Does deleting stuffs off your SSD lowers its lifespan?

[baK1ikan]

I got myself a small 480GB SATA SSD w/ USB-C enclosure thing for Video Editing stuffs. The idea is, i can use the SSD for scratch disk & a place to put everything im currently working on, and then move them to my HDD once im done. I figured out that editing videos directly from a slow ass HDD affects your video playback performance terribly.

 

My concern is, if i keep deleting (not writing) files on the SSD, will it eat my SSD lifespan away like what happens when it writes? Thanks in advance.

[filpo]

2 minutes ago, baK1ikan said:

My concern is, if i keep deleting (not writing) files on the SSD, will it eat my SSD lifespan away like what happens when it writes? Thanks in advance.

depends how much your deleting every time

[Naijin]

7 minutes ago, baK1ikan said:

I got myself a small 480GB SATA SSD w/ USB-C enclosure thing for Video Editing stuffs. The idea is, i can use the SSD for scratch disk & a place to put everything im currently working on, and then move them to my HDD once im done. I figured out that editing videos directly from a slow ass HDD affects your video playback performance terribly.

 

My concern is, if i keep deleting (not writing) files on the SSD, will it eat my SSD lifespan away like what happens when it writes? Thanks in advance.

Look at the TBW lifespan (terabytes written) of your specific model. Constantly writing is bad for an SSD, but a decent SSD will have a lot higher TBW to basically never reach for a normal consumer. You should also be able to check health (SMART data and total TBW) with some tools like CrystalDiskInfo.

[Ashleyyyy]

as far as i'm aware not really. SSD's usually have a max drive writes in the specifications you can look up for your SSD. you can then use a program like CrystalDiskInfo to see how much you've written to the drive. 

example here's my OS drive:

 

and here's what Samsung says:

 

 

so with my almost 30 terabyte written i'm about a third and a bit towards the cap for guaranteed reliabilty. 

in any case, you'll want to make backups of all your stuff always just in case, but this gives you a good indicator. 

Edited May 9, 2023 by Ashley MLP Fangirl
typo

[baK1ikan]

6 minutes ago, Naijin said:

Look at the TBW lifespan of your specific model. Constantly writing is bad for an SSD, but a decent SSD will have a lot higher TBW to basically never reach for a normal consumer. You should also be able to check health (SMART data and total TBW) with some tools like CrystalDiskInfo.

y---yeah i know. what i'm asking is whether DELETING stuffs will eat your TBW out or not.

 

i got a Samsung 870 or something whatever. I figured if kept writing a certain amount of data every once in a while, it'd last me quite a long time. what i hadn't realized is, what if i kept deleting stuffs from it

[Ashleyyyy]

1 minute ago, baK1ikan said:

y---yeah i know. what i'm asking is whether DELETING stuffs will eat your TBW out or not.

 

i got a Samsung 840 or something whatever. I figured if kept writing a certain amount of data every once in a while, it'd last me quite a long time. what i hadn't realized is, what if i kept deleting stuffs from it

i don't think so. here's an article explaining it: 

https://www.ontrack.com/en-us/blog/how-long-do-ssds-really-last

basically what wears it out is if you delete something and then put new data on it. that is measured in the write cycles myself and others have already mentioned. 

[porina]

You can't delete stuff until you've written it there in the first place, so I see it as going together. Basically just look at the writes.

[mariushm]

When you delete stuff, there's just some bytes changed, adding some records to a list, telling the ssd controller "these blocks of flash memory can be erased and made ready for new writes" 

 

That erasing process causes wear on the flash memory and after a lot of erases, that memory area becomes read only, can not guarantee it could store new data reliably. 

 

Think of the flash memory arranged in  24 MB or 32 MB or 64 MB chunks, each chunk having lots of 512 byte or 4096 byte "pages"  - the SSD controller can write a random page in one of these chunks, but can't overwrite it. It also can't erase an individual page,  it must erase the whole chunk. 

 

So when you write stuff to the SSD, the SSD controller finds empty pages in random chunks of flash memory and puts data there, trying to spread it across a lot of memory chips, to get faster write speeds by writing in parallel into multiple chips (or layers of a chip)

If you want to overwrite data in one of those small 512-4096 byte pages the SSD controller can't overwrite, so it just finds another available page in some random chunk, reads that data from the old page, makes the changes, writes the data in the new page and marks that old page for deletion. 

From time to time or when it needs free space to write new stuff, the SSD controller looks at chunks and where it finds a chunk with let's say 90% of its pages marked for deletion (having old data), it will copy the pages that still have valid data to other chunks and then erase the whole chunk to make the whole chunk available for writing again ... and by doing that erase, the whole chunk is now a bit more worn out. 

