A drive advertised at seven thousand megabytes per second sounds several times faster than one rated at five hundred, yet in daily use the two can feel oddly similar. The headline number measures one specific kind of transfer, and it is usually not the kind that decides how quickly your computer boots or opens programs. Reading a drive’s real speed means knowing which figures describe big continuous transfers, which describe the scattered small ones that fill ordinary use, and how to test your own drive honestly. This guide untangles the numbers so you can judge a drive by the right ones.
Sequential versus random speed
Sequential speed measures reading or writing one large, continuous stream, such as copying a single video file. This is where the biggest headline numbers come from, because the drive can march through address after address without interruption. It matters if you routinely move huge files, but it describes a minority of what most computers actually do. Sustained writing of this kind is also the heaviest sort for endurance, the slow process traced in how sustained writing wears cells.
Random speed measures many small reads and writes scattered across the drive, typically in four-kilobyte pieces. Loading an operating system, opening an application, or searching a folder generates exactly this pattern, thousands of tiny requests to unrelated locations. Random speed is far lower than sequential, often by a wide margin, and it is usually expressed in operations per second rather than megabytes, because the pieces are so small. This is the number that tracks how responsive a machine actually feels.
Queue depth and why it flatters drives
Benchmarks also vary the queue depth, the number of requests in flight at once. A high queue depth lets the drive work on many operations in parallel across its flash chips, which produces impressive peak figures. Marketing tends to quote these deep-queue results because they are the largest numbers a drive can post.
Everyday desktop use rarely reaches those depths. Booting and launching programs mostly issue one or two requests at a time, waiting for each answer before deciding what to read next, so the meaningful measure is random speed at a queue depth of one. A drive that posts a huge deep-queue number can still feel ordinary if its single-request speed is unremarkable, which is why two drives with similar advertising can behave quite differently in your hands.
Why boot and app loads are random
Starting the computer or a program is not one smooth read. The system pulls in many small files, libraries, settings, and assets, and it often follows a dependency chain where one file tells it what to fetch next. That makes the pattern small, scattered, and shallow in queue depth, the opposite of the long continuous stream that sequential figures reward.
This is why an inexpensive SSD and a headline-grabbing one often boot within a second or two of each other. Both handle low-queue random reads well enough that the processor and software become the limit, not the drive. The enormous sequential ratings separate them only during large file transfers, which for many people are occasional rather than constant. Seeing where the difference genuinely appears is the theme of where NVMe pulls ahead of SATA.
Interface limits versus drive limits
Two different ceilings can cap a drive’s speed. The interface it connects through sets a hard upper bound: SATA tops out around five hundred and fifty megabytes per second, an NVMe drive on four lanes of PCIe 3.0 around three thousand five hundred, and newer PCIe generations roughly double at each step. A drive cannot exceed the bus it sits on, so a fast SSD in a SATA slot is held to SATA speed no matter its potential.
The drive’s own flash and controller set the other ceiling. A budget drive on a fast bus may fall well short of the interface limit, because its flash cannot feed data quickly enough, and how long it can sustain a write depends on the kind of memory inside, a distinction drawn out in how NAND flash types differ. External enclosures add a third cap, since a USB bridge and its protocol impose their own maximum, and heat can force an NVMe drive to throttle during long transfers.
Testing your own drive
You can measure any of this yourself in a few minutes with a free tool. The steps below give a fair reading rather than a flattering one.
- Close background applications and confirm the drive has free space, since a nearly full drive tests slower than it truly is.
- Pick a tool that suits your system, such as CrystalDiskMark on Windows, AmorphousDiskMark on macOS, or fio for finer control.
- Set the test file larger than the drive’s fast cache, around four to eight gigabytes, to see sustained speed rather than a brief cache burst.
- Record both sequential and random results, and note the random figure at a queue depth of one separately from the deep-queue peak.
- Compare your sequential result against the interface ceiling to see whether the bus or the drive is the real limit.
- Run the write test twice in a row to reveal any thermal throttling or cache exhaustion on the second pass.
While the tool is open, glance at the drive’s health as well, because a failing drive can report odd numbers, and the sensor readings explained in reading a drive’s SMART health data show whether age or errors are dragging performance down.
Judging a drive by the right number
The speed that sells a drive and the speed that shapes your day are rarely the same figure. Large sequential numbers describe bulk file transfers and make for striking specifications, while low-queue random performance quietly governs how fast the system boots, how quickly programs open, and how snappy the whole machine feels. Knowing which is which stops you paying for headline throughput you will seldom actually use.
Before spending on a drive rated far above your current one, ask what you do that is genuinely limited by storage. If you move enormous files all day, sequential speed and a fast interface earn their keep. If you mostly boot, browse, and open applications, a drive with solid low-queue random speed will feel just as quick as one with a far larger number on the box. Test honestly, read the right row of the results, and let the real workload decide.
Frequently asked questions
Why is my SSD slower than advertised?
Advertised figures are best-case sequential speeds at a high queue depth, measured before the fast cache fills. Everyday tasks are low-queue random reads that run far slower by nature, so the drive feels slower without being faulty. A near-full drive, a slower interface such as SATA or USB, or thermal throttling during long writes can all widen the gap between the box and reality.
What is random versus sequential speed?
Sequential speed is the rate for one large continuous transfer, like copying a single big file, and it produces the headline numbers. Random speed is the rate for many small scattered reads and writes, the pattern of booting and opening programs. Random speed is much lower but matters more for everyday responsiveness, since ordinary computing is mostly small, unpredictable requests rather than long smooth streams.
How do I benchmark a drive?
Use a free tool such as CrystalDiskMark on Windows, AmorphousDiskMark on macOS, or fio for detail. Close other programs, keep some free space, and set a test size larger than the drive’s cache to see sustained speed. Record both sequential and random results, especially random at a queue depth of one, and compare the sequential figure against your interface’s ceiling to find the real limit.
