We can rule out one right away: there is no FAT64. 
FAT32 is the default formatting used on most thumb drives. Let’s review some of the alternatives, and consider when you might want to switch.
Become a Patron of Ask Leo! and go ad-free!
File system formats like FAT32, NTFS, and exFAT are standard ways of organizing data on hard disks. Each has strengths and weaknesses, and which is most appropriate depends on how the drive will be used. For USB thumb drives, FAT32 is typically the default, and is the most compatible for drives moved between machines with different operating systems.
File systems
File systems are how the bits and bytes of your data are organized on a disk. The type of file system is selected and initialized when you format a drive.
Let’s use a reference book as an analogy. As with most analogies, this relies on gross oversimplification and is not intended to describe the details of how disks work. Rather, it’s to understand the differences between how file systems organize data.
Regardless of the information in a reference book, it can be laid out and augmented in many different ways. Some books include only a table of contents. Some have an index at the back. Some include footnotes at the bottom of the page, others put them at the end, and some have no footnotes at all.
Similarly, some might use large text and wide margins, resulting in less information per page, while others cram a lot of words onto each. Some books are in color and have pictures; others do not.
The same information can be published in a variety of ways. All have the data, but with different characteristics. The same is true with file systems: they all store data, but organize it differently.
Different file systems
The three most common file systems you’re likely to encounter1 are FAT32, NTFS, and exFAT.
FAT32
FAT32 (File Allocation Table, 32 bit) is the oldest and perhaps simplest of the three. Think of it as a reference book with only a table of contents.
Because it’s been around longer (it dates back to MS-DOS and Windows 95), it’s by far the most compatible file system. All versions of Windows support it, and so do almost all other operating systems and devices that use disks. It’s well suited to transferring data between different devices, such as your camera, your phone, or your Mac, PC, and Linux computers.
Individual files on a FAT32 formatted drive are limited to 4GB in size. This limitation is becoming a more common reason that other formats might be used.
NTFS
NTFS (New Technology File System) was introduced with Windows NT, and is the default file system used by Windows. The Windows system drive (typically “C:”) must be formatted with NTFS. Think of it as a reference book with all the bells and whistles: table of contents, index, bibliography, footnotes, annotations, and more.
NTFS is optimized for fast, reliable, and perhaps most importantly, multi-user access. While FAT32 has no concept of “permissions” (other than generic “read only”, “hidden”, and “system” attributes), NTFS supports a robust, complex, and complete set of access controls at the user account and group level that can allow or deny read and write access separately, as well as access to many other characteristics.
NTFS carries much more overhead information as compared to FAT32.
Because of its performance and reliability, NTFS is the choice for hard disks and removable drives accessed only by Windows systems. Unfortunately, support for NTFS is less complete on other operating systems. Mac and Linux only read it, though addons can be installed to write to it. NTFS support in small, portable devices is very uncommon.
exFAT
exFAT (extensible FAT) is an attempt at a compromise: a FAT-like file system that can be supported with minimal changes to existing FAT32-supporting software. Think of this as our original FAT32 reference book with an index added and printed on bigger, easier-to-read paper.
exFAT is optimized specifically for flash-based disks, and is becoming the standard default file system format for memory cards over 32GB. It also supports individual file sizes well in excess of the 4GB FAT32 limitation.
You might think, then, that exFAT would be a great choice for a USB thumb drive or memory card — and it is, except if you need to use that thumb drive on other devices as well. While exFAT compatibility is better than NTFS, it’s still not uncommon to run across devices like older cameras that don’t support it. Current versions of major operating systems including Windows, Mac OS and most Linux distributions support it.
So, what file system should you use?
In your case, for storing a password vault, I’d leave it as FAT32. There’s no compelling reason to change. In fact, unless you have a compelling reason to change the default format, I’d leave the format of any USB stick or flash-based memory card exactly as it came.
