Tom's Hardware published an article last month on the findings of a YouTuber named HTWingNut who tested data stored on four SSD's. You can read the article for yourself, so I will not repeat it here. The bottom line from HTWingNut's experiment seems to be that data stored on unpowered SSD's begins to degrade after about two years, and that depends on whether the SDD is new or nearing the end of its useful life. This appears to mean that given the trade-off between the small amount of damage caused by writing data to an SSD and the fact that the data will be unreadable if left too long, we should be rewriting the data on all of our SSD's about once a year. In fact, this applies to all storage drives that are based on flash storage technology: SSD's, SD cards, and USB flash drives. Wikipedia corroborates this, citing a paper from 2017, so this has been known for a while. Wikipedia states, "SSDs based on NAND flash slowly leak charge when not powered, while heavily-used consumer drives may start losing data typically after one to two year[s] in storage."

This is important information because the standard wisdom when making data backups is to make three backups of all your important data, and those should be on devices that incorporate at least two different types of storage technology. This means most computer users will be backing up at least some of their data onto storage drives that are based on flash technology. Many readers may wonder why they should care at all because they backup all of their data into the cloud. If you feel comfortable with that, then I doubt I will be able to change your mind, but the little experience I have had with the cloud suggests to me that it is nowhere near as reliable as most people think. But hey, you be you.

I have heard that hard drives do not lose their data as a result of becoming demagnetized over time, but I have my doubts. I have seen several apparently perfectly good hard drives that I have put aside for months or years fail to boot after being reinstalled in a computer. These were not mechanical failures, because I could re-install an operating system to get them working in the computer again. I have not tested my suspicions by trying to refresh the files on any of my hard drives. Perhaps I will cover that in another article.

The results of HTWingNut's tests came as no surprise to me. I currently have in my possession about four dozen USB flash drives and SD cards of all sizes and ages. I have a 64 MB (not GB) drive, a 2 GB drive that was made in 2007, other 2 and 4 GB drives, and many 8 and 16 GB drives. And, I have many newer drives up to 256 GB in size. I frequently discover that drives that have been sitting for years cannot be read. Linux boot drives will not boot, USB flash drives appear to be unformatted, etc. This situation is a real mess that has given me headaches for nearly the past two decades. USB flash drives are an order of magnitude or more reliable than floppy disks were, but that isn't really saying much, since floppy disks were horribly unreliable.

Memory elements in flash storage devices store charges, like tiny capacitors. Those with charges have data values of 1, and those without charges have data values of 0. Electrons slowly and randomly leak out of flash memory elements, and as they do, the charges go down. When a charge falls below some threshold, it is read as 0. Data must be refreshed before this point is reached or it will be lost forever. "Refreshing" just means writing a "1" into every data location that currently contains a "1" to top off the charge. HTWingNut's cursory testing suggests this needs to happen about once a year to be safe.

Apparently this one-year recommendation was enough to finally push me over the edge, because last week I looked for software for refreshing the data on all of my storage drives. Everything I had read before this said data on USB Flash drives and SSD's should last for something like 10 years, but that wasn't exactly what I have experienced with all the failures I have seen over the years. I had already begun informally rotating some of my important data between USB drives a couple of years ago, but I didn't know how often it should be done until I read HTWingNut's article. I do suspect, however, that storage drives of different generations with different densities of capacitive elements may require different refresh cycle lengths. My guess is that the oldest flash drives with the largest physical dimensions of memory elements--thus, the most stored electrons, especially those made before say 2010--may not need to be refreshed as often as those that were manufactured this year. I will share below what I discovered about this during the refresh of my old drives that I just performed.

The best way of refreshing storage drives is to use software specifically designed for that purpose. I installed onto one of my Windows 7 laptops a free program called DiskFresh that supposedly runs on every version of Windows between XP and 11. It is very easy to use and seems to work well. Unfortunately, Linux doesn't seem to have any software designed specifically for refreshing data on storage drives, but some people have suggested using the badblocks command to perform a non-destructive read-write test. One article said, "While badblocks can detect physical issues on your hard drive, it's not capable of fixing them." I wondered about that but soon realized that repairing the physical media is not the same thing as pumping electrons into a capacitive memory element. The Linux command "sudo badblocks -nsv /dev/device" may have refreshed my USB NVMe SSD that Windows 7 does not support, but I first decided to try installing a Micron NVMe driver onto my Window 7 laptop, so I could use DiskFresh. As usual with random Windows drivers, Micron's driver didn't work. So that leaves me with badblocks, when I find the time to try it.

