Should I worry about remuxing on my recording SSD, effectively doubling usage, or is SSD write longevity a non-issue compared to years ago?

WarMom

New Member
I write my OBS recordings to a Samsung 860 QVO 1TB SSD, which is only used for recording / encoding and editing video, and store my files long-term on a 2TB HDD or some externals. The question is 'where should I remux'?

If I move my MKVs to the HDD first, I'm theoretically saving extra writing to my render / recording SSD, but because I run OBS as admin I can't drag-n-drop them in and it's slightly more cumbersome to get the remuxing going. The HDD - at present, at least - is a bit closer to full and if I were to remux a bigger big of stuff it might not all fit.

So remuxing on SSD is slightly more convenient, buuuuut there's the obvious problem of 'I'm essentially doubling the amount I write to that SSD'. I record at CRF14, and 1080p / 1440p and 30/60fps depends on what I'm recording (for example there's no point recording a 1080/30 console game at 1440/60). I could write upwards of 50GB for 3 hours, depending. Remuxing that MKV to MP4 bloats my usage from 50GB to 100GB. How much I do it varies; I could be as low as 100GB some weeks, as much as a few hundred in others.

Is SSD write longevity really much of an issue any more, or should I really commit to remuxing on the HDD? One of the reasons I switched to OBS from Nvidia Share for recording was that I could store my replay buffer in RAM rather than have Share perpetually writing to SSD with its bloated file sizes. Not that I'd ever go back, of course, it's just one reason.
 
D

Deleted member 121471

Saving to the SSD will wear out each used cell a bit more but, as long as you have TRIM enabled and you NEVER fill your SSD above 85 to 90%, it shouldn't affect SSD lifespan that much, if you stick to reputable brands and keep your eyes peeled for any changes in fabrication. Once you get past the 90ish % mark, the automatic wear leveling solutions do not work as well or at all, as you near 100% capacity filled.

Fortunately, you do not have a model of SSD that has any history of fabrication woes. Some companies manufactured some really horrible, cheaply made SSD models (crucial) or even changed controllers and other parts in the later half of a model's fabrication period to save a buck and that's what led to all the "SSD failing early" debacle. If I recall correctly, Samsung 840 EVO was one such model, as an example of a dud from a reputable brand.
 
Last edited by a moderator:

WarMom

New Member
Thanks for the response! Thankfully the drive doesn't tend to ever get close to full; I figure that 1TB would be plenty to hold all the files for a massive project if one arises in the future but at present it has only ever held a couple to a few hundred gigs at one given time before the project is done or the files get shunted to the storage drive until I actively need them in editing.

So what would you say the lifespan might be? Like is it assumed that even in this (relatively) heavy use-case scenario where it's got 10% permanently overprovisioned and rarely filled up what's remaining at once, the drive will last until I just sort of naturally buy a new one as price per size decreases? Or in other words I've next to no need to be concerned about it wearing out within ~4 years?
 
D

Deleted member 121471

Roughly, you'd need to write 100GB per day (not read, just writes) for 10 years before your SSD reaches its expected maximum writes before failure. It's more likely that your SSD controller will fail before you have too many cells worn out.

That being said, one reason SSD are becoming cheaper is how bits are "packed" into each cell so it might be worth reading about SLC, MLC, TLC and QLC. They have different pro and cons and might matter when purchasing new hardware, namely endurance vs cost vs capacity per unit.
 
Last edited by a moderator:

Lawrence_SoCal

Active Member
As SSD NAND has changed from SLC to MLC to TLC and now some QLC, re-write capability has dropped significantly (as in 100K read/write ops to 10K to 3K and less). For TLC, QLC and future (less resilient) versions, keeping spare area up for wear leveling is important as noted earlier.
Leaving 15% spare vs 10% can make a difference. Though as noted, you'll probably be fine regardless. And a sequential write operation like with a video file is much less demanding on an SSD than lots of little random small writes (potential write amplification issues) as is the case on an OS drive

Whether a SSD or HDD, a backup is still important for anything you don't want to lose. Drives (regardless of type) can and will fail. So backup what you don't want to lose, and don't worry about if/when the drive will fail. And yes, you are likely to replace the SSD with a larger capacity drive long before it fails (as long as you get passed the first few weeks of usage)
 

WarMom

New Member
Cheers for the info, I'll do some reading for future use, but for now I'm just glad I can rest easy that my drive isn't going to crap out - and of course, anything significantly important, that's kept backed up on other internal and external drives.
 

koala

Active Member
In addition to the things already said: there is no real danger to the data on the drive. If the drive should really fail after 10-20 years, it will not just breakdown and instantly lose all data. Instead, only write performance will drop to a crawl when the replacement memory cells are used up. All data is still available.
 

Lawrence_SoCal

Active Member
Uh.. .granted this was on older SSDs (from 3-4+ yrs ago)... but for me and others I know... they just died/bricked. no data retrievable. not unlike a failed HDD (with those of us old enough to remember the freezer trick which sometimes (rarely) worked)

So I wouldn't worry more about a SSD with adequate spare space from dying any more than a HDD.
but either can fail at any moment - sometimes you'll have warning signs, but not something I'd count on... and that is why if the data is important to you, you'll have an off-system backup. And if you really don't want to lose it, you also have an off-site backup (and those backups need to be such that ransomware can't delete/encrypt all of them)
 

koala

Active Member
This is different to what I read. I studied these test reports where some magazine or test lab did nonstop SSD write tests until the SSD died, then reported how long the ssd's lasted. As far as I remember, they reported the given guaranteed TB to write was reached by all of them, and usually twice the TB was also reached by them. Around this write volume (twice the guaranteed TB) write operations got slower, and only after that the device became readonly. Only very few really died completely.

Remember, newer SSD become readonly if the reserved sectors are used up, and if this is your Windows boot drive, Windows will not boot any more from it. This looks as if the drive is broken, but it's still readable if you boot from a different boot device. If the file system needs repair, it's also not possible to repair, because it is read only. In this case the device also looks as if it died, but it's still accessible. You need to image copy the raw SSD device to a new HDD/SSD and repair the filesystem on the new device.

You can monitor your SSD with ordinary SMART tools. These will report the current health state, which includes the remaining percentage of reserved volume. You will see a steady decline of this, so you can estimate when you need to replace the SSD: it's when the health gets under perhaps 10%. So there are no surprises.
 

FerretBomb

Active Member
Confirmed, I've had a couple of SSDs fail before. They didn't brick, they just became effectively read-only.

Really though, if you're worried about writes to an SSD, consider that you're writing repeatedly as you record. The remux will incur the same number of writes as just copying the file, since it's a simple video container rewrap, not anything like a re-encode.

Personally, I bought a cheap 150GB Kingston SSD as my recording disk. It's effectively disposable, but I've been using it for years now, recording 6-8 hours a day with no sign of problems and just copying the completed recording to a local spinning-rust archive disk and a NAS for safety (as I *did* have a spinning-rust disk die and take out all of my recordings).
 

Lawrence_SoCal

Active Member
My SSDs (and some clients) were bricked... dead... not read-only I am more than aware of difference, and of course attempted to read from alternate systems. Drives wouldn't even get a BIOS detect... on multiple systems, nor when trying using a few different eSATA and USB-SATA adapters, and vendor diagnostic tools
 
Top