RTPVid wrote in post #16561532
I read somewhere that your typical USB plug-in flash memory (i.e. "flash drive", "thumb drive", etc.) is made from cheaper components that an SSD. This makes sense, since an SSD must perform nearly 100% reliably or no one would buy them, whereas a corrupted flash drive will merely be tossed in the trash (with a little grumbling about having to re-copy the files... most people use flash drives for transport, not for "one and only copy" storage... well, at least that is how I use them). Long way of saying, I would not use a flash drive for backup storage at all.
There are no-name manufacturers who are willing to use non-tested memory in their products. But there are no-name manufacturers of trash of just about any type.
The cheaper factor involved in flash thumb drives is that they:
1) Often use very slow flash chips, which means that the write speed is very, very slow. The USB2 interface might manage up to about 40 MB/s for bulk transfers, while some USB thumb drives can only accept 1-2 MB/s. The slow flash cheaps aren't cheaper because they are bad, but because they are too slow to be used in lots of applications so the manfucaturers who makes them have to sell them at a lower price point just to get someone to buy the chips. And a large percent of USB thumb drive buyers never thinks about the write speed, so USB thumb drives are an excellent way to find customers for these slow flash chips.
2) Some USB thumb drives uses lousy memory controllers with bad wear leveling implementation. But archival storage doesn't have much use for extensive wear leveling since the thumb drive gets filled more and more without lots of erase cycles - it's the directory structure and the file allocation structure that has a number of rewrites. There might - just might - be USB thumb drives that doesn't verify each write to make sure the sector was usable, but that isn't an issue if getting a named drive. But such bad USB thumb drives aren't found under any known label unless buying faked USB thumb drives.
In the end - if you get a named USB thumb drive, you will get quite similar reliability as you get from an SSD drive. Just with a slower transfer speed.
It is only very expensive SSD drives that uses SLC-type flash memory, where each memory cell stores a single bit. The normal SSD uses same type of MLC flash memory as normal USB thumb drives, or normal memory cards. So each memory cell stores two - or maybe even three - bits of data. SLC can handle ten times as many rewrites before wearing out, since it is much simpler to detect charge/non-charge in a memory cell compared to measuring the amount of charge to decide if the cell represented 00/01/10/11 or maybe even 000/001/010/011/100/101/110/111. But look at the prices of SSD specifying SLC memory, and you would realize that they aren't the big sellers in the computer shop.
Another thing - normal thumb drives and SSD uses normal flash surface for wear leveling storage. Really good SSD uses F-RAM memory for the wear-leveling information, i.e. a non-volatile memory type that doesn't wear out. But that is also irrelevant for archival use - there just aren't enough rewrites in an archival media to make a difference.
So in the end: A USB thumb drive represents a good media for use as "one archival copy". The size also makes them excellent in situations where physical size is important - like when storing a backup at the bank. One problem right now is that the move from USB2 to USB3 means lots of USB thumb drives has prices matching an SSD - because to get the high transfer speeds, they need to compete for the same fast flash chips as used in the SSD. And a large USB thumb drive requires high-density flash chips, i.e. current-generation chips.
It's common to guarantee 10 years retention time from flash media. But that is often at maybe+40 or +60C. Every 10°C temperature higher halves the rentention time and every 10°C lower temperature doubles the retention time. Anyway - it's good to move the data to newer media every 5 years. Newer media has lower volume costs, refreshes the data and takes advantage of technological improvements (like newer interfaces, faster transfer speeds, ...)
But whatever media is used: there should obviously still exist multiple copies of important files. And it doesn't hurt if multiple media types are used.