Remaining SSD life

[Wolf9999]

I'm the happy owner of a Super Talent UltraDrive 64 GB FTM64GX25H (Indilinx controller) and I've never had a issue so far and I've upgraded the firmware each time a new one was released. Currently it's version 2030.

I've been using the drive as my boot and program drive for my Win 7 64 bit for the past 14 months and accordingly to both CrystalDisk Info and Indilinx SSD Status the remaing drive life is now 7%.

The disk has 38 GB free (of 64), the cell wearing is 4669, GB read are 7995, GB written are 3020.

Because the drive life is reported as 7% and going down a 1% or so every week, should I worry?
What happens when the reported life is 1% or 0%?

Should I buy a new SSD now and clone the old one to the new one?

Thank you!

 

[Mark R]

When it reaches 0, your BIOS may give you a SMART warning on boot.

It's unlikely that the drive would shutdown, or go read-only. The life counter is often very conservative, and I'd expect the drive to be able to do 3-5x as much as the counter says.

I find it strange that the counter dropped so fast though. I have an Intel G2, and have put 4200 GB writes on it in 10 months - the drive still reports 98% life span available.

 

[PlasmaBomb]

Back up your data... just in case.

Whats the warranty on the drive?

 

[taltamir]

interesting, a little bit of math shows that 3020GB written / 64GB = 47... aka, 47 writes per cell. Yet its showing 4669 writes total... which is a write amplification of ~100x? (there is much more complicated math if you want to be perfectly accurate, but this above calculation is a quick and accurate enough rough estimation)

also, it seems to think that the cells can only be written 5000 times to give the 7% figure (typical MLC is 10,000).
This looks pretty bad. I am very curious to see how it will fail, in theory you should not lose data. Keep us updated please.

As far as what you should do, back it up, and check on the warranty. No point in buying a new one until it actually fails though. Prices constantly go down, new tech constantly comes out.

I am curious, do you have TRIM? NCQ (via AHCI)? are your partitions properly aligned?

I just checked my 80GB intel G2 and it shows 7160GB written and 97% life left. (it will not tell me writes per cell), which gives me an expected life of 7.16TB*100/3 = 238.7TB
if the MLC is rated for 10,000k writes then 80GB*10,000 = 800TB expected. So a write amplification of 3.35x. If, like yours, it considers the MLC only good for 5000 wites, then it means half that write amplification (1.68x write amplification)

anyone else has an indilinx drive to compare?

 

[flamenko]

Yup....watching this thread as well with same questions...

 

[taltamir]

http://www.anandtech.com/show/3965/intels-3rd-generation-x25m-ssd-specs-revealed

notice anand specifies "total 4k random writes lifespan" at 7.5TB for my drive. I did some math, 4k random writes in drive without TRIM have a write amplification of at least 128x, ceause you have to delete a 512kb block and then rewrite all of it to write a mere 4k (anything less than 4k still takes 4k on the drive itself)...
So I suspect you have something generating tons of 4k random writes and no trim enabled. What is your OS? do you defragment your SSD? (well, if you aren't using win7 it will be automatically defragmented... very bad!)

 

[Old Hippie]

So I suspect you have something generating tons of 4k random writes and no trim enabled.

If your figures are correct something is prematurely wearing the drive?

Looks like this drive is basically an OCZ Vertex 60GB and didn't have TRIM when it was released but did have some kind GC and wiper utility.

W7 set my Defrag to manual when I installed an SSD (like it should) but I went in and disabled the service.

AAR, Wolf9999 you are in a position that AFAIK none here have experienced....yet.

If CrystalDisk Info and Indilinx SSD Status are correct (they probably are) it will be very interesting to how your drive/data reacts as it gets closer to the end.

I guess we're just a morbid bunch!

If I were in your shoes there's no doubt I'd be buying another drive to clone the origional but I'd have to use the origional to the bitter end just to see how it reacts.

If possible please keep us informed on the details.

I know I'd appreciate it as would many other inquisitive souls who prowl this board.

 

[FishAk]

Very interesting...
Make sure you have a good image file, so it's very easy to recover- just in case.

 

[BTRY B 529th FA BN]

as with all Sandforce SSD drives, make sure you let the machine sit in idle state with nothing running for an hour so it runs the garbage collection

 

[FishAk]

The OP's drive is Indilinx. Does GC work the same as SF?

 

[taltamir]

If CrystalDisk Info and Indilinx SSD Status are correct (they probably are) it will be very interesting to how your drive/data reacts as it gets closer to the end.

I guess we're just a morbid bunch!

I would say its less being morbid and more wanting hard proof and factual evidence supporting our hitherto theoretical claims of "what happens when your SSD runs out of writes".
Would it really just lose the ability to write without losing any data as we were promised? will it catastrophically fail causing your data to go poof? what will happen! stay tuned to find out.

if you do replace it, please don't throw out the old one though, let one of us run tests on it.
if you can get it replaced under warranty, please wait for it to reach 0 before sending it in.
I have never seen someone in your position, and as drives improve it will be less likely we see one again.

Also, if lack of trim helped cause it... well, trim might be a lot more important than we have believed thus far (we just saw it as a way to get a little more speed).

 

[Old Hippie]

I would say its less being morbid and more wanting hard proof and factual evidence supporting our hitherto theoretical claims of "what happens when your SSD runs out of writes".

Unbelievable.

 

[=Wendy=]

interesting, a little bit of math shows that 3020GB written / 64GB = 47... aka, 47 writes per cell. Yet its showing 4669 writes total... which is a write amplification of ~100x? (there is much more complicated math if you want to be perfectly accurate, but this above calculation is a quick and accurate enough rough estimation)

also, it seems to think that the cells can only be written 5000 times to give the 7% figure (typical MLC is 10,000).
This looks pretty bad. I am very curious to see how it will fail, in theory you should not lose data. Keep us updated please.

As far as what you should do, back it up, and check on the warranty. No point in buying a new one until it actually fails though. Prices constantly go down, new tech constantly comes out.

I am curious, do you have TRIM? NCQ (via AHCI)? are your partitions properly aligned?

I just checked my 80GB intel G2 and it shows 7160GB written and 97% life left. (it will not tell me writes per cell), which gives me an expected life of 7.16TB*100/3 = 238.7TB
if the MLC is rated for 10,000k writes then 80GB*10,000 = 800TB expected. So a write amplification of 3.35x. If, like yours, it considers the MLC only good for 5000 wites, then it means half that write amplification (1.68x write amplification)

anyone else has an indilinx drive to compare?

The first Indilinx based drives had 50nm NAND (10,000 PE cycles), but a few months into production, and as 34nm NAND became widely available, they started using 34nm NAND (5,000 PE cycles, except for some special binned stuff).

I've been saying for months that 34nm NAND (in most parts) is 5,000 cycles, but it is mostly ignored or contradicted by members here.
Having said that, the NAND on the OP's drive seems to be burning faster than normal.

I have a couple of Indilinx based drives, but they are review samples and use the original 50nm NAND.