Do Grouped Disks Actually Go Faster? RAID, and What Is Inside an SSD
19 August 2026
Last time I measured a NAS and concluded that storage does not need to be fast. This article is about what goes inside it.
https://kimono-oyaji.com/storage-hdd-nas-options-en/If you group several disks together, does it get faster in proportion? That is the grouping scheme (RAID), and alongside it, how an SSD’s internals make it a different product entirely. Buy without knowing those two and the cheap purchase turns out to be the one that dies early.
This article is compiled from published information. It is not the result of building and measuring an array here (checked 19 August 2026).
- 1. What I looked at, and where from
- 2. RAID is not only a scheme for going faster
- 3. SSDs are different products depending on what is inside
- 4. So does grouping SSDs in RAID solve it?
- 5. What happens when you group HDDs instead
- 6. I did not build and measure one: what this article cannot tell you
- 7. In summary: striping several cheap SSDs is the most dangerous thing you can do
- 8. Sources (RAID and SSD endurance)
What I looked at, and where from
Nothing here was built and measured. What follows is put together from published sources, and the range is worth stating up front.
- What: the RAID arrangements and what each trades away; how the type of NAND inside an SSD changes its endurance and its sustained write speed
- Sources: a drive manufacturer’s endurance documentation, a NAND technology explainer, and a NAS vendor’s own warning about RAID (all linked at the end)
- Checked: 19 August 2026
- What is from here: only the measured speeds quoted from the previous article (NAS 117MB/s, external SSD 1,000MB/s, internal M.2 3,500MB/s)
RAID is not only a scheme for going faster
It makes several disks look like one. Which arrangement you pick changes speed, capacity and resilience.
| Arrangement | Speed | Usable capacity | If one disk fails |
|---|---|---|---|
| RAID 0 | Fast | All of it | Everything is lost |
| RAID 1 | Reads are fast | Half | Carries on if the other survives |
| RAID 5 | Writes are slower | One disk less | Survives one |
| RAID 6 | Writes slower still | Two disks less | Survives two |
| RAID 10 | Fast | Half | Depends which |
Two things came up while reading that belong before anything else.
RAID is not a backup
A file you delete by mistake disappears from every disk at once. RAID protects against a disk failing, not against you.
Mistakes, malware, the whole machine being destroyed. RAID covers none of them. You need a copy somewhere else.
More disks means a higher chance that one of them fails
Obvious once said, easy to miss.
You get the speed and the capacity, but the protection is weaker than a single disk. Used knowingly that can be a fine choice; built in ignorance it hurts.
The repair window is the most dangerous moment
When a disk fails in RAID 5 or RAID 6 and you fit a replacement, the contents are recalculated from the surviving disks. That process is called a rebuild.
During it, every remaining disk is read end to end. With large HDDs that takes hours or days.
Which means the heaviest load of all is applied to every disk at the array’s weakest moment. If a second drive fails during that window, RAID 5 loses everything. That is the reason people say large drives should be on RAID 6.
SSDs are different products depending on what is inside
Now the SSD side. Two things labelled “SSD" can differ in lifespan by an order of magnitude.
They are separated by how many bits go into one memory cell. The more you pack in, the cheaper and larger it gets — and the shorter it lasts.
| Type | Bits per cell | Write cycles | Where it stands now |
|---|---|---|---|
| SLC | 1 bit | 90,000–100,000 | Barely sold at retail |
| MLC | 2 bits | 5,000–10,000 | Disappearing |
| TLC | 3 bits | Around 1,000 | The current mainstream |
| QLC | 4 bits | Fewer still | Common in cheap, large drives |
SLC is reported to be more than thirty times as durable as TLC. And most of the cheap high-capacity drives on the shelf today are QLC.
That said, normal use does not need to worry about SSD lifespan
Being honest about this.
From what I could find, at something like 10GB a day of ordinary use, running out of write cycles does not realistically happen. Looking at photos, watching video, writing documents — something else breaks first.
Where it does matter is continuous heavy writing: video editing scratch files, database work, and writing out data for generative AI. If that describes your use, it is worth checking which type you are buying.
The advertised speed only holds at the start
Another one that catches people out.
A QLC drive is slow to write as it stands. So manufacturers switch part of the area into one-bit-per-cell mode and use it as a fast landing pad. That fast write region is called the SLC cache.
So does grouping SSDs in RAID solve it?
This is the part I most wanted to write.
It is tempting to think “buy a few cheap SSDs, stripe them in RAID 0, and get speed and capacity". And the raw throughput does go up.
But none of the problems above go away.
| Problem | Striped across several drives |
|---|---|
| Heavy writing shortens life | Shared out, so each drive writes less |
| Speed collapses once the pad is used up | They all run out together. The same total is split evenly, so they stall in unison |
| One drive failing loses everything | More drives, more chances of that |
The second row is the one. Four fast landing pads do not last four times as long. Each takes a quarter of the same total, so they hit the wall at the same moment.
Which makes several cheap QLC drives striped in RAID 0 the most dangerous combination there is. Short life, simultaneous stall, and one failure loses the lot. Three weaknesses stacked.
An example of the side that is harder to get wrong. TLC, with a TBW figure (how much you can write in total) stated as a number. Drives without one are not necessarily bad, but there is nothing to compare.
Another with TBW stated. More expensive, but the figure appears as a warranty condition in the same way.
What happens when you group HDDs instead
Here the situation is different.
A single HDD does something like 150–250MB/s, and grouping them does add up. Unlike an SSD, its life is not consumed by how much you write, so the lifespan question is a separate one.
And some models can pair them with SSDs as a landing pad: keep what you use often on the fast SSD and write through to the HDDs behind it. The NAS here is built that way, and it shows when ingesting.
But the cable you connect it with runs out first
As measured in the previous article.
| Where | Measured |
|---|---|
| NAS (over a 1000Mbps link) | 117 MB/s |
| External SSD (USB, for carrying models) | 1,000 MB/s |
| M.2 inside the PC | 3,500 MB/s |
However many drives are grouped inside the NAS, a 1000Mbps link stops it at 117MB/s. Before adding drives, find out where it is actually blocked.
An example if you are grouping several HDDs. Built for NAS use, designed to spin continuously. Ordinary drives work, but they are not built for running around the clock.
I did not build and measure one: what this article cannot tell you
Everything above is compiled from published information. It is not the result of building a RAID array here and measuring it.
Three figures in particular should not be taken at face value.
- Write-cycle counts vary by product and generation. The table gives ballpark values
- The size of the SLC cache differs per product and often is not published at all
- “10GB a day is nothing to worry about" is also a rule of thumb from what I read. It changes with how you use it
Building one here would let me say something firmer. I have not, so I cannot.
In summary: striping several cheap SSDs is the most dangerous thing you can do
- RAID is not a backup. A deleted file disappears from every disk
- RAID 0 gets more dangerous with every disk added. One failure takes all of it
- With RAID 5 or 6, the rebuild after a failure is reported to be the most dangerous window
- SSDs differ by internals. Write cycles differ by more than thirty times between SLC and TLC
- But normal use is said not to need to worry about lifespan. It matters under continuous heavy writing
- The advertised speed does not hold. It drops once the fast landing pad is used up
- Striping does not stagger the stall — they all hit it together. Several cheap SSDs in RAID 0 is the worst combination
Reading through this, what struck me is that “just group more of them" deserved to be questioned from the start. Some things improve with quantity, some do not change, and some get worse.
Next comes the application of all this: with RAM prices up, can an SSD stand in for memory? To give away the ending — it is possible, and I do not recommend it.