DiskFresh notifies you when data on a disk is too far gone be refreshed. Instead of identifying individual bad files, it tells you which sectors are unreadable with the error "Bad or Inaccessible location at sector xxx". I had old flash drives with data stored on them as far back as before 2010. Some of the sectors with older files could not be recovered. However, with the exception of one bad sector on one drive, drives on which data had been stored for four years or less seemed to have had no problems. Several drives with files in the 5-10 year age range had dozens of files that could not be recovered. two or three even had hundreds of sectors that could not be recovered. I should also add that since these drives had mostly been used as storage drives, not as working drives, they were nowhere close to their lifetime write limits. After refreshing the data on all of my flash drives, I was left with five that had so many sectors with errors that I decided to set them aside for later when I can find the time to look at the data to see if anything can be salvageable.

Creating an iron-clad rule for how long a drive can safely go before being refreshed is not really possible based on the dozens of drives I refreshed. The drives were different ages, from different manufacturers, with different numbers of write cycles. The files were also of different types. While the loss of a single bit of a file containing code could be catastrophic, the same loss to an audio or video file should very likely have no noticeable affect. Therefore, I simply cannot generalize by saying that four, five, or six years is good enough to prevent data loss. The charge-draining phenomenon is also random, because electrons leak out randomly. So, saying anything definitive about the probability of losing a file of a given age is impossible. All I can say is that I found that for little used drives with files written four years ago or less, essentially all data seemed to have been recoverable, while a significant number of files written longer than four years ago had at least some unrecoverable data in them.

Since DiskFresh only gives sector numbers containing unrecoverable data, I was not able to identify the ages of specific files that were unrecoverable. I could only get a sense of the age ranges on specific flash drives that suffered from data loss. Because I could not identify individual files with bad data, I could not delete them and perhaps replace them with other good copies from other backups. This means that the only two options are keeping the data on the flash drive or wiping it all off completely. I wish someone would solve this problem by writing some disk refreshing software that identifies bad files, instead of bad sectors. I considered Microsoft's chkdsk command, and I could be wrong, but I it seems to only identify bad sectors and master file errors, not the names of bad file.

Some supporting evidence for my theory that older flash drives hold their data longer came from a flash drive that I had acquired in 2007. The data on this 2 GB Kingston flash drive was all written in March of 2012. Windows 7 could not even boot Kingston's DataTraveler software with which the data had been stored on the drive. I had to use Windows XP on my 2006 Dell D610 laptop, which was also able to run DiskFresh. Miraculously, 100% of this thirteen-year-old data was readable and refreshable.

Something I should add is that with when DiskFresh refreshes files, their dates are not reset, so you will have to keep track somehow of when you refresh your drives. I think my solution to this problem will be to refresh all my drives once a year.

I noticed that an encrypted drive with some bad sectors on it was completely unrecoverable. After refreshing this drive, I was able to use TrueCrypt to open the encrypted volume, but when I tried to open the individual files, not one of them could be opened. As far as I could tell, the data was a total loss. All I could do was wipe the entire drive and use it for something else. This suggests that if you plan to store data on a drive for a long period of time, you should seriously consider not encrypting it.

When DiskFresh recognizes a file format on a drive, it offers a choice between refreshing the formatted drive and refreshing the raw drive. I noticed that when I picked the formatted drive option, it refreshed only the files and none of the blank space on the drive. When I chose to refresh the raw drive, it refreshed the entire drive. So, refreshing the formatted drive when you only have a few files on it should save time and wear on the drive. My assumption was that refreshing a formatted drive also refreshes its partition table, but I could not see a way of verifying that. Refreshing a drive should make it faster to read afterwards, but I did not notice this happening on drives that had hundreds of errored sectors. I can't explain that, and it makes me wonder if maybe I understand the refreshing process as well as I think. Unfortunately, the developer of DiskFresh provides almost zero documentation.

Although evidence shows the gradual loss of electrons from memory elements in flash drives can be reduced by lowering their temperature, this also causes other problems. So, you may want to read more on the subject before deciding to store your backup flash drives in a Ziploc bag in your refrigerator or freezer.

I have now mentioned the most significant insights I gained from spending four days last week refreshing all of my SD cards and USB flash drives. I hope these insights will be of some use to readers who are backing their data up onto flash-based storage devices. The bottom line here seems to be that if you want the data you have stored on these types of drives to be good for many years, you need to use drives that have not been written too much and refresh them at least every four years, but every year or two would be even better.

If you have found this article worthwhile, please share it on your favorite social media. You will find sharing links at the top of the page.

Related Articles:

The Importance of making Regular Backups

The Problem of Long-Term Offline Data Storage for Consumers

Understanding Computer File Transfer Rates

Developing Computer Self-Reliance