The Corsair User Forums  

Go Back   The Corsair User Forums > Corsair Product Discussion > Solid State Drives (SSD)

Reply
 
Thread Tools Search this Thread Rating: Thread Rating: 8 votes, 5.00 average. Display Modes
  #1  
Old 10-28-2011, 12:16 AM
Synbios's Avatar
Synbios Synbios is offline
Registered User
Synbios's PC Specs
 
Join Date: Jul 2003
Location: Washington, D.C.
Posts: 1,157
POST ID # = 537814
Synbios Reputation: 10
Send a message via AIM to Synbios
Default Have questions about SSD Life?

I finished some tests on an original Force drive:

http://www.corsair.com/blog/cat/tech...-life-testing/

Scroll to the bottom if you're interested in the final results. I hope someone finds it useful.
__________________
Force drive not recognized in firmware update utility? Click here
Reply With Quote


  #2  
Old 10-28-2011, 01:31 AM
Mummy Mummy is offline
Registered User
Mummy's PC Specs
 
Join Date: Jul 2010
Posts: 220
POST ID # = 537825
Mummy Reputation: 10
Default

Damn im getting old.
Wasnt that "Life Time" info wrong on SSD's as it may show much under 100% healt to a new driver. ...must member wrong then.

Thx for the test :)

Edit: ok, so i have magic pc that runs with water
Attached Images
File Type: jpg Disk Not Found.jpg (848.8 KB, 311 views)

Last edited by Mummy; 10-28-2011 at 01:42 AM.
Reply With Quote


  #3  
Old 10-28-2011, 02:43 AM
parsec's Avatar
parsec parsec is offline
Registered User
parsec's PC Specs
 
Join Date: Aug 2011
Location: USA
Posts: 1,456
POST ID # = 537832
parsec Reputation: 10
Default

Very interesting, thanks for posting this. It sure seems as if all the hand wringing about wearing out a SSD from writing to it is much ado about nothing. I'd like to see if a standard HDD would survive after having 240 TiB written to it.

Yes, impressive endurance, and on a mere 40GB SSD. This on 25nm NAND as well, which is supposed to have fewer write operations than 32nm NAND.

A comment about the write and delete process, and please comment on the accuracy of my statements. In reality, a "delete" operation on a NAND cell (a bit), is really just a reset of the cell's state to it's ready-to-be-written-to state. The reset operation is operationally no different than a data write operation to the cell. So when N bytes are written and then deleted from a drive, the NAND has actually sustained 2N "writes" (P/E cycles), due to the delete/reset process. Thus writing and deleting 240 TiB (N) of data to the SSD is equivalent to 480 Tib (2N) of NAND state changes.

Do you think that every doubling in capacity of the SSD, such as 40GB to 80GB, will then double the amount of data that can be written to the SDD in its lifetime? (ignoring over-provisioning and real vs actual capacity,etc)

AFAIK, the standard manufactures policy regarding a SSDs operation when it can no longer be written to, is that the data still may be read from the drive for a significant amount of time. IMO, that is what I have read about that situation. The reality of that, I really don't know. It is interesting that it seems the controller expired in your testing, a circumstance that is ignored in the scenario I described. Which means the data could not be read by conventional means, and the SSD wear out scenario advertised by manufactures is not a given. Your thoughts on this?

Assuming that your testing process was executed more or less at a constant rate, why SSDlife, at the 57% life remaining point, estimated the remaining life of the drive to be eight and a half years, is strange. That does not make sense if the program is using the usage history of the SSD to determine it's remaining life.

IMO, the "health" status reported by SSDlife is simply based on an arbitrary threshold that is coded into the program. I'm not saying it is right or wrong, but that it changes from good to not good at a point chosen by the programmers, and using it as a true indication of a SSDs health is questionable. The actual percentage of remaining life is a very useful spec, assuming it is accurate.

I hope you don't feel my comments and questions are a criticism of your test and blog, they are not at all, and intended simply for discussion.
Reply With Quote


  #4  
Old 10-28-2011, 01:58 PM
Synbios's Avatar
Synbios Synbios is offline
Registered User
Synbios's PC Specs
 
Join Date: Jul 2003
Location: Washington, D.C.
Posts: 1,157
POST ID # = 537938
Synbios Reputation: 10
Send a message via AIM to Synbios
Default

Quote:
Originally Posted by parsec View Post
Very interesting, thanks for posting this. It sure seems as if all the hand wringing about wearing out a SSD from writing to it is much ado about nothing. I'd like to see if a standard HDD would survive after having 240 TiB written to it.
I definitely agree, it appears that you don't even need a secure erase to restore performance on these drives. I think that is mostly due to the SandForce controller and TRIM, in my opinion the best in the market right now. It would be interesting to see how other drives respond to similar treatment, but I am assuming older generation and maybe even just non-SandForce drives will not be able to withstand abuse like this and maintain performance. There was of course a slight decrease in performance, but it is VERY minimal and certainly not something to write home about. Unfortunately I never got to do an ATTO at the end, but it was because the drive died. I did perform one fairly close to the end and the app measures speed at which it was writing and deleting, which was always consistent.

Quote:
Originally Posted by parsec View Post
A comment about the write and delete process, and please comment on the accuracy of my statements. In reality, a "delete" operation on a NAND cell (a bit), is really just a reset of the cell's state to it's ready-to-be-written-to state. The reset operation is operationally no different than a data write operation to the cell. So when N bytes are written and then deleted from a drive, the NAND has actually sustained 2N "writes" (P/E cycles), due to the delete/reset process. Thus writing and deleting 240 TiB (N) of data to the SSD is equivalent to 480 Tib (2N) of NAND state changes.
I'm looking into this a lot more. Very interesting point.

