Why does a portable SSD that copied a 20 GB folder in two minutes crawl to a near-stop halfway through the next job? The drive is not broken, and the cable is probably fine. What most people hit is a well-known design trade-off hiding inside almost every consumer SSD: a fast SLC cache that fills up, then hands the work over to much slower NAND.
This article explains what is happening under the hood, how to tell cache exhaustion apart from heat and interface limits, and what actually helps in practice.
What Is the SLC Cache and Why Does It Exist?
NAND flash stores data in cells. TLC (triple-level cell) NAND packs three bits per cell, and QLC (quad-level cell) packs four. Denser cells mean cheaper storage, but they also mean slower, more delicate writes.
To hide that weakness, drive makers carve out part of the NAND and run it in SLC mode, storing one bit per cell. SLC writes are fast and simple. Incoming data lands there first at full speed, and the drive later “folds” it into the dense TLC or QLC area in the background while the drive is idle or lightly loaded.
This has been standard practice across the industry since roughly the mid-2010s, and it is one of the least advertised specs on any product page. Vendors rarely publish cache size or the folding algorithm, which is exactly why two drives with the same headline speed can behave completely differently on a large copy job.
Why Write Speed Collapses After the Cache Fills
Think of the SLC cache as a loading dock and the TLC area as the warehouse behind it. Trucks unload fast onto the dock. Once the dock is full, every pallet has to be walked into the warehouse one at a time, and the whole operation slows to the warehouse’s pace.
In practice, a drive might write at several hundred megabytes per second or more while the cache holds, then drop to the native program speed of its NAND. On TLC drives that post-cache figure often lands in a much lower band, and QLC drives typically drop further because their native write throughput is lower to begin with.
The exact numbers are model-specific and depend on capacity, firmware, and test method, so treat any single figure you see online with suspicion. What matters is the pattern: a fast burst, then a cliff.
Cache size is not fixed
Many drives use a dynamic SLC cache, which shrinks as the drive fills. A nearly empty 2 TB drive may have a huge cache; the same drive at 90 percent full may have almost none. That is why a drive that felt fast in month one can feel sluggish in month twelve. (Solid-state drive)
Some drives use a fixed cache instead. Without vendor disclosure, the only reliable way to know is to test your own drive with a file larger than the cache.
Sustained versus burst performance
Reviewers separate burst writes (small jobs that fit in cache) from sustained writes (large jobs that do not). Marketing numbers usually describe the burst case. If your workload is video capture, virtual machines, or big backup sets, sustained performance is the number that actually matters to you, and it is the one least likely to appear on the box.
Three Things That Look Like Cache Exhaustion but Aren’t
Before blaming the cache, rule out the impostors. Three common culprits produce a similar slowdown.
Thermal throttling
External SSDs live in small, bus-powered enclosures with no fan. Under a long sustained write, the controller heats up and firmware dials back speed to protect the hardware. On a metal-bodied drive you can often feel the warmth. The symptom looks identical to a cache cliff, and without instrumentation it is genuinely hard to separate the two. If the drive is fast again after a five-minute rest, heat is a strong suspect.
Interface and bridge bottlenecks
A USB 3.2 Gen 2 link tops out around 10 Gbps, and real-world throughput is lower once protocol overhead is counted. USB4 and Thunderbolt raise the ceiling, but the NVMe-to-USB bridge chip inside the enclosure, the host controller, and even the cable can cap throughput independently of the NAND. A drive rated far above what your port can deliver will never show its full speed, and a weak bridge can make a good drive look bad.
Small-file overhead
Copying 40,000 photos is not the same workload as copying one 40 GB video file, even though the byte count matches. Small files carry per-file overhead that can dominate. Test with one large file before concluding anything about the hardware.
How to Tell Whether It Is Really the Cache
A simple test separates the suspects without special software.
- Confirm the drive has plenty of free space, ideally at least 30 percent.
- Let the drive sit idle for ten minutes so it is cool and its cache has had time to fold.
- Copy a single file larger than the drive’s likely cache, such as a 50 GB video file, and watch the transfer graph in your file manager.
- Note where the speed drops and to what level.
- Repeat immediately without resting the drive. If the second run starts slow and stays slow, the cache was already full. If the first run started fast and the second started slow, heat is more likely.
A graph that starts high, holds flat, then falls to a lower plateau and stays there is the classic cache signature. A graph that declines gradually and continuously points more toward thermal throttling.
What Actually Helps in Practice
No setting turns a TLC drive into an SLC drive. But several practical choices meaningfully reduce how often you hit the wall.
Keep free space available
This is the single highest-leverage habit. Dynamic caches need room to work. Leaving a healthy margin of free space keeps the cache large and keeps garbage collection from competing with your writes. If a drive is chronically full, moving data to a larger drive is a better fix than any tweak.
Match the drive to the workload
If you routinely move very large files, prioritize a drive with DRAM cache and TLC NAND over a cheaper QLC model. DRAM helps the drive track where data lives, which keeps sustained performance steadier. For occasional document and photo transfers, a QLC drive is perfectly reasonable and often cheaper. Buying for the worst case you actually face beats buying for a benchmark you will never run.
