Is SSD wear linear?

[Shrek]

I have a solid state drive that has 85% life remaining and has 235 Erases; does this mean it will last 235/0.15 Erases, which is about 1600?

 

[Dr. Dro]

No. Modern SSDs feature TRIM, compressible write algorithms and wear leveling to improve write endurance.

Correct me if I'm wrong but I seem to recall there was another thread about this question?

 

[Shrek]

Don't know about another thread.

I thought 235 Erases was a TRIM thing, namely that each cell has been written to 235 times.

 

[ir_cow]

Each cell has a maximum number of writes. QLC (4bits per cell) is like 300 or so. The SSD controller will evenly spread the data out. TRIM is a background task that clears full cells marked as empty (as in deleted data) so when you write to the drive it doesn't have to clear the cell out first before writing new data on the fly.

 

[Shrek]

Which still leaves my question

I have a solid state drive that has 85% life remaining and has 235 Erases; does this mean it will last 235/0.15 Erases, which is about 1600?

 

[ir_cow]

does this mean it will last 235/0.15 Erases, which is about 1600?

Yes if it works out in your favor. Some SSDs have extra NANDs to help with wear leveling and if a dead cells happens you suddenly don't have less storage space. If you can look up the specs and see what the write limit is too.

 

[nomdeplume]

I have a solid state drive that has 85% life remaining and has 235 Erases; does this mean it will last 235/0.15 Erases, which is about 1600?

Somewhere on the no longer SEO optimized list of sites there was a fairly large test of SSD. Generalizing in the utmost...

- Upper tier SATA and m.2 were capable an order of magnitude or larger (on average) constant writes AND had a nonlinear reporting of drive life remaining. Failure came suddenly after physical polling showed anywhere around 90%, as in ~98% to a sudden drop. In other words they dropped very slowly due to large amount of extra NAND before falling off a cliff.

- Lower and mid tier SSD of both flavor DID on average have what might be termed something approximating linear reporting of drive life. Lots of abstracts and outliers here obviously. In general they fell short of official TBW figures and life dropped point by point before hitting a plateau of indeterminate length(s). At which point errors were diverse (poor reporting subsequent with a failing drive) despite continuing to write and rewrite until failure was accompanied with lots of flashing lights and bells going off. Or they just died absent an EOL statement per above.

 

[Wirko]

compressible write algorithms

To this day? That was a thing long ago, OCZ boasted with that. I haven't seen compression mentioned in years, neither in SSD datasheets nor in reviews.

 

[AusWolf]

I have a solid state drive that has 85% life remaining and has 235 Erases; does this mean it will last 235/0.15 Erases, which is about 1600?

If TRIM is working properly, and your drive has a good wear levelling algorithm, then I suppose so.

 

[GerKNG]

if you are at 85% of the TBW and it shows 85% "life remaining" it should be linear. but even after you've reached the TBW value printed on the box does not mean it's a dead drive.
you can ignore the TBW, and wear leveling. Just look outfor critical errors (some new cheap drives like the P3 can have media and data integrity errors and it seems to be a bug because other manufacturers that most probably rebrand them post on their website to ignore this value.)
Heise online tested the TBW warranty from Samsung back then when the 850 Pro released.
it was a 256GB Drive and has a 150TB TBW warranty. the SSD wrote 9.1 Million GB (9.1 Petabyte) before it died... that's 60 times higher than the warranty.
then they had the cheapest crucial BX200 low end SATA SSD and it was over 100% beyond the TBW.

 

[Shrek]

Good to know they can go way beyond expectation.

 

[Dr. Dro]

To this day? That was a thing long ago, OCZ boasted with that. I haven't seen compression mentioned in years, neither in SSD datasheets nor in reviews.

Yeah, SandForce made big waves about this but I believe it became kind of standard fare for SSD controllers since then. There's no point in writing redundant data to the NAND itself if you can keep tabs on the zeros in files and a map of where the controller should transparently insert them back after all

 

[Wirko]

Yeah, SandForce made big waves about this but I believe it became kind of standard fare for SSD controllers since then. There's no point in writing redundant data to the NAND itself if you can keep tabs on the zeros in files and a map of where the controller should transparently insert them back after all

That's very much improbable. Compression would increase capacity (like it does on archive tapes) and speed (like it did with Sandforce controllers). Manufacturers would use that as a marketing point, and at least some reviewers would do benchmarking with both compressible and uncompressible data. I don't see that anywhere. I see no compression mentioned on SMI product pages. Phison only has this to say, which I agree with:

Compression is easy to accommodate on the SSD and aligns with the streaming model concept, but it provides limited benefit given that most of the bulk data (photos, video or music) is already fully compressed. There are large data sets that can benefit from compression, but the use-case is relatively uncommon, so it tends to be relegated to dedicate server appliances.

What you mentioned is still possible; a sector with only zeros is never written, it does occupy memory cells, but those cells don't have to be erased before being written again.