TBW rating for SSDs - how to interpret?

[VirtualLarry]

TBW, if you didn't know, stands for "Total Bytes Written". Some people read it as "TeraBytes Written".

Most people interpret it as meaning total TB written to the drive.

But... what if... it doesn't natually mean "TeraBytes".

What if they mean, the "Total Bytes" of the drive, times the number of the rating?

Much like server drives are rated in DWPD - "Drive Writes Per Day". They mean, the number of times per day, that the entire drive can be effectively written to.

So, if TBW means a similar spec, then I think that some people are incorrect thinking that this automatically means TeraBytes. ESPECIALLY, if the mfg doesn't specify "TB" in the number, as in "TBW = 200TB". What if they just say "TBW = 200". You would have to go by the size of the drive.

Likewise, when a mfg has the same TBW spec, for various different drive sizes.

 

[BSim500]

Likewise, when a mfg has the same TBW spec, for various different drive sizes.

^ I find it really annoying when they do that as the number then becomes useless as a guide to P/E cycles. Probably the most accurate lifespan prediction used to be back when tech sites actually made an attempt to measure it. Eg, Anandtech used to measure how many cycles it took to drop from 100% to 99% SMART attributes then extrapolate. Not perfect, but as close as you'll get.

These days though, most SSD 'reviews' are "low information" glorified infomercials.

 

[BonzaiDuck]

My understanding of it was initially nurtured by scanning through some reviews -- some, I think, were at Anandtech. So last I'd looked, there was a range of TBW with maybe a Samsung SSD in the 900 TB to 1+ PB range. Sometimes I could find a particular drive's "TBW" spec in the spec sheet, other times not.

So the issue, from the standpoint of a "just-get-it-done-right" enthusiast, might be the consistency between the proprietary SSD software which shows a growing TBW over the drive's usage, and these accumulated TBW specs we see reported.

Now I'll be curious as to how someone illuminates Larry's initial question, observation and post.

 

[dave_the_nerd]

^ I find it really annoying when they do that as the number then becomes useless as a guide to P/E cycles. Probably the most accurate lifespan prediction used to be back when tech sites actually made an attempt to measure it. Eg, Anandtech used to measure how many cycles it took to drop from 100% to 99% SMART attributes then extrapolate. Not perfect, but as close as you'll get.

These days though, most SSD 'reviews' are "low information" glorified infomercials.

They stopped checking that stuff because it turned out even crap SSDs had 10+ years of projected use.

 

[dave_the_nerd]

TBW, if you didn't know, stands for "Total Bytes Written". Some people read it as "TeraBytes Written".

Most people interpret it as meaning total TB written to the drive.

But... what if... it doesn't natually mean "TeraBytes".

What if they mean, the "Total Bytes" of the drive, times the number of the rating?

Much like server drives are rated in DWPD - "Drive Writes Per Day". They mean, the number of times per day, that the entire drive can be effectively written to.

So, if TBW means a similar spec, then I think that some people are incorrect thinking that this automatically means TeraBytes. ESPECIALLY, if the mfg doesn't specify "TB" in the number, as in "TBW = 200TB". What if they just say "TBW = 200". You would have to go by the size of the drive.

Likewise, when a mfg has the same TBW spec, for various different drive sizes.

All the worrying about definitions of words is kind of silly - you can look them up.

Example: http://www.samsung.com/semiconductor/minisite/ssd/product/consumer/850pro.html

Definition of TBW as "total bytes written." http://www.samsung.com/semiconductor/minisite/ssd/support/warranty.html

You've got a drive with a ten year / 300TBW warranty, which has 1TB of 6k P/E flash (so, like, 6PB projected lifetime writes.) And it'll probably last 20 years.

And that's the thing - the warranty terms are largely arbitrary and way underbid. That's why drives with different sizes have the same TBW rating.

So don't worry about it.

 

[Elixer]

It is usually THW (Total host writes), and that can be expressed differently, but usually, it is in gigabytes.
They even have TNW (total NAND writes) which could be different that THW as well, and it is up to the manufacture to say what unit of measurement those are in.

I agree with Dave, no real reason to get into the fine details, unless you are troubleshooting something, and wonder what is writing 100's of gigabytes of data a day to your SSD.

