FAT32 size issues, _just_ about size?

[StargateFan]

Is the issue with FAT32 and Win9x, if I've understood the problem
correctly, simply about the size of one directory, or is it also an
issue of the size of the entire hdd. My reason for concern is, other
hardware issues aside first, could we run Win9x on a large 200gig
computer if it were partitioned into the 32gig (or whatever max size
actually is) pieces? Or is the 200gig an issue right off the bat?
Thanks.

 

[Jeff Richards]

The size of one directory has never been a significant issue. Perhaps you
meant to say the size of a partition.

You can run W98 on a machine with a 200Gb disk if you limit the size of the
partition to about 137Gb. You could use the whole disk by creating multiple
partitions provided the machine supports 48-bit LBA and the appropriate
drivers are used. However, some of the standard system tools will not
work properly.

See, for instance:
http://www.48bitlba.com/win98.htm
--
Jeff Richards
MS MVP (Windows - Shell/User)
"StargateFan" <IDon'tAcceptSpam@IDon'tAcceptSpam.com> wrote in message
news:ukvdr3tlkmg90aqr0hvmgtfvu0pijc6ujh@4ax.com...
> Is the issue with FAT32 and Win9x, if I've understood the problem
> correctly, simply about the size of one directory, or is it also an
> issue of the size of the entire hdd. My reason for concern is, other
> hardware issues aside first, could we run Win9x on a large 200gig
> computer if it were partitioned into the 32gig (or whatever max size
> actually is) pieces? Or is the 200gig an issue right off the bat?
> Thanks.
>

 

[philo]

"Jeff Richards" <JRichards@msn.com.au> wrote in message
news:%23Yxm%23rNcIHA.4880@TK2MSFTNGP03.phx.gbl...
> The size of one directory has never been a significant issue. Perhaps you
> meant to say the size of a partition.
>
> You can run W98 on a machine with a 200Gb disk if you limit the size of
the
> partition to about 137Gb. You could use the whole disk by creating
multiple
> partitions provided the machine supports 48-bit LBA and the appropriate
> drivers are used. However, some of the standard system tools will not
> work properly.
>
> See, for instance:
> http://www.48bitlba.com/win98.htm



That's completely correct of course...
I just wanted to add that Fat32 has a file size limitation of 4 gigs...
so if the drive would be used for storing large files, such as videos...
they could not exceed 4 gigs.

Additionally, a fat32 drive over 32 gigs would have a fair amount of wasted
space
due to large cluster size.

 

[Jeff Richards]

"philo" <philo@privacy.net> wrote in message
news:%23LjoOXOcIHA.4696@TK2MSFTNGP05.phx.gbl...
>
> "Jeff Richards" <JRichards@msn.com.au> wrote in message
> news:%23Yxm%23rNcIHA.4880@TK2MSFTNGP03.phx.gbl...
>snip <
> I just wanted to add that Fat32 has a file size limitation of 4 gigs...
> so if the drive would be used for storing large files, such as videos...
> they could not exceed 4 gigs.
>

and many W98 applications don't work with files larger than 2Gb.

> Additionally, a fat32 drive over 32 gigs would have a fair amount of
> wasted
> space
> due to large cluster size.

and will perform quite poorly for file changes due to the large size of the
FAT.

 

[98 Guy]

StargateFan wrote:

> Is the issue with FAT32 and Win9x, if I've understood the problem
> correctly, simply about the size of one directory, or is it also
> an issue of the size of the entire hdd.

The 32-bit protected-mode driver (esdi_506.pdr) that windows-98 uses
has a flaw, or design oversight, that prevents it from correctly
accessing hard drives larger than 128 gb (or 137 giga-bytes, depending
on how you define a giga-byte).

This issue has been discussed at length several times within at least
the past year here in this newsgroup, and if you really want to
understand the issue then you should use the google-groups usenet
search interface and look for those past threads. Look for any thread
containing "edsi_506.pdr".

> My reason for concern is, other hardware issues aside first,
> could we run Win9x on a large 200gig computer if it were
> partitioned into the 32gig (or whatever max size actually
> is) pieces? Or is the 200gig an issue right off the bat?
> Thanks.

If the drive in question is an ordinary IDE (aka PATA) drive then most
likely it will be accessed via ESDI_506.PDR and it will suffer data
corruption when a read or write is performed across the 128 gb point
on the drive. It does not matter how the drive is partitioned. If a
200 gig drive is partitioned such that nothing beyond the first 127 gb
are allocated to any partitions or logical drives then in theory there
will never be any drive access beyond the 128 gb point and you should
not encounter any problems.

