Are there wear problems from partitioning a SSD?

[Wirko]

Physics. This is because the logic gates are programmed by a change in physical state by means of an electrical current, and the constant changes in the physical state of the memory cause the material to lose its property over time, ceasing to function correctly, it's similar in concept to electromigration in CPUs. The more states that a given cell can accomodate, the more sensitive to this electrical current it becomes, and this is the reason why newer lithography nodes with ever smaller logic gates and multi-bit NAND cells all have a cost in write endurance.

Is it correct to call the programming process a change in physical state? No chemical reactions occur and atoms are not moved, it's just electrons that get pushed into the insulating gate, using the quantum tunneling effect. (However, in the slow process of degradation, atoms are moved and/or chemical changes occur.)

Single bit per cell NAND (SLC), operates with a single programmable state of "0" and "1", while the 2 bits per cell MLC design does "0" "1" and "2", for example. TLC and QLC have 3 and 4 programmable states in addition to empty.

Not true ... sadly. The misleding names are not helpful here. QLC stores four bits per cell, which means 2^4 = 16 distinct levels of charge. Those 16 may include "empty" or not, who knows.

It would even be possible to store something like 11 levels per cell, yielding 10 bits per 3 cells (11^3 = 1331 > 1024), or 13 levels per cell, yielding 11 bits per 3 cells (13^3 = 2197 > 2048). I know those are weird numbers but 232 layers is a weird number too.

 

[Dr. Dro]

Is it correct to call the programming process a change in physical state? No chemical reactions occur and atoms are not moved, it's just electrons that get pushed into the insulating gate, using the quantum tunneling effect. (However, in the slow process of degradation, atoms are moved and/or chemical changes occur.)


Not true ... sadly. The misleding names are not helpful here. QLC stores four bits per cell, which means 2^4 = 16 distinct levels of charge. Those 16 may include "empty" or not, who knows.

It would even be possible to store something like 11 levels per cell, yielding 10 bits per 3 cells (11^2 = 1331 > 1024), or 13 levels per cell, yielding 11 bits per 3 cells (13^3 = 2197 > 2048). I know those are weird numbers but 232 layers is a weird number too.

I've definitely oversimplified it a bit, but you are correct. I would argue for the change in physical state because a significant level of wear occurs, but perhaps it is just the speed in which it does occur naturally. It's fascinating stuff to which I'll admit I still have plenty to learn about

 

[Ferather]

I remember hybrid mechanicals with nand (SSHD), according to Google they still exist.

 

[Shrek]

I have three SeaGate FireCuda 2TB 3.5" SSHDs and like them; it is what I use at the moment.

 

[chrcoluk]

It wont add wear, the sectors are not 1 to 1 assigned like on spindles.

Manufacturers even recommend leaving unpartitioned space for over provisioning purposes.

The utilisation level is what affects wear as that reduces the ability of wear levelling.

 

[Shrek]

Now THAT is a useful observation.

 

[Wirko]

It wont add wear, the sectors are not 1 to 1 assigned like on spindles.

Manufacturers even recommend leaving unpartitioned space for over provisioning purposes.

The utilisation level is what affects wear as that reduces the ability of wear levelling.

This document by Intel explains some basics and also has some calculations for Intel drives. Basically, any free space within a partition is as good as unpartitioned (=guaranteed to be always free) space. As long as you're able to keep ~20% of total drive space free, most of the time, you're good, and you gain a lot of endurance. Most of the time doesn't even mean all of the time, so you can still do some large transfers that fill up all of the drive occasionally.

Edit: even on spindles it's not 1 to 1. The difference is that the mapping doesn't change every day, it only does when the drive finds new bad sectors.

 

[Count von Schwalbe]

To make it more confusing, Optane (3DXPoint) was true phase change storage; not that it is relevant now that it has been phased out.

 

[Edwired]

To make it more confusing, Optane (3DXPoint) was true phase change storage; not that it is relevant now that it has been phased out.

Intel thought it be a game changer but they were losing money in the whole deal the last i checked it

 

[Wirko]

To make it more confusing, Optane (3DXPoint) was true phase change storage; not that it is relevant now that it has been phased out.

I see what you did there; and yes, it just changed to vapour phase.

Seriously though, the phase change in Optane is a change in crystalline or amorphous structure, so it's a rearrangement of atoms/molecules, right?

 

[Mussels]

What if I make a small partition on a solid-state drive and then write to it a lot, will I cause unbalanced wear? or can the drive compensate?

Since over provisioning leaves a small unused partition and helps spread writes out over the drive, I assume the drive itself will spread the wear out
It's not being mapped like a mech drive, your partition still includes a portion of multiple memory chips, if not all of them


My OS SSD's always have a 100GB C: partition to reduce data loss and effort when reinstalling an OS, and i've not seen any increase in wear

 

[ThrashZone]

It wont add wear, the sectors are not 1 to 1 assigned like on spindles.

Manufacturers even recommend leaving unpartitioned space for over provisioning purposes.

The utilisation level is what affects wear as that reduces the ability of wear levelling.

Hi,
Yep if you've ever used samsung magician and had it do the over improvising for you you'd see unallocated space after this is all the firmware needs.

