My SSD hit a speed cliff after 19 seconds, and the spec sheet never warned me

Gaming has been Samarveer’s greatest passion, and the Literature graduate in him takes immense joy in dissecting games for their themes, messages, and impact. Samarveer holds a deep appreciation of gaming, and considers the platform to be the most immersive and impactful across all media. He can be found engaging with gaming communities online, always ready to debate the finer points of ray tracing or itching to write an 8-page collegiate thesis on any game that impacts him emotionally.

There isn't a single PC user who bought a new SSD and didn't base that decision on the huge read-write-speed numbers in the brochure. We now know that those numbers aren't the most important factor when it comes to purchasing an SSD, but they are still the major draw on the product page.

After all, who wouldn't want a drive that promises sequential write speeds of 7,000MB/s or more? I certainly did, until I watched my own SSD struggle to maintain anything remotely close to its advertised performance. When it comes to writing or moving huge files that are triple-digit gigabytes in size, most SSDs resort to much slower speeds, and that's actually by design.

The write-speed on my SSD ran out after 20 seconds

Large file transfers slow down to a crawl the longer they go on

My SSD, a 2TB NVMe Gen4 drive, boasts a write speed of 6,800 MB/s on the product page. That's a pretty impressive number, no doubt, but only if you don't know what it signifies. On any NVMe SSD, this write-speed number, which, of course, is the biggest (and loudest) number. That's impressive, but also fleeting.

I watched it happen on my own drive — it's a perfectly reasonable unit with TLC NAND and a DRAM cache, not a bargain-bin QLC stick I rigged to fail. I started a long, sustained write, and kept the throughput graph open while it ran. It started pretty strong, transferring a copy of my entire 235GB media server from one partition to another at a whopping 6,000 MB/s speed.

All it took was 19 seconds for that speed to dip hard, and go all the way down to somewhere between 1,000 MB/s and 3,000 MB/s. That was indeed the cliff I knew existed, but had never hit before. If your drive is cheaper or more full than mine, you won’t wait long. If it's a big, empty, premium drive like this one, you'll have to write a lot to get there — which is why most users have no idea this cliff exists.

The fast lane on the SSD is smaller than you'd expect

The SLC cache is doing the heavy lifting

NAND flash on an NVMe drive stores that data in cells, and the more bits you cram into each cell, the more voltage states the controller has to distinguish between. SLC (Single-Level Cell) stores one bit, TLC (Triple-Level Cell) stores three, and QLC (Quad-Level Cell) stores four. Pure SLC and MLC (Multi-Level Cell) SSDs exist, but they're extremely rare in the consumer market, despite being the absolute fastest. It's TLC and QLC SSDs that are the most common for consumers, with QLC drives being the cheapest.

Naturally, packing more bits into the same silicon makes storage cheaper, but writing to it becomes slower. For this, SSD manufacturers use a clever workaround. The drive temporarily writes data to a portion of its TLC NAND operating in SLC mode, effectively using each cell to store just one single bit. This is the SLC write cache — a fast lane that lets your SSD sprint before it settles into its regular pace.

There's a catch, of course, as I found out myself. This fast lane isn't actually a fixed size. NVMe drives employ Dynamic SLC caching, which borrows from the available NAND capacity. This directly means that the emptier a drive is, the larger SLC cache it can offer. My XPG GAMMIX S70 BLADE SSD, with about 24% of its maximum 2TB capacity free, managed roughly 117GB before its performance collapsed.

Once the dynamic SLC cache fills up, an SSD's speeds drop as dramatically as mine did. Consequently, new writes go almost exclusively to TLC or QLC, depending on the drive. The entire while, the controller keeps folding the cache data back into the native TLC storage so it can free up some of the SLC cache while the transfer remains ongoing.

Here's how you can test your own SSD's "cliff"

One free tool and five minutes is all it takes

I used HD Tune Pro, whose trial version is enough for this experiment. Open the File Benchmark tab, select your SSD's partition, and ensure that you set the data pattern to Random or Mixed (not Zero). I made the mistake of setting it to Zero myself, and since zeros can be compressed or handled specially by some drives, it can lead to misleadingly healthy graphs.

Next, set the File length high enough to exhaust your SLC cache. This option is in MiB (mebibyte), so just use Google for converting gigabytes to MiB. My test needed a 235GB file before the slowdown became obvious. What's important to remember here is that partitioning does not magically create separate caches — every partition on the same physical SSD shares its NAND. All you have to do is make sure your chosen partition has enough free space, and then start the test.

Keep the throughput graph visible. You're looking for a stable performance shelf followed by a sudden drop and a jagged, oscillating band. You can watch the same activity under Task Manager's Performance tab, too.

Your SSD isn't slow, but its spec sheet needs an asterisk

Buy for the marathon, not the sprint

If you're primarily gaming, you can stop panicking. Games mostly read data from your SSD, and even installing a 100GB title rarely means writing that entire amount in one uninterrupted burst. Your SLC cache can comfortably absorb smaller transfers, which is why most people never notice this cliff.

There are exceptions, of course. Video editors who dump hours of footage, or anyone migrating massive media libraries or running hefty backups, often run into such cliffs. Plus, they naturally have fuller drives, which makes matters worse because the dynamic SLC cache shrinks, bringing that cliff closer to everyday workloads.

Lastly, when shopping for new SSDs, make sure you prioritize TLC over QLC, consider drives with dedicated DRAM, and leave enough free space for your SSD to breathe. Most importantly, look beyond headline sequential write speeds and investigate sustained performance.

The next time you buy an SSD, look beyond the box

The advertised numbers for your SSD aren't necessarily "lies." Treat them as sprint times for hardware that you would occasionally ask to run a marathon. However, product pages rarely ever make this distinction obvious. After all, they wouldn't want to lead with the number that makes their drive look unimpressive and less exciting.

This number though, is important, and you have to go looking for it yourself by testing your own hardware. You don't need to be a storage engineer, or spend an entire evening (or ten) obsessing over NAND to understand what you're buying. Once you've watched your own SSD fall off that cliff, you'll never read a spec sheet the same way again.