Hacker News Reader

Comment by tjoff | Hacker News Reader

These tests are quite misleading as they typically (didn't find any details) just go through the entire drive writing to it evenly. Which, to be fair is probably not that bad of a test unless you have intimate knowledge about how the controller operates.

But it means that every single cell gets even use and there are no write amplification and it doesn't expose controller or usage characteristics.

Depending on use it will most likely fail way earlier under more normal conditions. Not saying that filling a drive time and time again is completely abnormal, but it the nicest thing you can do to an ssd in many aspects.

View on HN

This actually isn't even writing data to the disk, it's exploiting a bug in the Windows 11 storage subsystem to run up the drive's firmware's count of bytes written without performing any actual writes in order to disprove the myth that SSD firmware is programmed to self-destruct when the endurance rating is reached.

This is detailed in the thumbnail of the YouTube video embedded about halfway through the article.

Wouldn't the firmware on the drive typically ensure that it's written to evenly regardless of write pattern? If you repeatedly delete/update the 'same' logical block address on the drive, that won't actually correspond to the same physical location.

It will try to do that yes, but there are consequences, drawbacks and imperfections.

You can only cheaply do so with free blocks, and if 70% is occupied you can only spread it out over the remaining 30%. After you've done that for a while you'll have to rearrange existing data which incurs more writes and less performance. And fragmentation is still an issue.

There are tons of tradeoffs - which will be better or worse for different workloads. But to assume that it wear perfectly, which most people seem to do because it is easy, isn't particularly realistic.

That’s why SSDs have overprovisioning, more so for enterprise drives. And why TRIM is important. That being said most workloads are sequential, particularly retail usage. Much easier to deal with for the controller than enterprise workload ratings that assume 4k random writes

Yes? Not sure what your point is.

Sequential workloads and just filling the drive over and over are very different still.

The drive internally usually stores data in blocks larger than the 4k blocks it pretends to use to the OS (say it uses 16k blocks internally). So if your writes are sequential, it will most likely write the same 16k block in one go. If you write random 4k blocks, it will have to rewrite a full 16k block for every 4k block you write, hence write amplification.

So if you fill the drive over and over using sequential writes you can expect way more endurance than if you write random blocks.

Yes, that was one of my points in my first post...

Reply on news.ycombinator.com