I'll add the firmware also only needs zero activity to do it's work/ hibernation...

 

[OneMoar]

the fawks are you people smoking

No thats not how it works SSDs don't have physical partitions like a mechanical disk
the partition is just a box. where the data is put in the box is entirely up to the controller and the controller is going to manage its wear leveling automatically
can we please stop even thinking about SSD Endurance in the Year 2022 its completely irrelevant
./endofthread

 

[lexluthermiester]

and yeah, manual TRIM was possible on XP and Vista, they just weren't designed for automatic maintenance of SSDs at all.

ALL modern SSDs handle TRIM functions native to the drive controller. OS compatibility is no longer a concern and has not been since around 2014.

I am still trying to understand why NAND wears out, even the good stuff seems only good for maybe 600 writes.

The technical details are complicated, however the simple expaination is as follows: All NAND cells require phased voltage applications to change it's data state. With each change, the voltage degrades the cell to a small degree. With SLC, only a single voltage change is applied to set data and one to reset to zero. With MLC, two voltage applications are required and then one to reset to zero. TLC, three voltage applications are required and one to reset to zero. QLC, four+one. The voltage applied at each step is increased with each application on a per-write basis. This effect is cumulative, IE, as the cell wears more voltage must be used to alter it's states. At a certain point the NAND cell will literally burn out from extended voltage exposure thus rendering it non-functional.

So now you can understand the electro-chemistry involved in NAND flash is destructive by design. With each data-change function the data cells involved are degraded. Because of how NAND works, this is an unavoidable result. So the fewer changes made, the better and the more refined the chemical process(resulting in great resistance to degradation and better durability) the better. For this reason SLC or pSLC will always be the best option. MLC/pMLC is the next best and so on.

Perhaps it's now easier to understand why I have such a problem with QLC based NAND and why I consider it garbage for Boot/OS usage.

 

[dnm_TX]

OFF TOPIC but to to lighten the mood a bit.
@Shrek some wise guy,playing smart,trying to make a buck,selling them for almost double on Craigslist(or they could be very well the free once from Microcenter) :

 

[Mussels]

the fawks are you people smoking

No thats not how it works SSDs don't have physical partitions like a mechanical disk
the partition is just a box. where the data is put in the box is entirely up to the controller and the controller is going to manage its wear leveling automatically
can we please stop even thinking about SSD Endurance in the Year 2022 its completely irrelevant
./endofthread

People aren't gunna learn if they don't ask questions and get answers

 

[lexluthermiester]

the fawks are you people smoking

No thats not how it works SSDs don't have physical partitions like a mechanical disk
the partition is just a box. where the data is put in the box is entirely up to the controller and the controller is going to manage its wear leveling automatically
can we please stop even thinking about SSD Endurance in the Year 2022 its completely irrelevant
./endofthread

People aren't gunna learn if they don't ask questions and get answers

Then there are those who think they have the answers, but don't because they either misunderstand key parts of the equation or simply never learned the correct information to begin with.

 

[delshay]

I always use "Over Provisioning" This drive is around 8 - 10 years old. ...It eat's up 190GB on a 2TB SSD recommended setting (default).

 

[P4-630]

I always use "Over Provisioning" This drive is around 8 - 10 years old. ...It eat's up 190GB on a 2TB SSD recommended setting (default).

I also use the recommended magician overprovisioning on all my SSD's/NVMe's. (don't use mechanical drives anymore)

 

[Shrek]

OFF TOPIC but to to lighten the mood a bit.
@Shrek some wise guy,playing smart,trying to make a buck,selling them for almost double on Craigslist(or they could be very well the free once from Microcenter) :

View attachment 257029

Yeh, I knew about the offer, but the store is too far away, so paid $25 on ebay

 

[Valantar]

I always use "Over Provisioning" This drive is around 8 - 10 years old. ...It eat's up 190GB on a 2TB SSD recommended setting (default).

The 850 Pro 2TB launched in mid 2015, so 7 years at the most

 

[delshay]

The 850 Pro 2TB launched in mid 2015, so 7 years at the most

OK. can't remember when I bought it, but I know I got it when it first came out.

EDIT: 1TB came out in 2014 which is 8 years old ( I have this drive also). So I got mixed up which year the drives came out. 2TB came out a year later.

 

[P4-630]

I always use "Over Provisioning" This drive is around 8 - 10 years old. ...It eat's up 190GB on a 2TB SSD recommended setting (default).

Got an 850 Pro 512GB using since 2016 (did not set overprovisioning at that time though.)
But still going strong now using with over provisioning.

 

[Ferather]

....
P.S. For example. Here is my secondary SSD,barely used for it's age(at least 8+ years old) 1TB written but look at the hours. It's at 97% life.
But if logically thinking here that only extensive writing is killing them then i should be at 100% still:

View attachment 256909

Mine is about the same age, slightly less hours, but a lot more total writes, 3% more wear than yours.
Originally this drive was the primary for 7-8 years, with separate page file partition.

 

[Palladium]

~10TB writes in 6 years is like a drop in the endurance bucket though.