PDA

View Full Version : Trying to wrap my head around the concept of SSD "wear"



Rart
08-05-2011, 06:52 PM
At least from what I have seen thus far with my very limited knowledge of hardware, I have come across two types of SSD "wear".

1. Performance degradation from filling up the drive/hammering it with writes. To a certain extent, TRIM commands from the OS/the native garbage collection on the SSD controller will help alleviate this.

2. The "lifespan" of the SSD. After a certain amount of usage, the flash memory cells will lock, becoming read only.

My main question is, are these types of wear completely independent of eachother, intertwined somehow, or maybe the exact same thing?

And as for the "2nd" type of wear, is it an all or nothing sort of deal, it makes no discernible difference in the performance of the drive until it happens, at which point the entire drive is immediately unusable? Or is it also associated with some sort of gradual performance degradation, whether dependent or independent of the 1st type of degradation?

Finally, I was wondering exactly how exactly Secure Erase command factors in to all this, which is supposed to restore your drive to a factory like state in terms of performance, since it completely wipes everything. I (think) that it completely alleviates the wear due to the "1st" point I mentioned. Does it resolve the wear of the "lifetime" of the SSD in the 2nd point? And if theres any sort of gradual performance degradation associated with the 2nd point, would it help resolve that as well?

clocker
08-05-2011, 08:51 PM
It's magic, just accept it.

tesco
08-05-2011, 11:26 PM
They are two seperate things to worry about (or not).

The first is the degraded speed from filling the drive. This problem doesn't exist anymore since TRIM takes care of it. Just make sure you buy a drive that supports TRIM and use win7 or another modern OS and forget about it.

The second is the degrading life of the SSD. Each block can take something like 10,000 writes before it can't hold a charge anymore. A modern ssd controller keeps track of how many writes it has made to each block, and tries to strategically wear the device. A good SSD has extra space beyond what is advertised (for example a 90gb drive could actually have 130gb). This is so that blocks that cannot hold a charge can be marked as 'bad' by the ssd's controller and still be able to keep the advertised size. Some SSDs (more expensive ones) have more of this buffer space than others.
For average use you don't even need to worry about this, it will be obsolete long before it even gets close to the end of its life.

Oh, and also there's that magic thing.

Rart
08-06-2011, 11:09 PM
It's magic, just accept it.

Are we talking Winter is coming type magic or some Harry Potter shit?


They are two seperate things to worry about (or not).

The first is the degraded speed from filling the drive. This problem doesn't exist anymore since TRIM takes care of it. Just make sure you buy a drive that supports TRIM and use win7 or another modern OS and forget about it.

The second is the degrading life of the SSD. Each block can take something like 10,000 writes before it can't hold a charge anymore. A modern ssd controller keeps track of how many writes it has made to each block, and tries to strategically wear the device. A good SSD has extra space beyond what is advertised (for example a 90gb drive could actually have 130gb). This is so that blocks that cannot hold a charge can be marked as 'bad' by the ssd's controller and still be able to keep the advertised size. Some SSDs (more expensive ones) have more of this buffer space than others.
For average use you don't even need to worry about this, it will be obsolete long before it even gets close to the end of its life.

Oh, and also there's that magic thing.

Thanks for clearing it for it up for me. Just for clarification though: is the SSD degrading life issue an all or nothing type of deal? As in the the cells will act as a factory fresh SSD would, up until the point when it reaches a certain magic number of writes and then dies? Or does this type of wear also slowly degrade performance as the cells begin to lose their ability to hold a charge?

tesco
08-07-2011, 02:26 AM
Are we talking Winter is coming type magic or some Harry Potter shit?


They are two seperate things to worry about (or not).

The first is the degraded speed from filling the drive. This problem doesn't exist anymore since TRIM takes care of it. Just make sure you buy a drive that supports TRIM and use win7 or another modern OS and forget about it.