There are remedies that will allow the use of hard drives larger than
128 gb for windows 98:

1) obtain an alternate (modified) version of ESDI_506.PDR.
2) obtain a PCI-based hard drive controller card that comes with it's
own replacement for ESDI_506.PDR (this can be either a PATA or SATA
controller card).
3) Obtain something called the "Intel Application Accelerator" (this
only applies for system motherboards with certain Intel-based
chipsets).
4) The use of SATA hard drives (if so supported by the motherboard)
will allow the use of large hard drives (larger than 128 gb) as long
as certain settings in the system bios are used.

 

[StargateFan]

On Sat, 16 Feb 2008 10:24:35 -0500, StargateFan
<IDon'tAcceptSpam@IDon'tAcceptSpam.com> wrote:

>Is the issue with FAT32 and Win9x, if I've understood the problem
>correctly, simply about the size of one directory, or is it also an
>issue of the size of the entire hdd. My reason for concern is, other
>hardware issues aside first, could we run Win9x on a large 200gig
>computer if it were partitioned into the 32gig (or whatever max size
>actually is) pieces? Or is the 200gig an issue right off the bat?
>Thanks.

Thanks for everyone's replies. Yes, I meant "partitions" rather than
"directories". I do that all the time, use the wrong word, sorry.

I'll look at the google group archives. Thanks for the specifications
if the PDR issue. I don't know any of the terms involved, so it will
be good to learn.

So the issue with FAT32 generally is one of the size of the entire HDD
and not just of individual partitions. A colleague quote 30+ gigs yet
here 4 gigs was used as maximum. Interesting discrepancy, but thanks
for setting me straight (perhaps he, too, was referring to overall HDD
drive and _not_ individual partitions).

Thanks for potential workarounds, too. Will look into those as well.
Again, it's all greek to me so I have a lot to research and learn.

Cheers.

 

[Bill in Co.]

StargateFan wrote:
> On Sat, 16 Feb 2008 10:24:35 -0500, StargateFan
> <IDon'tAcceptSpam@IDon'tAcceptSpam.com> wrote:
>
>> Is the issue with FAT32 and Win9x, if I've understood the problem
>> correctly, simply about the size of one directory, or is it also an
>> issue of the size of the entire hdd. My reason for concern is, other
>> hardware issues aside first, could we run Win9x on a large 200gig
>> computer if it were partitioned into the 32gig (or whatever max size
>> actually is) pieces? Or is the 200gig an issue right off the bat?
>> Thanks.
>
> Thanks for everyone's replies. Yes, I meant "partitions" rather than
> "directories". I do that all the time, use the wrong word, sorry.
>
> I'll look at the google group archives. Thanks for the specifications
> if the PDR issue. I don't know any of the terms involved, so it will
> be good to learn.
>
> So the issue with FAT32 generally is one of the size of the entire HDD
> and not just of individual partitions. A colleague quote 30+ gigs yet
> here 4 gigs was used as maximum.

The largest FILE (i.e. filesize) you can create using FAT32 is 4 GB.
File <> Directory <> Partition. They're all different.

> Interesting discrepancy, but thanks
> for setting me straight (perhaps he, too, was referring to overall HDD
> drive and _not_ individual partitions).
>
> Thanks for potential workarounds, too. Will look into those as well.
> Again, it's all greek to me so I have a lot to research and learn.
>
> Cheers.

 

[Lil' Dave]

"StargateFan" <IDon'tAcceptSpam@IDon'tAcceptSpam.com> wrote in message
news:075fr354jkisl2rff6buma0fb5esserdac@4ax.com...
> On Sat, 16 Feb 2008 10:24:35 -0500, StargateFan
> <IDon'tAcceptSpam@IDon'tAcceptSpam.com> wrote:
>
>>Is the issue with FAT32 and Win9x, if I've understood the problem
>>correctly, simply about the size of one directory, or is it also an
>>issue of the size of the entire hdd. My reason for concern is, other
>>hardware issues aside first, could we run Win9x on a large 200gig
>>computer if it were partitioned into the 32gig (or whatever max size
>>actually is) pieces? Or is the 200gig an issue right off the bat?
>>Thanks.
>
> Thanks for everyone's replies. Yes, I meant "partitions" rather than
> "directories". I do that all the time, use the wrong word, sorry.
>
> I'll look at the google group archives. Thanks for the specifications
> if the PDR issue. I don't know any of the terms involved, so it will
> be good to learn.
>
> So the issue with FAT32 generally is one of the size of the entire HDD
> and not just of individual partitions. A colleague quote 30+ gigs yet
> here 4 gigs was used as maximum. Interesting discrepancy, but thanks
> for setting me straight (perhaps he, too, was referring to overall HDD
> drive and _not_ individual partitions).
>
> Thanks for potential workarounds, too. Will look into those as well.
> Again, it's all greek to me so I have a lot to research and learn.
>
> Cheers.

