010||8 Gbps||984.6 MB/s||1.97 GB/s||3.94 GB/s||7.88 GB/s||15.8 GB/s|
|4||2020||16 Gbps||1969 MB/s||3.94 GB/s||7.88 GB/s||15.75 GB/s||31.5 GB/s|
|5||2022||32 Gbps||3938 MB/s||7.88 GB/s||15.75 GB/s||31.51 GB/s||63 GB/s|
Note: 1 Gigabyte per second (GB/s) = 1000 Megabyte per second (MB/s) | 1 Gigabit per second (Gbps) = 125 MB/s
In my experience, a good PCIe Gen 3 NVMe SSD (like the Samsung 970 EVO) can deliver sustained copy speeds of over 2,000MB/s.
Future NVMe SSDs will be even faster, thanks to the faster lane speed of next-generation PCIe. For now, PCIe Gen 4 SSD, such as the Samsung 980 PRO, can deliver a rate some 50% faster than a PCIe Gen 3 drive.
PCIe NVMe SSDs are generally backward compatible. A PCIe Gen 4 drive will work with a PCIe Gen 3 M.2 slot. And vice versa, Gen 3 SSDs will work with a Gen 4 slot.
When it comes to consumer storage, NVMe SSDs are the way of the future.
All new motherboards and laptops now use NVMe as the standard internal storage choice. Some even come with multiple M.2 sockets.
But the hard drive won’t go away, either. It still provides low-cost storage solutions, such as NAS servers, where speed is not a big issue.
No matter what type of storage you use, in the end, it’s always what you want to convert into those zeros and ones — the information — that matter.