The second is the degrading life of the SSD. Each block can take something like 10,000 writes before it can't hold a charge anymore. A modern ssd controller keeps track of how many writes it has made to each block, and tries to strategically wear the device. A good SSD has extra space beyond what is advertised (for example a 90gb drive could actually have 130gb). This is so that blocks that cannot hold a charge can be marked as 'bad' by the ssd's controller and still be able to keep the advertised size. Some SSDs (more expensive ones) have more of this buffer space than others.
For average use you don't even need to worry about this, it will be obsolete long before it even gets close to the end of its life.

Oh, and also there's that magic thing.

Thanks for clearing it for it up for me. Just for clarification though: is the SSD degrading life issue an all or nothing type of deal? As in the the cells will act as a factory fresh SSD would, up until the point when it reaches a certain magic number of writes and then dies? Or does this type of wear also slowly degrade performance as the cells begin to lose their ability to hold a charge?I understood it as an all or nothing deal. Something like, the ssd writes to the cell, it either takes the charge or it doesn't, and the controller can instantly detect if it failed and if so writes to a different cell. This should all be unnoticed to you.
Might be wrong about that though.
But I do know for sure that unless you're running a busy database server or something else that's constantly writing to disk you'll never reach these limits.

BTW a S.M.A.R.T reader can tell you how much life you have left in your ssd (in percent), and my drive which I've been using for 10 months now is still claiming 100% life.

megabyteme
08-14-2011, 09:45 PM
I did a bit of reading into SSD's before installing one in my Asus 901 (it takes an unusual, specific SSD), but I seem to remember it being an all-or-nothing catastrophic failure. Which makes sense, or people wouldn't worry until they ran short of space. It was a major problem, but was fixed quite awhile ago.

Prices are still coming down, and I recently purchased a new netbook, but decided to hold off for the next price drop before doing an SSD upgrade to the new computer. I do have to say having a nearly indestructible (drop proof) netbook has been EXTREMELY nice while in school. Right now, the expense vs drive size is not quite worth exceeding our tight budget.

EDIT- I do remember that the issue actually was a result of physical wear and tear on the drive itself and cannot be alleviated by formatting, etc.

For the most part, the technology is quite good. Being a tightwad myself, I am quite a long ways from putting an SSD in any of my home-based machines, but for laptops they are fantastic. If you do want some extra piece of mind (speaking of a laptop/netbook), purchase a 16GB SDHC card rated "class 10", and make quick backups of your important files. If you are holding off as I am, a "class 10" card makes it possible to do "ReadyBoost" which will pick up some of the slack of the typical 5400rpm laptop drives. It is still not quite as fast as an SSD, but is a cheap boost without the full SSD cost. Also, if the SDHC card is pulled out, nothing important is lost since the ReadyBoost files are OS duplicates. It's a smart system.

clocker
08-15-2011, 01:30 PM
I have personal experience with about 20 SSDs, either customer installs or personal use, and the only failures I've seen to date have been my own OCZ Vertex drive (which has been RMA'd twice but seems fine now).
My personal machine ran for over a year on a RAIDed pair of the OCZs and performance didn't degrade one iota...till the one drive just upped and disappeared.
It's replacement did the same thing after two weeks but the third try seems to be working.

So, based on my experience, I'd say fear of performance degradation is unfounded, if the drive is recognized, it's probably working just fine but you'll get no warning when it decides to die.

Artemis
08-16-2011, 08:55 AM
So, based on my experience, I'd say fear of performance degradation is unfounded, if the drive is recognized, it's probably working just fine but you'll get no warning when it decides to die.

If you turn on S.M.A.R.T. reporting in the BIOS it should give you a heads up of imminent failure, if it is a physical degradation, where as if the controller leaves the building it is usually a surprise.

clocker
08-16-2011, 11:42 AM
Yeah, except SMART doesn't work on RAID arrays.

tesco
08-18-2011, 10:30 PM
Yeah, except SMART doesn't work on RAID arrays.I can't believe of all the things they can do, they cannot come up with a way to make that work. :wacko:

clocker
08-18-2011, 10:51 PM
I know.
Damn our technological overlords!