 

So basically , on a new 250 GB drive, you could write 200 GB of movie, erase those 200 GB, and the drive's health would still be 100%, because the controller didn't actually erase any chunk yet, the content is just marked as "can be erased when needed". 

When you write another 100 GB,  the controller will write 50 GB in the cells that were never used so far and then it will have to erase some chunks to get another 50 GB of free disk space. So after writing 350 GB of data, you would have 50 GB worth of memory chunks erased once, and the rest of 200 GB never erased.

With TLC, you have 1000-3000 erases and then the memory chunk is done for. 

 

Modern SSDs use a portion of the flash memory chips in pseudo-SLC mode, where each cell of memory holds only 1 bit instead of 3-4 bits in the case of TLC or QLC.  By using this mode, these chunks can be erased more "gently" and this pseudo-SLC portion can do even 10-30k erases before it's worn out completely. 

So if you have content that only stays on the drive for an hour or so, the SSD controller may not even move the data from the pseudo-SLC write cache area into "permanent storage".  If this happens , the consequence of erasing the data from pseudo-slc cache is less serious, it does  less damage to the cells in theory. 

 

 

 

[baK1ikan]

10 minutes ago, porina said:

You can't delete stuff until you've written it there in the first place, so I see it as going together. Basically just look at the writes.

11 minutes ago, Ashley MLP Fangirl said:

i don't think so. here's an article explaining it: 

https://www.ontrack.com/en-us/blog/how-long-do-ssds-really-last

basically what wears it out is if you delete something and then put new data on it. that is measured in the write cycles myself and others have already mentioned. 

aoh okay that makes sense. so it depends on how much rewriting is done to a certain storage cell. thanks guys

[kokosnh]

I just want to point out, that to delete stuff from ssd, you first have to write it to it. 
The SSD endurance is measured in program erase cycles of it’s NAND, as you have to first write to the ssd to delete it,  the whole cycle count as one, ssd usually have 100-3000 P/E cycles dependent on NAND used.

 

And the SSD actually don't delete the data that you delete. It waits sometimes weeks, or months for the trim command, to do it all in one time, to better spread the other data on the NAND ( wear leveling algorithms ), and use the minimal P/E cycles possible, doing it.

[kokosnh]

5 hours ago, mariushm said:

the SSD controller can write a random page in one of these chunks

That is not accrued enough, in today dense NAND, you go from 0 to max page in „block”, some SSD use shadow sequence or other algorithms to minimalizm leak current, affecting other NAND cells, when writing data.
Usually you can write to the next page, from the last written in the block, but you usually can’t write the one of the earlier pages, as to not corrupt the data in its neighbor NAND page cells.
it’s not random. SSD technically can write to any random empty NAND page, but they don’t, as to not corrupt the data.

 

I know what you wanted to say, just wanted to clarify it for anyone reading.

 

[Timme]

5 hours ago, mariushm said:

Modern SSDs use a portion of the flash memory chips in pseudo-SLC mode, where each cell of memory holds only 1 bit instead of 3-4 bits in the case of TLC or QLC. By using this mode, these chunks can be erased more "gently" and this pseudo-SLC portion can do even 10-30k erases before it's worn out completely.

 

Wait, the memory is still tlc. From what I've gathered, the way SSDs imitate SLC is by finding empty 3D cells and simply writing 1 bit per cluster, which virtually grants that SLC read/write speeds, but the more you fill up your drive the fewer free nodes remain, and at some point, there won't be any "cache". Am I wrong?

[kokosnh]

Yes, the SSD do the Pseudo SLC by writing only the LSB (least significant bit) to the TLC, or QLC NAND cells, so limiting it to store only 1 bit (but the voltage range is lower, and you have more space for error, resulting in higher lifespan it that mode), so the capacity of the NAND cell used in Pseudo SLC mode, drops from respectively 3 bit (TLC) or 4 bit (QLC), to 1 bit (Pseudo SLC mode on TLC/QLC NAND).

In almous all modern SSD, the Pseudo SLC cache is dynamic (NAND can be in TLC/QLC mode, or Pseudo SLC mode), and the less free space you have on SSD, the less the capasity of the pseudoSLC cache you get.
And from what I gather, the SSD usually sacrifice whole CE lanes from controller for the Pseudo SLC mode buffer (no individual blocks, or pages), to get the best performance (of course, if it can, it can transform any TLC/QLC NAND in one of those Pseudo SLC, but then that whole buffer, cannot be used to store data, and you have to have also a space to empty the buffer to)
And It's still not as fast as real SLC NAND, but fast enough.