Lifespan question on SSD

[yacoub]

I've had an Intel X25-M 120GB SSD in operation as a boot drive for just shy of one year. It's been flawless.

In that time, SSD Toolbox says I've written 1.55TB of data on it.

At this rate, what is the estimated/average lifespan?

How many TB of writes is the norm for the lifespan for this series/model of SSD?

And is there somewhere I can view that information for various SSD drives?

I'm currently reading this article about enterprise use of SSDs and see references to how many writes per NAND cell but not sure how that equates to how much I've potentially written to the cells on mine: http://www.anandtech.com/show/5518/a-look-at-enterprise-performance-of-intel-ssds
For a different series of Intel SSD (320 series, which is newer and different die size), this info is stated:

If you divide the column on the right by the column on the left you'll come up with 125 program/erase cycles per cell (if you define 1TB as one trillion bytes). If we assume that each cell is good for 5000 p/e cycles (Intel's 25nm MLC NAND spec) then it means that we're actually writing 40x what we think we're writing. This 40x value gives us an upper bound for write amplification on Intel's SSD 320. That's far lower than the peak theoretical max write amplification of 256 (writing 2048KB for every 8KB write sent to the host), but it's safe to say that Intel's firmware won't let things get that bad.

Write amplification of 40x isn't very good but it's also not very realistic for the majority of workloads. Our database workloads are heavy but they are not perfectly random writes over all LBAs for the life of the drive. Those workloads do exist, but we're simply not an example of one. A more realistic, but still conservative estimate for write amplification in our case would be 10x (just based on some internal estimates for write amplification). The actual write amp is likely less than half that but again, I wanted to be conservative.

But I don't know the write amplification for my X25-M, nor the average p/e cycle count, so I'm not sure what formula to use to calculate that 1.55TB of writes against whatever the estimated p/e lifespan of the NAND cells on the X25-M is.

http://www.anandtech.com/show/5518/a-look-at-enterprise-performance-of-intel-ssds/6 doesn't seem to help since my X25-M's E2 and E4 values are crazy (i.e., probably not supported). E2 shows 20477320 and E4 shows 943001248.

 

[Seven]

I have almost 5TB written and still works as new.

 

[Puppies04]

Don't worry...

If you don't want to crunch through the math, Intel estimates that the 80GB X25-M will last for five years with "much greater than" 100GB of write-erase per day. That's a relatively long time for much more data than most folks are likely to write or erase on a daily basis.

Full article.

http://techreport.com/articles.x/15433

 

[exdeath]

Assuming decent write amplificaition and wear leveling firmware, total writes until flash failure = max capacity of drive (including hidden slack space) x max program erase cycles

120GB = 128GB raw NAND, so 128 x 5,000 = 640 TB.

Lets say the 40x write amplification is true (which is absolutely crazy, should be around 1.5x to 2x) then 640GB / 40 = 16 TB before failure.

You have written 1.5 TB in 1 year, so that gives you a total lifespan of 16 TB / 1.5 TB per year = 10.67 years at WORST case scenario of 40x write amplification (which is absurd), and going by the minimum specified program / erase count (which is a low side estimate to CYA, in practice the cells don't magically die at exactly 5,000 erases).

Even with all this scary terrible worse case info with low program erase counts and obscene write amplification, you still have around 10 years left. I never had HDDs that I've kept for more than 7 years before upgrading to something bigger/faster, and as a machine ages past the first 3-4 years you use it less and less and aren't benchmarking and installing stuff like mad, it just kind of sits in the corner as a backup system.

Basically NAND exhaustion is out of the question at the consumer desktop level. If the drive fails on you in it's lifetime, it will be a firmware glitch or controller failure.

Keep your data backed up, and in 10+ years when it does die, buy a new SATA 48g 6 GB/second 1 million IOPS 2 TB drive for $200 to replace it. If you even keep it that long. Most people decide it's a great time for a new generation storage device at the time they do a clean install of the newest Windows release.

 

[yacoub]

Thanks guys, very helpful!

I do back up important files every couple months just in case.

The biggest concern I'd have is that when an SSD does fail, there's generally no way to recover data from it the way there is with an HDD. Or have I been misinformed?

 

[Coup27]

When NAND has been written to the maximum number of times it will it will enter a read only mode. Your data will be safe on the NAND for 12 months before it will fail entirely.

 

[Mark R]

I have almost 5TB written and still works as new.

Several people have tried "testing to the death".

A 40 GB Intel X2 got to about 350 TB before dying. (This took about 3 months of 24/7 heavy stress testing).

An 80 GB is currently up to about 385 TB and still working fine.

 

[bryanW1995]

Assuming decent write amplificaition and wear leveling firmware, total writes until flash failure = max capacity of drive (including hidden slack space) x max program erase cycles

120GB = 128GB raw NAND, so 128 x 5,000 = 640 TB.

