Is SSDLife software accurate/reliable?

[wpcoe]

I saw some folks joking online about how SSDLife showed their SSD's to be projected to be good ("T.E.C. dat") until 2020, and having a few chuckles. So, I thought I check up on my (1.5 year old, admittedly low-end) Kingston SSD :

Yikes! Can that be right? Here's the report from CrystalDiskInfo:

Should I be planning on a replacement of my SSD, within six months, or so? Or is SSDLife engaging in a bit of "FUD"?

 

[DirkGently1]

That drive will probably outlive you!

 

[groberts101]

yep.. I've seen the author explaining how the algorithm works. These programs are purely estimates of the life used/remaining based on current workflows(over a shorter timespan) and even firmware updates that make minor adjustments to the controllers/smart base code can change the results.

Many have hit 0% and watch the utility return back to 100% once more as it starts another countdown. Destructive flashes have also been known to reset them as well.

Dirk is right on target above. Those tools are not worth much more than the code they contain aside from some ocassional entertainment value. It would be better to use the mfgr designed specific toolbox than a 3rd party app like that. But even then?.. they have been known to go well beyond the expectations of a simple utility's/algorithm's estimate.

 

[Coup27]

How has the MWI been calculated in those graphs? I also take it that when the green starts, thats over and above what they are "officially" rated at?

 

[fffblackmage]

According to SSDLife, my OCZ Agility 60GB is at 13% health remaining. D:

(Not worried)

 

[Binky]

Per crystaldiskinfo, my Vertex 30g went from 100% down to 0%, then jumped to 40%, then slowly increased to ~45%, then I flashed it to a "vertex turbo" and it's back at 34% again.

These estimates are useless. I'm sure that ssdlife is good for making the author some money.

 

[Jeff7]

I'll see if I can remember to run this on my home PC. It's got an Intel SSD, and Intel's utility reports the media wear indicator.

I'd think that it should be a matter of knowing:
- When the drive was installed (no real way for the software to know this)
- How many on-time hours it's logged.
- Current media wearout indicator, or else the total # of bytes written to it.

Use the media wearout or bytes-written vs on-time, and that'll tell you about how many more hours you can reasonably expect to get out of the drive, then figure the hours per day it's powered up, and then project out an end-of-life date.


Update: Intel's utility says:
6888 hours on-time
5.73TB written
Wearout indicator: 97
1 reallocated sector


SSDLife is estimating October 27th, 2020. (Estimated lifetime: 8 years, 8 months, 25 days.

The drive was bought July 18th, 2010.
569 days.
6888 hours on-time.
12.105 hours per day.
5.73TB. (Terabytes or teribytes?)
Let's go with 5730GB
0.8319 GB per hour. (Cripes, what the heck am I doing to this thing? :twisted


Huh, according to the "How do you calculate it?" link in SSDLife:

Note: by the way, some manufacturers give the total amount of data written to the drive as one of the drive lifetime indicators. For example, Intel guarantees that the total of about 37 TB of data will be written to X25-M drives (20 GB per day for 5 years: “The drive will have a minimum of 5 years of useful life under typical client workloads with up to 20 GB host writes per day.”).

I'm at 19.965GB/day average. That's exceptionally average.

So anyway, assume I'm down by 3% in 6888 hours, that should get me to 0 in 222,712 hours, assuming it's a linear relationship, sometime
around Christmas of 2035.

Though according to this PDF, they're only expecting 7.5TB of life, though that's when filling 100% of the drive with 100% random data.

Oh well. I'm not expecting to wear this thing out anytime soon.

 

[frostedflakes]

SSDlife uses the media wearout indicator from SMART to estimate drive health. As others have pointed out MWI isn't necessarily indicative of how many writes your SSD can handle, in practice most SSDs seem to be able to handle many times more writes than MWI suggests, so it should be considered a worst case scenario. You could probably get another couple years of usage out of that drive even after MWI hits 1%.

SSDlife is actually useful for predicting MWI exhaustion with drives that wear out quickly. If you have a drive where the MWI isn't exhausted quickly, though, there is no way of extrapolating MWI exhaustion. On my drive, for example, it's still at 100% after over a year. Until it goes down to 99% or 98%, you can't obtain an accurate wear rate and don't really have enough information to extrapolate how long it would take the MWI to reach 1%. The most you could say is the drive would last for at least X years. For example, with my Vertex 2 that's currently at 10972 hours, you could assume at 10973 it will drop to 99% and that gives you a wear rate of 1% MWI wear per 10972 hours, and extrapolating that you get 1075354 more hours (~123 years) before MWI is exhausted. This would give you a worst case estimate and you could say "we don't know how exactly long it will take to exhaust writes, but based on the info we do have we can estimate at least 123 years." But in this situation, instead of just saying that or "insufficient info" SSDlife just assumes about ~8 years, which is stupid IMO, but I guess I understand why they do it, for SSDs that can handle a lot of writes you can use one for years without MWI even dropping from 100%, and users would just uninstall the program and never bother buying the Pro version if it couldn't give them an estimate. I guess from their point of view it makes more sense to just pull an estimate out of their rear to give to the user instead of telling them the truth.

