• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Sandforce Overprovisioning Levels and SSD Longevity

B

BBMW

Member
Various Sandforce controllers (and I guess the various firmware releases) support various different levels of overprovisioning. I've heard about 28%, 13%, and 7%.

I will likely use one of these drives as a boot/installed software disk for a new build, and will have have a magnetic HD for transactional/bulk data. There are now a few 90GB Sandforce drives out there. This is right in the wheelhouse of the capacity i'm looking for, but I could go anywhere from 80 to 120GB. The pricing on the 90s is pretty good also, compared to the bigger drives.

BUT...

The 90s seem to be based on the 7% overprovisioning. Is this enough spare area to give the drive sufficient longevity? I tend to keep my machines a while, and I DON'T want the SSD checking out on me a few years down the line. The 100 GB drives seem to be run at 18% overprovisioning, so maybe they're safer, but then again, they're nearly as much as the 120 GBs.

What does everyone think? Am I making too big a deal about this? Should I care about the overprovisioning level?
 
People call the various sandforce capacities, various levels of overprovisioning - but there is no evidence to say that is actually the case.

Sandforce don't publicly release information as to what the various capacities mean or don't. Similarly, nor do the drive manufacturers - probably due to some form of NDA.

I suspect the 28% 'overprovisioning' is not. Instead, my hypothesis is that the 28% is a combination of Sandforce's 'RAISE' data protection system, in conjunction with 13-15% overprovisioning.

The 13% probably is genuinely 13% overprovisioned - giving a reasonable buffer for more efficient wear leveling for systems that don't support TRIM, or for systems where the drives may operated at close to capacity.

7% seems to be the industry minimum. But, given a suitable controller, should be more than adequate. E.g. Intel's controller, with 7% overprovisioning, gives outstanding performance, and very low write amplification. By contrast indilinx's controller, seems to have write amplification of 10-20x as high as Intel's, when run with 7% overprovisioning. Intel's drive life time is sufficiently high, that it would be nearly impossible for a home user to wear one out. (I severely abuse mine, and I estimate that it's got another 35 years to go).

Sandforce actually claims to have even better wear leveling than Intel - so I would have thought that all things being equal, the sandforce should either equal or outlast an Intel - which are already so careful with their flash, that it isn't really an issue.
 
As long as you have trim, and you don't use all the space, the over provisioning won't matter.
 
As long as you're talking about a controller that doesn't have horrible write amplification, then the over-provisioning amount has almost zero to do with drive longevity. (The caveat there is if you fill your drive to the brim and leave it there, then you're exacerbating the situation and it can become a problem for any controller.) But it still almost completely comes down to how much data you write to the drive determines its life. You're worrying about something you don't need to worry about.

Oh, and the reason the 90's are all provisioned at 7% as you mentioned is they were an SKU added later, and they've determined that over-provisioning higher than 7% was pointless (read Anand's article on this if you need more details/confidence). All the new drives should be provisioned at the lower level. Those on the market with higher over-provisioning have likely been in the channel longer (sat in a warehouse somewhere, etc.).
 
Last edited:
well sandforce needs the space to prepare aligned empty blocks (part of their speed) and then there is compression/decompression. 7% is competitive standard.

I believe 13%+ was optimal but you can't have a 50gb drive sell for the same price as a 55GB (usable) - people will @!#$@#$%@# return it.


tis sad that 64gb = 55gb (or 50gb) - pretty soon we'll be down into the forties lol.
 
I don't think anyone has managed to toast even one of the SSDs based on a good controller through media wear-out. I've not seen a single reference to it anywhere. One would think someone somewhere had some SSDs sitting in a test bench writing and reading at the max data rate 24/7 to find out what really happens when the device wears out.

All the SSD failures I read about are usually catastrophic : the SSD completely fails and the data cannot be read. Since worn out media is still readable, these failures must be from faulty hardware or bugs in controller firmware.
 
i've had two crash (intel)
one had a bad sector and would continue to operate
one had a bad sector you hit it and the drive went bye-bye

just like hard drives
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top