Does reading from an SSD decrease its lifetime?

Question

SSD manufacturers advertise TBW (terabytes written). Why is "terabytes read" not part of the spec? Is that because reading from an SSD is virtually irrelevant for its lifetime?

Added: For a bit more context, assume we have a 2TB drive that we want to backup (e.g. to a NAS). If we backup every day, that's already about 700TB of reads per year. And if TBR is roughly comparable with the currently advertised TBW (i.e. around 1200) then we might consider doing the backup not daily but every other day to increase its expected lifetime from 2 to 4 years. Does it make sense?

Answer 1

When you change the electric field, the SSD does wear down. Writing will change the electric field. Reading itself is just getting the state of the SSD and in theory there is almost to no wear for doing just that.

Chips do die from just having current running through them, which correlates with the reading part.

Because almost everyone will write enough to the SSD that this causes the failure, the read is not even measured because failure through writes is most likely to occur anyway. If you decide to write data to an SSD once and then just keep it as backup, do note that there still is a lifetime. It depends on the drive how much that is, but I would estimate that to be 5 to 10 years, depending on how much the drive is being powered on in between.

Keep in mind, not plugging the drive into power for many years is not good either.

Answer 2

In principle, reading does not wear down an SSD, only writing does.

However, you are not the only one writing to the disk - the disk's firmware is always busy working on avoiding Wear leveling, which it does by consolidating partially-empty blocks into full blocks.

This means that even if the operating system shows the disk as idling, there is a good chance that read and write maintenance operations are being done, which are invisible to the operating system.

Answer 3

TBW is a decent specification and should be used if available.

Good quality SSD drives have ample capacity to use good cells and should normally last longer than the machine they are in .

Reading does not cause excessive wear and good drives can take much writing without excessive wear.

I have a 5 year old ThinkPad X1 with a 1 TB Samsung SSD. Its lifetime is estimated to be 600 TBW. It is at 60 or so TBW at this point and I have replaced the machine with a newer X1 with a similar 1 TB SSD and Windows 11.

Good drives should include the TBW spec and my Samsung drives do. Lenovo diagnostics report on current TBW used and so look for a counterpart on your machine.

SSD drives last a long time and are now better than good consumer Hard Drives.

Answer 4

Quoting Transcend:

A read disturbance problem, in short, occurs after a large amount of read tasks are performed to pages within the same block, causing the digital values of cells nearby the cells being read to change, and thus inducing data errors.

How does read disturbance happen?

Every NAND flash memory cell has a floating gate transistor with a control gate connected to the word-line, and a source and a drain connected to its neighboring cells on the same bit-line. The floating gate transistor is situated beneath the control gate. The amount of electrons stored in the floating gate determines the threshold voltage of the transistor.

To find out if there are any electrons trapped on a particular floating gate, the memory device must read the entire word. In order to read the cells in a selected row (shown in light green), a higher voltage must be applied to its neighboring unselected word-lines (shown in deep green) in the same block. In the meantime, one selected cell will be read per bit-line for the device to determine its digital value, i.e. whether it stores a 0 or a 1. The high voltage applied on the neighboring transistor gates attracts electrons to the floating gate, slightly raising the cell threshold voltage with each read, and “disturbing” the cells in the process. Over time, the threshold voltage of a cell in an “un-programmed” state, which means that it stores a 1, increases and accumulates enough to eventually shift to a “programmed” state, which means that it stores a 0. This is known as the Read Disturbance phenomenon. The change of state is an irreversible process, and once changed, the bit value will not “flip” back unless the block is otherwise erased.

Transcend’s solutions

Read disturbance can be reduced by minimizing excessive reads. Transcend offers three different solutions to address this problem.

1. Wear leveling algorithm: This feature spreads the NAND flash memory cell usage over the available memory array, making sure that the data are equally written within the block.
2. Early move: This feature detects and corrects potential data errors. If error bits in a block reach upper limit, then the data should be moved to another block and the original block should be erased. (Note: Some products do not have this feature.)
3. Read retry: This feature is designed for flash memory to adjust the read reference voltage and eliminate the read error.

Delkin also addresses this issue:

When NAND flash memory is used only for read operations, many designers believe that they do not have to worry about product lifespan, since no writing is occurring. In reality, even read operations can eventually wear the flash memory out, as repeated read operations can lead to inadvertent reprogramming and damage caused by cross-coupling noise. When these issues occur, they are called read disturb errors. Fortunately, industrial-grade NAND flash memory typically comes with read leveling and ECC integrated to mitigate these risks.

On the other hand, I haven't seen any read load limit in the the specifications for regular SSDs, so it's probably not a big issue (because the the manufacturers have taken counter measures), unlike the write load limit.