Question do frequent partition restores wear down an SSD?

[swapjim]

Every couple of months I restore my Windows installation with ntfsclone. It's a command line Linux utility similar to dd or ddrescue, only that it specializes in NTFS partitions. It's included in Clonezilla.

My question is, will doing this every 2 months wear down the drive?

If you're unfamiliar with ntfsclone, dd, ddrescue, or Clonezilla, but you have knowledge of another partition restore utility, like Acronis, then you can answer about that.

 

[mikeymikec]

SSDs have a finite lifespan measured in writes (and the standard MTBF in hours), so doing lots of unnecessary writes will wear down the drive faster, but it would depend on how much data is being written per occasion: just a modern version of Windows? That plus a load of user data? etc.

If you ignore the standard expected lifespan and start factoring in a weakness in that particular SSD and say a cell fails prematurely, it'll likely fail a lot more prematurely with lots of unnecessary writes.

Also, SSD manufacturers' warranties also mention a certain number of TB written that they'll guarantee the drive for.

Dare I ask: Why are you resetting your Windows installation on a regular basis, let alone that often?

 

[WilliamM2]

As stated, how much wear depends on the amount of data. My Windows install is about 40GB with all programs (data and pics are stored on other drives). So restoring 6 times a year would only be about .25 extra terabytes written a year.

My 980 Pro is rated for only 300TBW. So that extra .25TB plus the 3.5TB to 4 TB I already write a year means it's going to wear out in just 75 years. I think you'll be okay.

 

[Billy Tallis]

A partition restore will by definition do less than one full drive write (though software updates after the restore will cause some additional writes). Consumer SSDs are typically rated for 0.3 drive writes per day for five years, though low-end drives often have only a three-year warranty and QLC drives are more often rated around 0.1 drive write per day. But regardless, one partition restore every few months is insignificant compared to the volume of writes that SSDs are designed to handle. If your day to day use is relatively light on writes, you could do this restore process every week and could still reasonably expect the full 5-year lifespan from your drive.

 

[IntelUser2000]

My question is, will doing this every 2 months wear down the drive?

It's extremely difficult to wear down your SSD as a consumer. Only guaranteed way is Chia mining. Everything else whatever you do is inconsequential.

People are still so worried about TBW lifespans they don't consider the way they really fail. The electronics inside die, just like it has always been. For nearly everyone the electronics that power the setup will die before you get even close to the TBW lifespan.

And actually electronics are not really more reliable despite using "solid state" parts. They fail just like everything else.

 

[bononos]

It's extremely difficult to wear down your SSD as a consumer. Only guaranteed way is Chia mining. Everything else whatever you do is inconsequential.
......

Maybe with QLC drives, consumers could finally see the effects of wear, although the drives aren't going to just fail but get too slow.

 

[mindless1]

No, your every-two-months plan won't wear it out soon enough to matter, but yes, a consumer/home-user on a TLC or QLC SSD can wear it out before they'd otherwise want to, if a power user that writes a lot of data and if conservative in their upgrade cycle.

For example I have one box with an MLC SSD, now showing slightly more than 50% life left. It's about 6 years old. If it was a QLC, I'd expect it'd be dead by now, but dead means... fewer available cells right? Not really dead, weird time to be in, where we are now finally wearing out SSDs and get to see what happens.

One thing is clear, to have plenty of free space on it. If you have, say 10GB of free space, and write 10GB, it's one wear cycle. If you have 30GB of free space, you have 1/3rd the wear cycling. Wear leveling does not make it that 1:1 an analogy but having the free space on an SSD is better than not.

Either way, this once every two months partition restore is not consequential. I just wonder why you even bother to do it if there isn't some problem warranting it. I mean if it's a highly sensitive system, two months seems far too long, like you should instead have a sandbox that flushes anything when you reboot and still not needing the two month period, just making the backups anyway and having them but hopefully not needing them *until* some event causes the need.