Break up huge jobs
Writing 200 GB in one continuous stream guarantees a cache cliff. Writing it in chunks with short pauses lets the drive fold data in the background between chunks. It sounds like a hack because it is one, but it works, and backup software with throttling or scheduling options can do it for you.
Improve cooling
A small USB fan pointed at the enclosure, or simply elevating the drive so air can move around it, can delay thermal throttling noticeably during long jobs. It will not fix cache exhaustion, but it removes one compounding factor.
Check the whole chain
Use the shortest, highest-rated cable you own, plug into a port that actually supports the speed you expect, and avoid unpowered hubs for heavy transfers. If you are moving an existing system to faster storage, the same care applies when you clone a hard drive to an SSD in 2026 and want the migration to finish in a sensible time rather than overnight.
Burst Speed Versus Sustained Speed: A Practical Comparison
The table below summarizes how the same drive behaves under different conditions. Figures are illustrative patterns, not promises about any specific product.
| Scenario | What you see | Likely cause | Practical fix |
|---|---|---|---|
| Small copy, drive mostly empty | Full rated speed | Data fits in SLC cache | None needed |
| Large copy, drive mostly empty | Fast start, then a cliff | Cache exhausted | Chunk the job, or accept the plateau |
| Large copy, drive nearly full | Slow from the first second | Tiny or no dynamic cache | Free up space |
| Long copy, warm enclosure | Gradual decline, stays slow | Thermal throttling | Cool the drive, add pauses |
| Copy always capped low | Flat ceiling, never higher | Interface or bridge limit | Better cable, better port |
Common Mistakes People Make
A few habits reliably make the problem worse.
Judging a drive by its peak number is the big one. A drive advertised at a very high speed may hold that speed for only a few seconds. Reading sustained-write reviews, or running your own large-file test, tells you far more.
Filling a drive to the last gigabyte and expecting steady performance is another. Firmware needs working room, and a full drive has nowhere to put a cache.
Blaming the drive for a slow port is equally common. Before returning hardware, verify what your computer’s port actually supports. A quick way to rule out software weirdness on the host side is to clear the browser cache and cookies and retest, though that rarely affects file transfer speed directly; it mainly eliminates background sync activity that can compete for disk time.
Finally, many people buy one drive for everything. Splitting roles, a fast TLC drive for active projects and a cheaper high-capacity drive for archives, usually costs less than one oversized premium drive and performs better in daily use. If you are still deciding which models are worth the money, a rundown of the best external SSD options is a sensible starting point, and pairing it with a guide to the best external hard drives for backup helps you put the right job on the right device.
When the Drop Is a Real Fault
Cache cliffs are normal. Some slowdowns are not.
If the drive drops to single-digit megabytes per second and stays there even when cool and mostly empty, if transfers fail outright, or if the drive disconnects mid-copy, those are fault symptoms rather than design behavior. Check the cable and port first, then the drive’s health with the manufacturer’s utility. Persistent errors after that point usually mean a warranty claim, not a settings problem.
Conclusion
An external SSD that slows down after a fast start is almost always doing exactly what it was designed to do: absorbing a burst of writes in a fast SLC cache, then folding that data into denser NAND at a much lower speed. The cliff is the handoff.
The practical response is not to chase a magic setting. It is to keep meaningful free space on the drive, choose TLC and DRAM if you regularly move huge files, break up very large transfers, keep the enclosure cool, and verify that your cable and port are not the real ceiling. Do those things and the slowdown becomes predictable, which is far easier to live with than a mystery.
FAQ
Why does my external SSD slow down after copying a few gigabytes?
Most consumer SSDs write incoming data to a fast SLC cache first, then fold it into slower TLC or QLC NAND. Once that cache fills, speed drops to the native write speed of the denser NAND. The size of the drop depends on the drive, its capacity, and how full it is.
Is the slowdown a sign that my SSD is damaged?
Usually not. A speed drop after a burst of writes is normal design behavior. Warning signs of a real fault include transfers failing, the drive disconnecting, or speeds stuck at single-digit megabytes per second even when the drive is cool and mostly empty.
Does a fuller drive slow down more?
Often, yes. Many drives use a dynamic SLC cache that shrinks as free space disappears, so a nearly full drive may have little or no cache left. Keeping a healthy margin of free space is one of the most effective things you can do.
Can heat cause the same slowdown?
Yes, and it is easy to confuse with cache exhaustion. External enclosures are small and fanless, so a long sustained write can trigger thermal throttling. If the drive recovers after resting and cooling, heat is likely a major factor.
Would a more expensive SSD avoid this problem?
It can reduce it, but not eliminate it. Drives with DRAM cache and TLC NAND generally hold sustained speeds better than budget QLC models, and larger capacities usually have larger caches. No consumer SSD writes at its peak speed forever.
Does the USB cable really matter for write speed?
It can. The port, the cable, and the bridge chip inside the enclosure all cap throughput independently of the drive’s NAND. If transfers hit a flat ceiling and never go higher, the connection is the likely limit rather than the cache.