In my experience, the actual issue I ran into is total data in terms of
files exceeding 128GB on a given hard drive (ide) with no ide driver mods in
98SE. Didn't matter if had multiple FAT32 partitions, FAT32 partition(s)
not and NTFS partition(s). When the total file data exceeded 128GB amongst
a single or multiple FAT32 partitions combined, file corruption followed....
And, yes. Total file data size on the NTFS partitions is also a factor in
the equation. (Even though 98/98SE can't "see" NTFS).
Dave

 

[Bob Harris]

While you might seem to get 98 to work on this drive, sooner or later
something will not work as you expect, and then you may lose some valuable
data.

If you really want to use a very large hard drive, switch to XP or LINUX.
If XP, you might also want to choose NTFS instead of FAT32. NTFS permits
single files larger than 4 Gig.

As for WHY you want to upgrade to 200 Gig, that is clearly not just to run
98. I therefore assume that it is for storage of many and/or large personal
files. An alternative would be to leave 98 and installed programs on a
smaller drive, but move personal files to one or more USB drives, each under
127 Gig. Or, if a desktop, you might be able to mount up to 4 internal hard
drives, again, each under 127 Gig.

"StargateFan" <IDon'tAcceptSpam@IDon'tAcceptSpam.com> wrote in message
news:ukvdr3tlkmg90aqr0hvmgtfvu0pijc6ujh@4ax.com...
> Is the issue with FAT32 and Win9x, if I've understood the problem
> correctly, simply about the size of one directory, or is it also an
> issue of the size of the entire hdd. My reason for concern is, other
> hardware issues aside first, could we run Win9x on a large 200gig
> computer if it were partitioned into the 32gig (or whatever max size
> actually is) pieces? Or is the 200gig an issue right off the bat?
> Thanks.
>

 

[Bill in Co.]

And once again, was the problem with the entire hard disk size being 200 GB?
(I know that 128 GB is the critical point)

That is, unless the BIOS supports the 48 bit LBA thing, he wouldn't even be
able to put the HD into the system and try to repartition it to a smaller
size (less than 128 GB) in the first place?

So that it's the TOTAL HD capacity that counts here, and not just the
partition size (which could be adjusted down later IF it could be installed
in the first place)??


Bob Harris wrote:
> While you might seem to get 98 to work on this drive, sooner or later
> something will not work as you expect, and then you may lose some valuable
> data.
>
> If you really want to use a very large hard drive, switch to XP or LINUX.
> If XP, you might also want to choose NTFS instead of FAT32. NTFS permits
> single files larger than 4 Gig.
>
> As for WHY you want to upgrade to 200 Gig, that is clearly not just to run
> 98. I therefore assume that it is for storage of many and/or large
> personal
> files. An alternative would be to leave 98 and installed programs on a
> smaller drive, but move personal files to one or more USB drives, each
> under
> 127 Gig. Or, if a desktop, you might be able to mount up to 4 internal
> hard
> drives, again, each under 127 Gig.
>
> "StargateFan" <IDon'tAcceptSpam@IDon'tAcceptSpam.com> wrote in message
> news:ukvdr3tlkmg90aqr0hvmgtfvu0pijc6ujh@4ax.com...
>> Is the issue with FAT32 and Win9x, if I've understood the problem
>> correctly, simply about the size of one directory, or is it also an
>> issue of the size of the entire hdd. My reason for concern is, other
>> hardware issues aside first, could we run Win9x on a large 200gig
>> computer if it were partitioned into the 32gig (or whatever max size
>> actually is) pieces? Or is the 200gig an issue right off the bat?
>> Thanks.

 

[98 Guy]

"Bill in Co." wrote:

> And once again, was the problem with the entire hard disk size
> being 200 GB?

Did you not see my post from yesterday?

 

[Tim Slattery]

"Jeff Richards" <JRichards@msn.com.au> wrote:

>The size of one directory has never been a significant issue. Perhaps you
>meant to say the size of a partition.