 

[KentState]

I would wager the repetitive operations you have to perform between restores ads an equal amount of wear. Depending on your SSD brand, you may be able to track the TBW over time and see the actual impact.

Anecdotal, but I tried to destroy my Sabrent PCIE 4 drives with mining Chia and after 100TB of plots, they barely showed any wear.

 

[IntelUser2000]

Maybe with QLC drives, consumers could finally see the effects of wear, although the drives aren't going to just fail but get too slow.

It's partially countered by the fact that the storage space is larger, so even if the per bit life is lower, if you have twice the storage the effective differences are half compared to say a TLC that has half the space.

Of course the differences aren't that big, but you can see why even for QLC it's not a problem. The older drives may have used 2-bit cell MLC, but they were very small. My X25-M is 80GB in capacity. So even if the new technology has 1/10th the lifespan, if the drive storage is 800GB the write cycle life will be same as the X25-M.

 

[bononos]

It's partially countered by the fact that the storage space is larger, so even if the per bit life is lower, if you have twice the storage the effective differences are half compared to say a TLC that has half the space.

Of course the differences aren't that big, but you can see why even for QLC it's not a problem. The older drives may have used 2-bit cell MLC, but they were very small. My X25-M is 80GB in capacity. So even if the new technology has 1/10th the lifespan, if the drive storage is 800GB the write cycle life will be same as the X25-M.

The TBW of QLC drives generally looks like 1/2 that of TLC drives (eg Transcends 500Gb qlc ssd - 100TBW). Some QLC drives like WD's SN350 is even worse - 100TBW for both 1Tb/2Tb models, now that is just downright bad.

 

[BonzaiDuck]

I still don't see the point of cloning the hard disk every two months. At one time I thought to clone a boot disk once a year as a fundamental backup.

The only time I would clone a disk now would occur when I want to swap in new hardware. When I was building a machine that would be primarily using NVME drives, I initially created the boot disk on an SATA SSD. I made the transition with a clone.

Otherwise, unless there is some reason to swap hardware, Macrium Reflect Workstation does scheduled backups on my local system with both differential and incremental daily backups. I've had to restore my NVME boot disk maybe two times in 2017 from those backups, and after that, never needed to.

 

[mindless1]

The TBW of QLC drives generally looks like 1/2 that of TLC drives (eg Transcends 500Gb qlc ssd - 100TBW). Some QLC drives like WD's SN350 is even worse - 100TBW for both 1Tb/2Tb models, now that is just downright bad.

That's more about warranty though, both that and their 3 year period. TBW is hard to equate to true lifespan since there is wear leveling and it's taken from remaining free space, while a typical (major brand SSD grade) QLC chip write endurance is somewhere around 1000 cycles last I checked.

The main take away from this topic should be that doing a partition restore is only a single write cycle, so bimonthly for a decade is a mere 60 cycles, and even that, only if the drive were nearly full.

 

[IntelUser2000]

The TBW of QLC drives generally looks like 1/2 that of TLC drives (eg Transcends 500Gb qlc ssd - 100TBW). Some QLC drives like WD's SN350 is even worse - 100TBW for both 1Tb/2Tb models, now that is just downright bad.

And that's a problem how?

You'll be hard pressed to reach 10TB, nevermind an order of magnitude greater than that.

 

[swapjim]

My drive is the Samsung SSD 850 EVO 500GB which as I see is an MLC drive, so it's good news!

I've got one more question, out of curiosity: a fresh Win10 installation is around 25GB of data. Doing a restore is like writing 500GB or 25GB of data to the disk?

 

[KentState]

A partition restore is whatever the size of the backup is. Depending on the software used, it either destroys the partition and recreates them and writes the files or is a complete bit by bit over write. Not familiar with the app you use.

 

[swapjim]

Thanks everyone!

 

[aigomorla]

It's extremely difficult to wear down your SSD as a consumer.

not true... Games are now in excess of 60-70gb.
UE4 updates requires the entire data rewritten.
That means each update you are writting 60-70gb of data.
Call of Duty Modern Warfare was in excess of 120GB... Compound the fact CoD would do serveral updates each week., yeah, your (on average 250TB writes) aren't going to last you past a few sequels in that game.