Quote:
Originally Posted by parsec View Post
Do you think that every doubling in capacity of the SSD, such as 40GB to 80GB, will then double the amount of data that can be written to the SDD in its lifetime? (ignoring over-provisioning and real vs actual capacity,etc)
Yes I definitely do, I have seen similar tests over at Xtremesystems.org and the larger drives last MUCH longer, simply because they have more blocks to write in the first place. Is life 1:1 with data capcity? Yes, I certainly believe so. The actual "life" is at the cell level.

Of course, in my scenario the controller most likely died, so in that case the life has very little to do with the cell life but rather the abuse of the onboard SSD controller.

Quote:
Originally Posted by parsec View Post
AFAIK, the standard manufactures policy regarding a SSDs operation when it can no longer be written to, is that the data still may be read from the drive for a significant amount of time. IMO, that is what I have read about that situation. The reality of that, I really don't know. It is interesting that it seems the controller expired in your testing, a circumstance that is ignored in the scenario I described. Which means the data could not be read by conventional means, and the SSD wear out scenario advertised by manufactures is not a given. Your thoughts on this?
This is certainly a concerning point, and I thought about it as well. I was expecting the drive to become read-only after a certain point. However, it appears that SandForce lets these drives run full speed into the end. It does depend on the manufacturer. Other manufactuers implement what's called LTT (life time throttling). If it detects that your drive is being significantly worn, it actually throttles the write speed! It does this in order to 'attempt' to extend the life of the drive. Personally, I completely agree with SandForce and Corsair's decision to not include this feature. Full speed until the end!

Because of the possibility of the drive disappearing one day, I will always encourage backing up data. Regardless, as long as you monitor the health status, I would not be concerned until it reaches the 10% mark. Even after that, you have a significant buffer of time to get your data backed up and moved to a new drive at that point.

Quote:
Originally Posted by parsec View Post
Assuming that your testing process was executed more or less at a constant rate, why SSDlife, at the 57% life remaining point, estimated the remaining life of the drive to be eight and a half years, is strange. That does not make sense if the program is using the usage history of the SSD to determine it's remaining life.
A huge error on SSDlife's part, IMO.

Quote:
Originally Posted by parsec View Post
IMO, the "health" status reported by SSDlife is simply based on an arbitrary threshold that is coded into the program. I'm not saying it is right or wrong, but that it changes from good to not good at a point chosen by the programmers, and using it as a true indication of a SSDs health is questionable. The actual percentage of remaining life is a very useful spec, assuming it is accurate.
The health status percentage is very ambiguous between different SSD manufacturers. It is a SMART value located at address E7. SSDlife simply reads the E7 value and reports it. The rest of SSD life (time remaining, expected life, etc) is all calculated and is on the SSDlife side of things. I personally would not trust it.

Part of my goal was to give more meaning to the "health status" percentage for Force drives in particular. What i disocvered is that it will stay at 100% for quite some time, then slowly begin to drop. It will taper at 10% for a while, and then die. Needless to say I think the take home message is that for Force drives in particular, replace them at 10%.


Quote:
Originally Posted by parsec View Post
I hope you don't feel my comments and questions are a criticism of your test and blog, they are not at all, and intended simply for discussion.
Not at all. I appreciate the feedback :)
__________________
Force drive not recognized in firmware update utility? Click here
Reply With Quote


  #5  
Old 10-28-2011, 03:14 PM
Mummy Mummy is offline
Registered User
Mummy's PC Specs
 
Join Date: Jul 2010
Posts: 220
POST ID # = 537954
Mummy Reputation: 10
Default

i run in RaiD0 so thats why it dont find any Disk, inside of my system.
So trim isnt working or ever will in RAID but in normal use i dont see secure erasering is any big deal for people that even get RAID0 in his/hers system.

But yeah, i never get post's where some get fraid on ssd's wear of in use and that was for helping side and not the one that was getting the ssd.

*This is my 3 system with only SSD in and im still happy. Thanks for testing :)
Reply With Quote


  #6  
Old 10-29-2011, 11:23 PM
parsec's Avatar
parsec parsec is offline
Registered User
parsec's PC Specs
 
Join Date: Aug 2011
Location: USA
Posts: 1,456
POST ID # = 538155
parsec Reputation: 10
Default

Synbios, it's a shame that the controller on the SSD you tested seemed to expire. Given the posts on Xtremesystems, most if not all of the few SSDs tested did not experience that problem. As you know, many of the SSDs tested have far surpassed the amount of writes you were able to apply to your SSD. They also continued to function beyond the MWI value reaching its minimum. One of the 40GB SSDs was still being written to, and was beyond the 400TiB mark. The participants did not seem concerned with the failure of the controllers.

Given the results seen in the testing on Xtremesystems, with not very many, yet a wide spectrum of SSDs (an "old" Intel G1 to a new M4) all the fuss over NAND wear and failure is a non-issue. The SSDs used in this testing could not be a more random sample, which bodes well for SSD owners. It seems the PC enthusiast journalist community picked up on this, and as they say, became viral.

I have one PC where early on I decided not to baby it's SSD OS drive, except for the truly unnecessary things like defragmentation. It works just fine, never secure erased, one OS installation, and performance is per specs, going on two years of age.
Reply With Quote


Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes Rate This Thread
Rate This Thread:

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -4. The time now is 04:23 PM.


Powered by vBulletin® Version 3.8.7
Copyright ©2000 - 2020, vBulletin Solutions, Inc.