Not the amount of space that the files in the directory occupy, but
there is a limit on the number of entries in a directory: 65,536. File
or subdirectory names that don't fit the DOS 8.3 specification take
from 2 to 13 entries, so your directory can fill long before you
expect it to.

--
Tim Slattery
MS MVP(Shell/User)
Slattery_T@bls.gov
http://members.cox.net/slatteryt

 

[jason.mangiafico@verizon.net]

quoting:
> > Additionally, a fat32 drive over 32 gigs would have a fair amount of
> > wasted space due to largecluster size.

The default cluster size, that is. Any good third party partitioning
utility will let you adjust the cluster size of FAT volumes. For
example, a 32gb drive with 4kb clusters is well within the limitations
of fat32.


> and will perform quite poorly for file changes due to the large size of the
> FAT.

Not a real issue since either ntfs or fat32 with have roughly the same
file system overhead for the same cluster size.

 

[98 Guy]

Jeff Richards wrote:

> > Additionally, a fat32 drive over 32 gigs would have a fair
> > amount of wasted space due to large cluster size.
>
> and will perform quite poorly for file changes due to the large
> size of the FAT.

I've posted here about a year ago on the topic of using small cluster
size for win-98 on large hard drives.

Basically, I've run win-98 on 160gb and 500 gb drives (formatted as a
single partition using 4 kb cluster size).

There is _no_ performance hit because of the huge increase in the size
of the FAT. The OS does not load in the entire fat into memory during
normal operation either - it might only do that during drive
maintainence operations like scandisk and defrag. There is

And that's where the use of small cluster size runs into problems for
win-98, and there are some remedies (like using Win-me versions of
scandisk and defrag). I also found that win-98 doesn't like putting
the swap file on a 500-gb hard drive formatted as a single partition
with 4kb clusters.

-------------------------

Newsgroups: microsoft.public.win98.gen_discussion
From: 98 Guy <98@Guy.com>
Date: Mon, 26 Feb 2007 10:49:46 -0500
Subject: Update 4: Cluster size and exploring the limits of FAT-32

I think this will be the last update, as I've made a rather important
discovery, which is that DOS scandisk (both the original win-98
version and the windows ME version) don't really seem to have a
cluster size limitation.

Specifically, here's what I've just tested:

drive c: 25 gb / 4kb cluster size / 6.3 million clusters
drive d: 31 gb / 4kb cluster size / 7.8 million clusters
drive e: 125 gb / 4kb cluster size / 31.2 million clusters
drive f: 156 gb / 32kb cluster size / 4.9 million clusters

drive (c) is SATA drive 1
drive (d) and (e) are SATA drive 2
drive (f) is SATA drive 3

Drive C has win-98se installed, and instead of starting 98 I instead
brought up the start menu and started in DOS. From the command
prompt, I ran scandisk (Win-98se and Win-ME).

Both of them ran fine on all of the above volumes.

Previously when I was testing dos scandisk.exe, I was booting from a
floppy, and probably wasn't loading himem.sys. Himem.sys is needed by
scandisk to run properly.

So here is the master summary of this thread:

---------------------------

1) Scandisk (DOS scandisk.exe, not Windows scandskw.exe) does not
appear to have a cluster-count limitation. Both Win-98se and Win-ME
versions of scandisk have been run on drives with up to 31 million
clusters and have executed properly with no errors. Himem.sys must be
loaded for scandisk to function properly. Microsoft states that
FAT-32 volumes are limited to 4.17 million clusters because of
scandisk.exe, and that scandisk.exe is limited to a memory or data
array size of 16 million bytes. It could very well be that this 16 mb
limit is based on Microsoft's stated minimum system requirements for
Windows 98 (which is 16 mb of system RAM) and that scandisk will
automatically make use of all available system memory if required.

----------------------------

2) Win-98se has been installed directly on 160 gb volumes formatted
with 4kb cluster size (resulting in 40 million clusters) and has not
shown any instability. This was performed on a 160 gb SATA drive
assigned to a RAID controller (but not used in RAID mode). To test
for 137gb data corruption (which theoretically takes place when a read
or write across the 137 gb boundary occurrs) a series of 1 gb VOB
files were copied repeatedly in order to fill the drive. The drive
was eventually filled with 150 of these 1 gb files, and no drive
corruption occurred.