Also important to point out is that MWI is just the percentage of data the SSD has written compared to the total number of writes the flash can handle. With a good SSD and a desktop workload you're usually looking at decades at least before writes would be exhausted, these things just don't really die of write exhaustion. Firmware bugs and to a lesser extent hardware failure (the controller dieing, for example) seem to be what renders most SSDs inoperable, and these things can happen at any time. So it's important to keep in mind that MWI isn't an indicator of lifetime, only how much your SSD has written to. If you estimate that it would take 40 years to exhaust MWI, for example, that obviously doesn't mean the SSD would last 40 years, chances are something else on it would fail long before then.

 

[Coup27]

Excellent post frostedflakes.

Incidentally SSDlife and another program (forget the name) don't even work with my Samsung. I tried a few pieces of software out of curiosity but all said they were unable to report back any information.

 

[wpcoe]

I'm a bit more comfortable with the SSD, now, but every time I look at that report, I still feel a bit insecure. For my own sanity, I probably should just uninstall SSDLife.

 

[Motorheader]

I was going to post the similar experience. I purchased 2 brand new Samsung drives last year and SSDLife showed 77% health/Drive Health is EXCELLENT right from the start and subsequent updates show the same thing with the last update showing over 15000000000 GB written.

Excellent post frostedflakes.

Incidentally SSDlife and another program (forget the name) don't even work with my Samsung. I tried a few pieces of software out of curiosity but all said they were unable to report back any information.

 

[DirkGently1]

How has the MWI been calculated in those graphs? I also take it that when the green starts, thats over and above what they are "officially" rated at?

The MWI was taken off the SMART data. The graph shows how much data was written when the indicator reached 1. As you can see, different manufacturers set the MWI at different levels but all drives will continue writing way beyond that point.

The question now is how well the drives retain data after that point. Prevailing theory from a warranty stand-point is 1 year guaranteed once MWI reaches 1, and then decreasing rapidly as more data is written. Drives that have reached the 1PB writes mark may not be expected to retain data too well!

The Kinsgston that is doing so well with that endurance test can still retain data although others that reached similar heights either just died without recovery, or started failing retention tests. (MD5 checks are run against the static data on the drives to verify the integrity). The entire drive has been written to ~17,500 times!

Of course, this kind of constant write testing is much harder on the drives than normal use, and it's thought that NAND may have far greater endurance, to orders of magnitude, given more time to recover between writes.

The ongoing testing is here:

http://www.xtremesystems.org/forums/showthread.php?271063-SSD-Write-Endurance-25nm-Vs-34nm

I'm a bit more comfortable with the SSD, now, but every time I look at that report, I still feel a bit insecure. For my own sanity, I probably should just uninstall SSDLife.

I would. There are better tools but none will tell you when a controller will fail, how long the nand can be written to or how long the data will be retained for when you can't write any more. There are simply too may unknowns right now.

 

[Yellowbeard]

I think the single biggest problem in this whole equation is that there is little if any standardization in SMART reporting for SSDs. So, if your data is not reliable, then the utility cannot be.

 

[groberts101]

I think the single biggest problem in this whole equation is that there is little if any standardization in SMART reporting for SSDs. So, if your data is not reliable, then the utility cannot be.

That's exactly right. When you can see a drive reach 0% life left and reset back to 100%?.. not all is as it seems with these types of utilities.

 

[VirtualLarry]

Here's mine:

Haven't had the drive long, funny how the writes are 10x the reads, you would think it would be the other way around.

 

[groberts101]

LOL.. see post #13 as it would surely apply to you Larry.

 

[fffblackmage]

My "dying" SSD :

VirtualLarry's post reminds me that I haven't check for firmware updates for some time...

 

[DirkGently1]

My "dying" SSD :

VirtualLarry's post reminds me that I haven't check for firmware updates for some time...

4TB written. That NAND is just warming up!

EDIT: This new AT article is relevant:

http://www.anandtech.com/show/5518/a-look-at-enterprise-performance-of-intel-ssds

Nice to see some Enterprise testing and to hear what the man himself has to say about endurance!

 

[MadAd]

I had something like 6 or 7 years left but after no changes in firmware, drives or software, it now says 67 years 1 month...yeah right, thats believable - NOT

however the work time and powered on count seems to be about right ... from believing it to be good I now think this software is unreliable

 

[wpcoe]

That's exactly right. When you can see a drive reach 0% life left and reset back to 100%?.. not all is as it seems with these types of utilities.

Shades of ruby slippers and "there's no place like home."

 

[Yellowbeard]

I had something like 6 or 7 years left but after no changes in firmware, drives or software, it now says 67 years 1 month...yeah right, thats believable - NOT

however the work time and powered on count seems to be about right ... from believing it to be good I now think this software is unreliable

ALL software that depends on the SMART data is potentially unreliable due to the unreliability of the SMART data. The problem is two-fold.

1. I have seen quite a few utilities look at the same SSD and report different SMART data/events/errors.

2. No 2 SSD manufacturers report SSD SMART data the same way. Since it is not standardized, the information is for the most part unreliable.

Without some sort of reporting standard, both the data and the utilities are suspect. I would not rely on any of them for making any meaningful decisions.