“Compelling reasons” include the additional performance that might be gained from an NTFS or exFAT file system, or the ability to store files larger than 4GB. In either of these cases, I’d probably use exFAT on flash-memory devices.
What happened to FAT64?
The concept of FAT64 never existed.
While we think of the operating system as being 32-bit or 64-bit, that’s a characteristic of the CPU inside your computer, which operates on data either 32 or 64 bits at a time.
When it comes to file systems, the “32” in FAT32 refers to the size of some information stored about the data on the drive. It’s not dependent in any way on the CPU.
Limitations like the 4GB file size are a result of that “32” choice. I suppose one way to remove that limit might have been to move to a 64-based format, but it’s very likely that the structural differences would have been too great and would seriously impact storage efficiency and size as well as performance.
Alternatives like NTFS and exFAT were developed instead.
Do this
Subscribe to Confident Computing! Less frustration and more confidence, solutions, answers, and tips in your inbox every week.
I'll see you there!
Podcast audio
Related Video
Footnotes & References
1: At least in the world of Windows.
Leo, can you address partition size limitations on FAT32 and solutions for larger drives? My brother needed a 256 GB SD card formatted as FAT32 for a Tesla camera application . Windows would not go over a certain size (don’t remember the limit). It would not format a 256 GB card. Had to find a third party utility that would do the job. Funny thing is that this same 256 GB SD card worked fine in a windows machine after being formatted by the third party utility. Hmmm…
I’d have to see the error thrown by Windows when trying to format it. 256GB is well within what FAT32 supports in terms of a single volume size.
Many, many years ago, I owned one of the very first DOS computers. It came with two 320K (or was it 360K?) 5.25 inch floppy drives and no hard drive. I purchased a 20 MB hard drive and installed it. I had to partition that into two drives because 20 MB was too large for that version of DOS (memory’s vague but either 1.0 or 2.0) to support.
Yep. 360k floppies. I also remember getting overly excited when I saw a memory test go beyond 640k on the first PC-AT I encountered.
Hey, I used (and, occasionally, I STILL use!) a C64 computer whose floppy disks held at most 168,656 bytes of data!!! :)
I couldn’t resist:
Started with punch cards, but sadly have little call for them these days.
I used those at Texas Instruments. My supervisor wanted us to write our program on paper and have the keypunch operators type the Hollerith cards. I rebelled and did everything at the terminal (after all, that was 1980) but I still had to use cards for the Job Control Language.
Our mainframe also had a card punch. So I suppose one solution would have been to do everything at a terminal, and then send the program to the card punch to make someone happy.
I usually did that
You guys are newbies! When I started on the IBM 704 in 1956, the only way to load and run programs was from “binary cards”. IIRC, they held 22 36-bit instructions each, two each in the first 72 columns of the first 11 rows on a card. The 12th row of the card was reserved for information needed to load the instructions into the proper memory locations. The last eight columns were used for sequence numbering so that the cards could be kept in order.
The cards were punched out by an assembler program via a 721 Card Punch and later read by a 711 Card Reader. A program error would normally force correction of the source program and a reassembly which resulted in an entire new deck of cards. However, sometimes a simple error could be patched by manually punching a binary card (via a 010 Card Punch, which punched a hole at a time) with corrected instructions (in binary) which would be placed at the end of the deck and loaded over the erroneous instructions. Manually punching a binary card was obviously very error-prone, and an error would usually force you to repunch the whole card. But if one or two holes were punched where they shouldn’t have been, you could avoid that by inserting two of the punched-out chips (chad–remember “hanging chad” in the 2000 election?) back into the holes and putting scotch tape over them to hold them in.
Well, you remember what the standard instruction for that was, don’t you?
“Do not fold, spindle, or mutilate.”
(I never did know what the word “spindle” meant in this context.)
And, a long way back, when our town stiil had a Montgomery Wards catalog store (!), the employees there allowed this curious kid (me) to create a strip of Teletype tape (!!). :)
Spindle: don’t wrap the card around your pencil and expect it to work later.