----------------------------

3) The only drawback that I've seen when running a volume with a large
cluster count is that DOS will take a much longer time to perform the
first DIR command. This might also happen in Windows as well - I may
have seen this effect but I haven't specifically looked for it. The
issue is the computation and display of free remaining drive space,
which is part of the DIR command and also happens when browsing the
drives in windows. Related to this is the question does windows store
the amount of remaining drive space somewhere on the drive (instead of
requiring it to be re-calculated every time it's needed).

---------------------------

4) Standard DOS tools like fdisk and format can be used to partition
and format hard drives in excess of 137 gb. Fdisk was used to
partition a 160 gb drive into a 32 gb primary and 121 gb secondary.
The updated or "fixed" version of fdisk.exe was used. What has not
been tested (yet) is the undocumented /Z:n command line parameter for
format, which allows the user to specify a particular cluster size (n
x 512 bytes). Third-party drive utilities (based on On Track's Disc
Manager) can also be used to partition and format hard drives, but I
have found those utilities to be very unstable and to lock up/crash
about 75% of the time I use them.

---------------------------

5) There is evidence that 6,291,204 clusters may represent some sort
of "magic number". A third-party drive partition tool (PartitionMagic
Pro Server 8.05) resorted to this cluster count when an existing 32 gb
partition was manually resized to 4kb cluster size. Norton Disk
Doctor and Speed Disk was found to work properly using this cluster
count, but not on a volume with a slightly larger cluster count of 7.8
million clusters (see note 7 below). This 6.3 million cluster count,
combined with 32kb cluster size, results in a volume size of 206 gb.
Perhaps this set of parameters is the reason for the 200 gb hard drive
size which emerged in early to mid 2003. A dir command is also
performed instantly with no delays in computing free space given a
volume with 6.3 million clusters.

--------------------------

6) Win-98 versions of Scandisk (scandskw.exe) and Defrag did not
function on a volume with 6.3 million clusters but seems to be limited
to the MS stated value of 4.17 clusters. However, Windows ME versions
of dskmaint.dll and defrag.exe does appear to function correctly with
Windows 98se and compatibility with volume size of up to 31.2 million
clusters has been observed. It is not know what their ultimate limit
is.

----------------------------

7) Norton Utilities is a very common third-party set of applications
and their compatibility with large hard drives with a large cluster
count may be of importance to some people. I have found that Norton
Disk Doctor and Norton Speed disk were compatible with volumes with up
to 6.3 million clusters, but not more without using the command-line
parameter /NOLBA. When using this parameter, NDD and SD worked on
volumes with 7.8 million clusters but not 31 million. The exact
cluster-count limit is therefore unknown at this point and I may
explore that in the future.


The switch /NOLBA forces NDD and SD to skip the drive configuration
check. This can also be done using a registry entry by adding a DWORD
registry value named NOLBACHECK at this location:
HKEY_LOCAL_MACHINE\Software\Symantec\Norton Utilities


When this option is set to 1, Norton Disk Doctor and Speed Disk skip
the drive configuration check.

----------------------------

8) Anyone considering adding a large hard drive (a drive larger than
137 gb) to an existing win-98 computer needs to consider certain
issues that include the drive type (IDE/PATA vs SATA) as well as how
the drive is controlled by the motherboard BIOS (mapped to IDE channel
or controlled by RAID controller). The main issue here is that you DO
NOT WANT the win-98se 32-bit driver (ESDI_506.PDR) to be used to
access a hard drive larger than 137 gb. Many or most motherboards
made for the past 3 years will have built-in SATA ports. Windows-98
users are advised to obtain SATA drives instead of the older
conventional IDE drives when adding a new drive (larger than 137 gb)
to a system or if building a new system.

----------------------------

From July 2007 posts (regarding running win-98 on a 500 gb hard drive
partitioned as a single volume using 4kb cluster size):

Installed win-98 on a 500 gb volume (with 121 million 4kb clusters)
and filled up the drive to about 90% with many 1gb .VOB files.
Windows will not create or place the swap file on it, hence virtual
memory will not be enabled. It will create / place the swap file on a
secondary drive, even if that drive is another 500 gb drive (an
otherwise empty drive but formatted with 32kb cluster size resulting
in 15 million clusters).

Those seeing win-98 report a negative number for the amount of free
space on the hard drive should look at this:

http://support.microsoft.com/kb/272620

Even with 1 gb of installed memory, I got an "insufficient memory"
error when running Windows (ME) Scandisk and Defrag on the 500 gb
drive.

DOS scandisk does not give an error, but it would have taken 4 days to
run (given it was at the 30% point after 30 hours).