Lets say the 40x write amplification is true (which is absolutely crazy, should be around 1.5x to 2x) then 640GB / 40 = 16 TB before failure.

You have written 1.5 TB in 1 year, so that gives you a total lifespan of 16 TB / 1.5 TB per year = 10.67 years at WORST case scenario of 40x write amplification (which is absurd), and going by the minimum specified program / erase count (which is a low side estimate to CYA, in practice the cells don't magically die at exactly 5,000 erases).

Even with all this scary terrible worse case info with low program erase counts and obscene write amplification, you still have around 10 years left. I never had HDDs that I've kept for more than 7 years before upgrading to something bigger/faster, and as a machine ages past the first 3-4 years you use it less and less and aren't benchmarking and installing stuff like mad, it just kind of sits in the corner as a backup system.

Basically NAND exhaustion is out of the question at the consumer desktop level. If the drive fails on you in it's lifetime, it will be a firmware glitch or controller failure.

Keep your data backed up, and in 10+ years when it does die, buy a new SATA 48g 6 GB/second 1 million IOPS 2 TB drive for $200 to replace it. If you even keep it that long. Most people decide it's a great time for a new generation storage device at the time they do a clean install of the newest Windows release.

Intel claimed a write amplification of only 1.1 on the x25m, so that would give you close to 400 years total lifespan at your current rate. From what I can recall, it's safe to use a number like 5 or so on even the worst/most aggressive usage pattern, so figure 90 years or so before the nand wears out.

edit: and the crazy thing is that there's no guarantee that the drive won't actually last those 90 years +, either. with an estimated failure rate of only about .4%/year, it's a very real possibility that it will still be working in your great-grandkids computer (though exdeath will probably start burning down foundrys well before then if we don't have some sort of lightspeed cloud storage crap by then )

 

[nismotigerwvu]

When NAND has been written to the maximum number of times it will it will enter a read only mode. Your data will be safe on the NAND for 12 months before it will fail entirely.

This needs to be stressed, however controller/firmware failures might not be so rosy.

 

[BFG10K]

Intel expects the 320 series to last 5 years if up to 20GB is written per day.

 

[Voo]

Intel claimed a write amplification of only 1.1 on the x25m, so that would give you close to 400 years total lifespan at your current rate. From what I can recall, it's safe to use a number like 5 or so on even the worst/most aggressive usage pattern, so figure 90 years or so before the nand wears out.

Several people already estimated the WA for x25 g2 drives on the forum. Personally my 160gb desktop drive got a WA of somewhere around 3 and I was at the upper end of the range there. So yep a WA of 5 seems a reasonable, conservative assumption for any normal client load (and I do run some development servers from the SSD which isn't that normal anymore)

Anyhow the math works out to be much, much more data than anyone is ever likely to write to a SSD for a client desktop. Even Intel's extremely conservative guarantees are more than most people would ever need (and they're off by an order of magnitude).


Any faulty SSD I've ever seen/heard of died because of a firmware/controller problem - Intel's really good in that department but still possible. But then I hope everybody agrees that only an idiot would store important data on any kind of storage facility without at least one backup.. even if the drive itself was indestructible I can think of lots of other things that can destroy data.

 

[razel]

There an SSD torture write test thread off another forum referenced on this forum and 1.5TB is fine. If you want real-life average lifespan, then these drives haven't been out enough. Even if you do get drives returned for over-wear, only Intel would know and be able to validate it. Add those two factors together and you end up with DON'T WORRY ABOUT IT. Just back up your stuff which is CRAZY easy and fast with Win7 these days and make sure you bought it from a place reputable so that Intel will have no excuse to reject your warranty.

 

[exar333]

Keep your data backed-up and buy a new SSD in 3-4 years that will be 2-3x faster and will have 5-6x the capacity for the same $$$.

 

[bryanW1995]

Funny you should say that, I spent ~ $220 for an 80gb x25m g2 on BF 2009, and my goal this year is to spend ~ $220 for 2x256 gb ssd's (each). Or, even better, $500 for a 500gb ssd.

 

[Puffnstuff]

If you're running an sf1222 based drive you better have a back up handy all the time. Ask me how I know that to be true. My second failed drive, ocz agility 2, made it the longest of the two. The adata s599 failed after two months and it took them one month to complete the rma. When that drive came back in I cloned the then new ocz to it and put it away. Lo and behold the ocz failed in the same way and out came the adata and I was right back up and running. When the ocz returned from rma I cloned the adata to it so when it fails again I'll pop the ocz in and keep on keeping on. This is the ssd circus I run and it's also the chance I take for speed. I'll never go back to a mechanical boot drive, they're just too slow for me.

 

[bryanW1995]

Would have been cheaper for you to buy one quality drive.