 

[BonzaiDuck]

It is usually THW (Total host writes), and that can be expressed differently, but usually, it is in gigabytes.
They even have TNW (total NAND writes) which could be different that THW as well, and it is up to the manufacture to say what unit of measurement those are in.

I agree with Dave, no real reason to get into the fine details, unless you are troubleshooting something, and wonder what is writing 100's of gigabytes of data a day to your SSD.

Well, it probably made the difference in my decision to get a Pro versus an EVO, since the drive has three major partitions/volumes on it in addition to the sys-reserved and EFI.

But I don't think anything other than obsolescence, much higher capacity and other factors will change the fact that this drive will die after I do. Maybe my modest guesses about when I'll punch out are less than they need be, but it could shake out like that.

 

[Elixer]

I got a few SSDs that went belly up with absolutely 0 indications of anything wrong.
They basically went into a coma, and never came back out. I suspect firmware bugs/issues, and there is just no way to fix them into working units again.

 

[BonzaiDuck]

I got a few SSDs that went belly up with absolutely 0 indications of anything wrong.
They basically went into a coma, and never came back out. I suspect firmware bugs/issues, and there is just no way to fix them into working units again.

I could imagine that they might be resurrected but I wouldn't know.

I killed a 250GB 840 EVO when I made the mistake of pulling the plug on it while it was in "sleep state" during a secure erase. So the drive is "locked." Someone here had indicated a password or some kind written on the drive's paper label that would enable you to revive it, but I never pursued it.

I'm embarrassed to say that I only recently discovered how replacing a laptop spinner (in a laptop) with an SSD did the reverse of what was intended -- extend the battery time by reducing power consumption. All of those 2.5" devices run on 5V with no 12V input. An SSD typically uses 1.7A, while a spinner only runs at 0.85A -- more or less depending.

I've been on these forums for years and you'd think I would've figured all that out before I replaced the WD Blue 2.5" in the first place.

 

[Elixer]

A locked SSD is different that a SSD that won't even show up in the BIOS anymore.
(I took them apart, and some boards do have JTAG points on them, so, I assume something could be used to 'wake' them up or perhaps flash firmware, but, no company would give that info out)
Typically, for a locked SSD, you actually do pull the power on said SSD, then plug it back in, and that will unlock it.
That is a protection feature, since you don't want to secure erase the OS drive usually.

 

[FFFF]

A locked SSD is different that a SSD that won't even show up in the BIOS anymore.
(I took them apart, and some boards do have JTAG points on them, so, I assume something could be used to 'wake' them up or perhaps flash firmware, but, no company would give that info out)
Typically, for a locked SSD, you actually do pull the power on said SSD, then plug it back in, and that will unlock it.
That is a protection feature, since you don't want to secure erase the OS drive usually.

Even if a SSD doesn't show up in the BIOS doesn't automatically mean it's dead. In some cases it's only the firmware who bricked itself to protect from permanent damage so it could easily be reflashed by someone with the right tools. A SSD is dead for good only when the controller or the NAND cells themselves die out.

 

[Elixer]

Even if a SSD doesn't show up in the BIOS doesn't automatically mean it's dead. In some cases it's only the firmware who bricked itself to protect from permanent damage so it could easily be reflashed by someone with the right tools. A SSD is dead for good only when the controller or the NAND cells themselves die out.

That is what I said...via the JTAG.

 

[BonzaiDuck]

A locked SSD is different that a SSD that won't even show up in the BIOS anymore.
(I took them apart, and some boards do have JTAG points on them, so, I assume something could be used to 'wake' them up or perhaps flash firmware, but, no company would give that info out)
Typically, for a locked SSD, you actually do pull the power on said SSD, then plug it back in, and that will unlock it.
That is a protection feature, since you don't want to secure erase the OS drive usually.

I think it's time I set up one of the machines -- possibly even my laptop using a USB docking station, and attempt what you describe there.

Thanks.

 

[VirtualLarry]

Typically, for a locked SSD, you actually do pull the power on said SSD, then plug it back in, and that will unlock it.
That is a protection feature, since you don't want to secure erase the OS drive usually.

It's not for "locked" SSDs, it's for SSDs that are "security frozen", it's totally different.