Used punch cards at school and first year of university (1978). Thereafter my shoe box of punch cards was a great source of stationery for speaker notes. Lasted many years and there are wedding photos of me holding a bunch of them.
BTW, regarding that photo — what, exactly, IS the “Fortran Statement” represented by those punched holes, and what does that statement do? (Just curious; I know no Fortran. My computer language is BASIC.). :o
It doesn’t appear to be a statement card. It might be a data card.
My second computer in college was a PDP8 (the 8 stood for 8 kb) machine from DEC. We used a punch tape reader. Debugging was hell. Our department couldn’t afford a punch card reader.
Not really. The 8 stood for 1968, which is when that machine series debuted.
The first PDP 8 came out in 1965.
I remember starting with 7″ floppies. I don’t remember the capacity but I do remember each successively smaller diskette had a higher capacity than the previous larger diskette. I thought the 3.5″ diskettes were the greatest. Built-in dust protection, a write protect switch tab, and they fit in my shirt pocket. Our HDDs looked like washing machines and had one fixed and one removable platter each holding 5 MB. Now I have a million megabyte (1 TB) Micro SD card less than half the size of my pinky fingernail. Holding almost the equivalent of a million floppies.
You guys are making me old. I started in IT in the 60’s and yes, used 80 col cards, 7″, 5.25″, and 3.5″ floppys. Began on an IBM360. At on time used a Burroughs computer with 9600 bytes of memory to run Shop Floor Control and Inventory system in a mfg plant. No disk, 3 reel-to-reel tape drives–try doing a sort on that–watch tapes spin for hours!! Have worked on most everyting since, Retired 5 years ago…
You sound like you’re around my age I first used Hollerith cards in college in 67 writing programs in Fortran.
I thought they were 8″ floppy disks, not 7″. ???
One comment on formatting USB sticks / thumb drives: if you put audio or video files on them and want it to play in your car or via a TV’s USB port, you’ll want to format as FAT32. Like Leo mentioned, unless I there’s a need to format differently, I always use FAT32 for thumb drives.
You guys are making me feel old. I started with an abacus with five strings and ten beads on each. Colour coded for powers of ten. You think you have problems when a hard disk crashes. Try crawling around the floor looking for beads when an abacus string breaks.
When I was in Ukraine 20 years ago, every seller in the market used an abacus (Russkie Compyooter). They calculated the cost on the abacus held it up to show you how much to pay. Needless to say, I was clueless.
I am able to both read and write to NTFS using Linux Mint and Ubuntu. I use it with no problems on flash drives.
All modern OSes read (not necessarily write) NTFS formatted drives. The problem is also with other devices, such as, cameras, dedicated audio and video players, and such.
Thanks hugely, Leo — wonderfully explained! I’ve been wanting to get my head around this for a while — this article paints it out perfectly!
Warm regards,
Dorian
Leo, You guys are really making me feel old. I remember in 1955 the company where my dad worked had to build a whole new floor on the building to accommodate an IBM360 computer. I went and seen the computer and stood in awe of the sheer size of this monster and all the punch cards flying through it. They were just starting out using it for billing I believe. My father was an accountant and he said that machine will eventually take his job. He was right because he retired early because of computers.
In 1979, when I retired from the USAF, I ended up in a classroom with 36 Tandy Computers with 3.5 inch floppies.
Not to be critical, Glenn…. In 1962, I was programming in MAP (36 bits plus a parity bit) and binary on IBM 7090 Scientific machines (full 32K and a dozen tall tape drives). We were forbidden to stand in front of the magnetic tape drives if we were wearing white shirts since someone thought that reflected light would cause read/write issues. One of the Navy 7090s was water-cooled! We also had a little IBM 1401 character machine. The 1401 was so popular that, at the time, there were more 1401s on order than computers existing in the entire world. We also had a CDC 6600 and a USQ-20 as mainframes. When I left that job in 1964, the IBM 360 appeared.
