Most SSD reviews test the wrong thing. They run a benchmark that lasts thirty seconds, capture the peak number, and call it a verdict. That number is real, but it is also the least useful piece of information on the entire spec sheet. The number that decides whether your video project finishes tonight or tomorrow morning lives a step deeper, in something called sustained write speed. No box prints it. No marketing slide mentions it. If you do not know to look for it, you cannot make a smart purchase.
I learned this the slow way, by buying three wrong drives in a row. The lesson stuck. Here is what I tell anyone who asks me which SSD to buy now.
The peak number is not a lie, it is just incomplete
Manufacturers are not making up the headline read and write speeds. Those numbers come from a controlled test: a brand-new drive, freshly formatted, with a small queue of writes hitting a clean cache. The result is the fastest the drive will ever be in normal use.
That sounds fine until you realize what “normal use” actually means. After a week of ownership, your drive is half full, the cache has been exercised a thousand times, and the NAND cells have started showing wear. The peak number is now a memory.
The thing the box measures is essentially a sprint. The thing you care about is a marathon. They use the same hardware but reveal different properties. A good marathon runner is not always the fastest sprinter, and a fast-sprinting SSD is not always the one that survives a long copy job.
If you only do small things, browsing, booting, launching apps, the sprint number is what matters. If you copy hundreds of gigabytes at a time, render video, run backups, or compile large codebases, the marathon number is the one that decides your day.
Inside the drive: the SLC cache and the NAND underneath
Every modern SSD has two layers. The first is a small slice of the drive’s flash set aside to act as a write buffer. This buffer runs in SLC mode, which is the fastest and most durable way to store data on flash because each memory cell holds a single bit. The buffer is called the SLC cache, even though the underlying flash might be a different type.
The second layer is the main storage: the NAND that actually holds your files when the power is off. That NAND comes in three or four flavors, depending on how many bits each cell stores. More bits per cell means more capacity in the same physical space, at the cost of slower writes and shorter cell life.
When you write a file to the drive, the controller sends the data first to the SLC cache. The first few gigabytes of any copy fly at near-peak speed. Once the cache fills, the controller has to drain the buffer into the main NAND, and the speed you actually feel for the rest of the job is the speed of the main NAND.
The ratio between the cache speed and the main NAND speed is where SSDs differ most. A drive with a large cache and fast underlying NAND will look fast on a benchmark and stay fast during your real work. A drive with a small cache and slow NAND will look fast on a benchmark and stall halfway through your job.
What the four NAND types actually mean
You will see four abbreviations on spec sheets. Each one describes how many bits each memory cell stores:
- SLC: one bit per cell. Fastest, longest lasting, and most expensive. Almost never used in consumer drives because the cost per gigabyte is brutal.
- MLC: two bits per cell. Used to be the performance standard a decade ago. Now mostly found in industrial and prosumer gear.
- TLC: three bits per cell. The mainstream choice for good consumer drives. Fast enough for almost everyone, durable enough for years of daily use, affordable enough to compete on price.
- QLC: four bits per cell. Cheapest per gigabyte, which is why it dominates the budget end of the market. The catch is that sustained write speeds fall off a cliff once you exhaust the cache, and the cells wear out faster than TLC under heavy writes.
The industry has moved toward QLC for capacity drives because the price per terabyte keeps dropping. That is good news for anyone who stores a lot and rarely writes. It is bad news for anyone who treats the drive as a working scratch space.
What reviewers test vs what you actually feel
If you want to see the real difference between SSDs, you need a test that runs longer than the cache. Most review sites run benchmarks that write 10 to 20 GB and call it done. That only exercises the cache. The result is a number that tells you how fast the drive is when it is not really doing anything.
The reviews that matter are the ones that write hundreds of gigabytes, plot the speed over the entire run, and report what happened in the second half of the test. Those reviews are rarer than they should be, but they exist. Search for “sustained write test” or “SLC cache exhaustion” along with the drive model.
A useful shortcut: if a review site only shows a single average number, the review did not catch the cache drop. If it shows a graph that flattens out or slopes downward over a long write, the reviewer tested the drive the way you would actually use it.
A practical way to test your own drive
You do not need a lab to see the SLC cache effect. Copy a folder that is larger than your drive’s rated cache. A 100 GB copy on a 1 TB QLC drive is usually enough. Most operating systems can show you the current transfer speed. Watch what happens in the first minute versus the tenth minute. If the speed drops by half or more, the drive is relying heavily on its cache. If the speed holds steady, the underlying NAND is doing the work.
I do this whenever I install a new working drive. It takes twenty minutes and tells me more about the drive than any spec sheet.
Trade-offs
There is no free upgrade here. A TLC drive costs more per gigabyte than a budget QLC drive. If your work fits inside the cache, you will never notice the difference. The honest answer is that most people do not need a TLC drive. A QLC drive from a reputable brand will boot fast, launch apps fast, and feel responsive for years.
The reason I pay the premium is that my jobs do not fit inside the cache. A 4K video project, a database dump, a backup of a developer machine: each one is hundreds of gigabytes. Once you have watched a QLC drive crawl through the back half of a copy, the math changes. The premium for TLC becomes the cost of not waiting.
There are also limits to what you can verify before buying. Manufacturers do not publish sustained write speeds, and most review sites do not measure them. The reliable signal is reviews from sites that do long-form write testing. If you cannot find one for the drive you want, the drive is probably not aimed at buyers who would care.
If you are shopping for an SSD today, decide what kind of work the drive will do. For casual use, a QLC drive is fine and saves real money. For sustained writes, pay for TLC and ideally a drive with a DRAM cache (a dedicated memory chip on the SSD that holds the mapping table, which improves both speed and endurance). The headline speed on the box is a starting point, not a verdict.