Locked == has a password set, need to use ATA password command to give it a password to unlock.

Security Frozen == cannot set a password, which is necessary to do a secure erase. Must remove and re-apply power to "unfreeze". NOT "unlock".

 

[BonzaiDuck]

It's not for "locked" SSDs, it's for SSDs that are "security frozen", it's totally different.

Locked == has a password set, need to use ATA password command to give it a password to unlock.

Security Frozen == cannot set a password, which is necessary to do a secure erase. Must remove and re-apply power to "unfreeze". NOT "unlock".

I'm not sure which situation I'm in with that 256 GB EVO. We discussed this before -- I think it was you who had volunteered the information about the password.

I was using PartED Magic to Secure-Erase. that program puts the system and or the SSD to sleep. I was confused, and shut the system down. After that -- no EVO.

 

[dave.nelson@sitechsolutio]

All the worrying about definitions of words is kind of silly - you can look them up.

Example: http://www.samsung.com/semiconductor/minisite/ssd/product/consumer/850pro.html

Definition of TBW as "total bytes written." http://www.samsung.com/semiconductor/minisite/ssd/support/warranty.html

You've got a drive with a ten year / 300TBW warranty, which has 1TB of 6k P/E flash (so, like, 6PB projected lifetime writes.) And it'll probably last 20 years.

And that's the thing - the warranty terms are largely arbitrary and way underbid. That's why drives with different sizes have the same TBW rating.

So don't worry about it.

hi Dave and everyone
in this very old thread
just found it when searching about the same subject TBW after seeing drives, both SSD and HDD 's with this specification

Dave, that first link you gave to the Samsung site clearly states Terabytes written

"With heavy workloads, endurance is everything. The 850 PRO is a game changer, withstanding up to 600 terabytes written (TBW) "

my bold ( maybe you typo'ed and really meant to say that ?

Even the datasheet states that

https://download.semiconductor.samsung.com/resources/data-sheet/Samsung_SSD_850_PRO_Data_Sheet_Rev_3.pdf

Samsung Data Sheet

Enhanced Endurance and ReliabilityEnhanced Endurance Samsung's 3D V-NAND technology is built to handle 150 Tera Bytes Written (TBW) for 128GB and 256GB, 300 TBW for 512GB and1TB, 450TBW for 2TB and 600TBW for 4TB. Plus, it comes with an industry-leading tenyear

makes me so much happier as eg. 150 total bytes written is meaningless on a say 1TB drive

Best regards
Dave

 

[mikeymikec]

I'm embarrassed to say that I only recently discovered how replacing a laptop spinner (in a laptop) with an SSD did the reverse of what was intended -- extend the battery time by reducing power consumption. All of those 2.5" devices run on 5V with no 12V input. An SSD typically uses 1.7A, while a spinner only runs at 0.85A -- more or less depending.

IMO your original conclusion was the correct one - Windows 10 for example (from power off) takes about 1.5 minutes from the beginning of the OS boot process to the point of loading up the desktop icons then maybe another 10 minutes before the OS has stopped saturating the HDD.

With a SSD in my experience most laptops' boot times are shaved down to ~30 seconds without the extra saturation time. Then shave off all that power usage of the HDD being saturated and all the extra wait times caused by the HDD's comparatively slow response times when the user and OS asks the fixed storage to do something. Overall a SSD is going to spend a lot less time at max power usage than a HDD.

Also, I checked the specs of a random 'current' hard drive and compared them to a Samsung 870 EVO:

Samsung 870 EVO 1TB: Max 4W, idle 0.003W
Seagate ST1000LM048: Doesn't state maximum, "average read/write" is 1.7W, idle 0.45W

While I personally doubt that a SSD will hit max power usage in typical usage scenarios (my bet is sustained writes will provoke that scenario), but let's assume that the moment you ask it to do anything it hits max power usage. In most scenarios in Windows it will get that job done an order of magnitude faster than a HDD and then enter a much lower usage idle state than the hard drive.

*Maybe* a test could be concocted that shows a laptop being more power efficient than a hard drive, say by having both computers doing something like sustained writes non-stop, no factoring in how long a job takes, no idle time, etc., but I don't think that's a particularly realistic workload.