It is difficult for a typical end user to eat though those writes, but no so much for a gamer who probably will go though several TB writes each year just on updates.

Watch vendors start throwing gaming edition SSD's with double the write endurance, and charge enterprise prices for them, oh wait they already do that, and gpu price increases due to mining isn't helping either.

 

[kschendel]

Several TB per year is nothing. 50+ TB/year might be getting interesting. If a gamer is actually doing a 60+ GB update multiple times a week, I suppose you could get there; but then I'd love to see your ISP data bill.

 

[igor_kavinski]

Just use CrystalDiskInfo or Samsung's own Magician software to keep track of your SSD's health. If it starts getting lower than you are comfortable with, go easy on the partition restores.

 

[mikeymikec]

Remember that TBW in warranty terms means very little in terms of actual longevity: They're basically saying that it should always reach that figure unless something is very wrong. The Samsung 850 PRO 256GB has a TBW of 150GB yet an 840 PRO (which doesn't have a TBW warranty figure) in techreport's endurance trials managed >2400TB.

Even the early casualties in that trial lasted about 800TB.

 

[igor_kavinski]

Are we overthinking SSD endurance? : chia (reddit.com)



Looking at that graph, all I can say is, WOW!

Totally unexpected, apart from Samsung SSDs performing so great, is the relatively weak performance of Optane. Damn, do these Koreans over-engineer stuff!

 

[mikeymikec]

Are we overthinking SSD endurance? : chia (reddit.com)

View attachment 54965

Looking at that graph, all I can say is, WOW!

Totally unexpected, apart from Samsung SSDs performing so great, is the relatively weak performance of Optane. Damn, do these Koreans over-engineer stuff!

It's weird to think of that WD Blue at the bottom as laughable with 'only 82TB' written, but considering the last time I checked that my ~6 year old 840 PRO has only had 14TB written, the WD Blue isn't that terrible really.

I'm surprised to see the 860 PRO so vastly beaten by some TLC Samsung EVOs though, and again I have to remind myself that its figure is similar to the 840 PRO's figure, which is what earnt my loyalty to Samsung in the first place

 

[aigomorla]

Totally unexpected, apart from Samsung SSDs performing so great, is the relatively weak performance of Optane.

You are aware that 3dxpoint is about the most resilient form of memory you can have on a SSD in terms of write protection.
Its actually leagues greater in data protection then anything samsung has, which is why its used more in mission critcal systems that relies on heavy I/O caching which 3D Xpoint was designed for.

They even ranked 3DXpoint up with SLC in terms of resiliance, so no, Optane is not weak, im lost in how you got that its weak when its up at the top of that chart sitting at 1.

 

[igor_kavinski]

They even ranked 3DXpoint up with SLC in terms of resiliance, so no, Optane is not weak, im lost in how you got that its weak when its up at the top of that chart sitting at 1.

Perhaps Intel's Optane marketing slides are to blame. They advertised magnitudes more endurance than anything flash could offer. Yet, No.2 is Samsung flash not that far behind. I would have expected Optane to have at least 10 times more endurance than the best flash technology. Of course, it's possible that Intel's enterprise Optane products like the P4801X have that kind of endurance. Pity it wasn't tested here.

 

[IntelUser2000]

Perhaps Intel's Optane marketing slides are to blame. They advertised magnitudes more endurance than anything flash could offer.

Interesting. They got 7.7PB when the rated is 5.1PB. The 375GB version of the P4800X is 20PB rated, and the later versions are rated at 40PB.

Don't forget, they do live up to the claims in the DIMM version. The 256GB version has 500PB rating for writes, which means 5 years writes at maximum speed 24/7, and Intel has said it'll go way beyond that.

The DIMM version also has 200-400ns latency which is 500x lower than SSDs. It's the SSD versions that are hobbled.