Do SSDs slowly lose storage capacity over time?

[Turbonium]

I'm still not sure how SSDs work exactly, but I've been reading up on degradation and I can't seem to figure out the answer to the following question: does an SSD lose capacity over time?

Like say one day it's at 100GB, then a year later, it's at 100GB-XGB, where X is a function of how degraded it has become.

Also: degradation affects performance as well, right? And eventually, bad performance turns into something that just plain doesn't work, right (i.e. the drive needs to be replaced)?

 

[tweakboy]

No. As for degradation and performance going down that is what TRIM is for.

You can use Crystal disk info and see percentage on drive. It should say 100 percent, if it doesn't and says 97 percent like my old mans pos a-data ssd he used to be 375mbps which is slow for sata 3 ,, now he gets 300mbps and crystadiskinfo says 97 percent and you can see the error is on wear and lifetime and what not.

 

[hofan41]

If your ssd has good wear leveling, then your drive should just go from usable to flat out dead.

If your ssd has bad wear leveling or no wear leveling at all, then what you are talking may occur if the firmware is such that it can adjust its total capacity that it announces to your computer.

 

[imagoon]

The closest to "losing capacity" is the same idea as "Bad sectors" on a spinning disk. Cells do wear out, in most cases they replaced from a space pool of some sort and stop being used.

Performance loss over time is normally related to garbage clean up. The issue is that to even change 1 byte of NAND, you have to erase and entire page which could be say 16kb. So a single 512byte sector on an spinning HDD, would require 16kb of data to be read, the 16kb page erased and the modified 16kb page written back. The original disks did this. Then newer disks used TRIM to proactively erase cells that should be empty via garbage collection. Yet newer disks use copy on write where the 16kb cell is read, the data merged in and then the 16kb page is written someplace else that is already blank and ready to go. Then the SSD would mark the old page as ready for garbage collection. Copy on write drives and "spare area" drives benefit from TRIM far less (if not at all.)

Garbage collection is jut the process where the drive goes through erases empty pages for use again. Of course there are quite a few ways to do GC such as idle time, on demand scheduled etc.

Basically a drive that only does idle GC, needs idle time to clean up. Power save can shut the disk down leaving no idle time etc.

Eventually the SSD will fail due to page failures. Most drive fail in to read only to allow data extraction.

 

[blastingcap]

Drives have overprovisioning to deal with real-life usage. If you strain a SSD enough, it will eventually have so many bad sectors and not enough spare capacity to patch it up with, or go read-only because even with TRIM/wear-leveling it hits its useful read/write cycle life. Some just outright die, but that's usually some other problem like firmware or controller failure.

 

[tweakboy]

If your ssd has good wear leveling, then your drive should just go from usable to flat out dead.

If your ssd has bad wear leveling or no wear leveling at all, then what you are talking may occur if the firmware is such that it can adjust its total capacity that it announces to your computer.

SSDs do not just die according to Crucial tech support.

Theirs at least. that is what crystaldiskinfo is for to see how its health is, as long as you dont buy pos ssd , and stick with sammy or crucial , then your fine..

:\

 

[Turbonium]

Thanks so far. Complicated, but I'll get the hang of it.

It sucks that all consumer-level storage technology has a lifetime; spindle disks last "indefinitely", but usually break eventually (especially newer drives... the longest lasting drives I've ever used are older... still have a 2GB drive that works), while SSDs literally have a very definite and predictable expiry date.

 

[Ketchup]

Thanks so far. Complicated, but I'll get the hang of it.

It sucks that all consumer-level storage technology has a lifetime; spindle disks last "indefinitely", but usually break eventually (especially newer drives... the longest lasting drives I've ever used are older... still have a 2GB drive that works), while SSDs literally have a very definite and predictable expiry date.

I think it's safe to assume that if you buy a recent drive (not something that's been out for 2 years) it will last until you are ready to buy a new one.

These things remind me of when 'traditional' hard drives were getting cheaper and cheaper, bigger and bigger, faster and faster. SSD's are doing that now, and before it gets noticeably bad (in general) you will be wanting a new one anyway.

 

[Turbonium]

I think it's safe to assume that if you buy a recent drive (not something that's been out for 2 years) it will last until you are ready to buy a new one.

These things remind me of when 'traditional' hard drives were getting cheaper and cheaper, bigger and bigger, faster and faster. SSD's are doing that now, and before it gets noticeably bad (in general) you will be wanting a new one anyway.

Yea I'm about to pick up a Samsung 830, based on recommendations and some of my own research. It's for a light to moderate usage machine, so I would agree.

 

[Emulex]

extremesystems has 5PB on their samsung 830 so far. Many people use them server enviornments with battery protection.

Intel 320 is even better at 40-50% OP and capacitors.

LSI megaraid controllers can do ssdguard which can save a failing raid(0,1,5,6/10) by migrating to a hot spare. raid-0 - one drive shows signs of failure or read-only mode and it moves to the hotspare, pretty slick. Not all drives support this. google ibm 5014 on ebay