There was a highly popular desk ornament at the time. I still have one. Most said THINK. Some said DENKE or some other translation. THIMK was wildly popular proving that programmers had a sense of humor. If I can’t attach anything, look up ‘IBM THINK sign.’
CDC 6600 was my first computer in college programmed with Hollerith cards.
Same. Fortran first, then Compass … the assembly language for that beast.
I use ntfs as it seems faster and handles over 4gig avi files .
You all are making me feel old and incompetent. In Vietnam, 1971, I was a wiz at operating file cabinets and manual typewriters which was much better than humping the bush, but then I was suddenly assigned to operate a computer in a conex box thirty degrees colder than outside, working with cumbersome punch card trays, with an impatient sargeant looking over my shoulders who assumed that I knew what I was doing (it involved processing officer records). It was my intro into computers, almost literally, and I couldn’t wait to get out. Maybe someone knows what I would have been operating because I can’t remember.
One of my summer jobs in college was to do the wiring on Conex boxes. I got proficient with a soldering iron.
Thank you everyone for making me feel young. I started with what I called an expensive typewriter in the 80’s with DOS. It is nice to be the youngster on the block for a change
Reply to TheGrandRascal
(I never did know what the word “spindle” meant in this context.)
The office spindle was a pointed metal rod 4 to 6 inches (100 to 150 mm) high.
So I always assumed that the verb “spindle” meant to impale the object on that rod.
If you want to see a photo, use the title below for a google search.
1930s 1940s 1950s RETRO OFFICE SHARP SPINDLE SPIKE WITH PAPER MESSAGES BILLS IMPALED STUCK ON IT ANTIQUE ORGANIZATION SYSTEM – Image ID: AAMTEY
That explanation makes sense. I couldn’t imagine enough people spindling a card around a pencil to merit that warning. I remember going into shops where the sales personnel would spindle the store copy of the receipts. I never knew it was called a spindle till I saw your post.
Here’s a link:
https://www.alamy.com/stock-photo-1930s-1940s-1950s-retro-office-sharp-spindle-spike-with-paper-messages-12665138.html
I thought that was a spike. Learned something today. :-)
Well there goes my promise to myself to go to bed early for a change.
Thanks Leo for your clear and detailed explanation. Thanks everyone else for an interesting and humorous discussion on computing history (if you see this – I just noticed this thread is 3 years old).
I Just wanted to wipe some used thumb drives clean by reformatting (They were from a lady de-cluttering prior to moving into aged care and had personal stuff on them). I wasn’t sure if I should accept the default Fat32 option and I found my way here to geriatrics anonymous (or unanimous?). A suggestion considering the age of most respondents – if anyone has a go at me about my age I just say “you’re just jealous ‘cos I got to see the dinosaurs before they died”
I too had a (very small) part in early computing. As an apprentice mechanical draughtsman ( a failed experiment instigated in the 60’s by desperate employers faced with skilled labour shortages), I was farmed out to customise artwork for the printing of data cards ( the weird and boring jobs – that’s what apprentices are for, right?). Using ancient technology, a draughting pen and lettering stencils, I added company names and other wording (like the “Fortran statement” and “mass werk data center” on Leo’s example) to 3x enlarged dummy cards which were then photo-reduced to normal size and printed. I can still recall the name of the company, Data Card Australia (Duh, that was easy to recall)
Like Glen Cowgill’s Dad, I too lost my career to computers but by a different path. In the early 80’s when Legionnaires Disease and Asbestos suddenly became everyone’s priority, I was asked by my employer to swap my drawing pens for water testing kits and an asbestos surveying toolkit. Unfortunately this coincided with the introduction of CAD drafting, so when the asbestos work dried up several years later my lack of CAD skills resulted in redundancy.
Oh, and I think I still own a spindle, used at home as a BC (before computer) bills and receipts “folder”.