On 21 Oct 99, at 16:34, Gregg Pfaff wrote:

> Could somebody explain FAT16 & Fat32 to me? I am getting ready to
> buy a new HD and then reformat my old one to see if it's usable as
> a slave. My old one is a Conner 1.6GB that is fomatted to one
> partition? FAT16 win95a. I'm intending to install win95b which I
> believe is OS2? and FAT32?

  The FAT (File Allocation Table) is a special area on the volume
(drive/partition) that consists of one "number" for each "allocation
unit" (or "cluster").  [A file consists of some whole number -- zero
or more -- of clusters; clusters consist of a whole number of
*sectors*, which on the PC are always 512 bytes.]
  The directory entry for a file contains a starting cluster number;
when tho OS looks that entry up in the FAT, it finds the number of
the next cluster (or a magic value saying there aren't any more).

  In FAT16, each FAT entry is 16 bits, allowing for a maximum of
65,535 clusters in a partition.  Under DOS and Windows 9x, the
largest cluster size is 32KB (64 sectors), so the largest size of a
partition is 2GB.  [NT includes support for 64KB clusters, so it's
possible to have a FAT16 partition up to 4GB in size, as long as NT
is all you care about.]
  But hang on a sec.  Recall that a file consists of a whole number
of clusters.  That generally means that the last cluster of a file is
only partially used; on average, half a cluster (half of the last
cluster) is wasted for each file on the partition.  This may not be
bad if you have a few big files, but it's more common to have lots of
relatively small ones.  16KB/file (half of 32KB cluster size) times
thousands of files adds up to a fair bit of disk real estate.

  FAT32 uses 32-bit cluster numbers instead of 16-bit cluster
numbers.  At 1 sector/cluster, I figure that's enough to manage a
partition up to 2TB in size -- some other limitations come into play
before we get to that kind of size.

  FAT32 may be a bit slower than FAT16 on some machines.  Although it
reduces the amount of wasted space at the end of files (reducing the
number of data sectors that need to be read or written), it increases
the number and size of FAT entries that need to be consulted, and the
larger FAT makes it less likely that this information will be held in
cache when needed.

  [Historical note:  Early versions of DOS used a 12-bit FAT
structure, and I think had a cluster size limit of 8KB.]

David G

                Curious about the people moderating your
                   messages? Visit our staff web site:
                     http://nospin.com/pc/staff.html