Swap file optimization

Hi everyone,

I am trying to figure out what is the best possible location for the =
Windows 98 swap file on this computer, which has one 80 GB hard disk. I =
was setting the computer up last summer, and I read through the =
following stuff on PCGuide.com:

http://www.pcguide.com/opt/opt/osSwapLocation-c.html
http://www.pcguide.com/ref/hdd/file/partSpecial-c.html

Currently, I have a separate partition for the swap file, geographically =
somewhere close to the middle of the disk. The partition has 32k =
clusters (I had to make it FAT16 in order to have the clusters that =
large). I am now thinking what I did might not have been that smart or =
might simply have been unnecessary. I have the following questions:

1. Is it better to have 32k clusters (smaller # of clusters, less =
fragmentation - although this wouldn't matter much if we're talking =
about a dedicated swap file partition, right?) than smaller ones? I =
remember Gary Terhune or someone mentioning that 4k clusters are =
actually best for the swap file, since Windows 98 processes memory in 4k =
sections?

2. Does any of this really matter on this machine, which is a Pentium 4 =
2.40 Ghz with 256 MB of RAM? I find that the swap file is used very =
rarely. Would it thus be smart for me to get rid of the special swap =
file partition and put the swap file on C: or on a different partition =
which contains temporary files?

3. Does fragmentation really matter that much on a machine this fast? I =
have not seen defragmentation cause any great performance gains.

4. How would answers to the above be different for a Pentium III, 450 =
Mhz machine (but also 256 MB)?


Thanks in advance for anybody's thoughts on these issues.

Ivan

P.S. Nothing like having Windows 98 on a modern system - shuts down in =
about 2-3 seconds!

Only real gain can be seen by putting the swapfile on a separate hard
drive's partition at the beginning of that hard drive. And that hard drive
should be as fast as the hard drive containing the OS. Since there's never
actual dual access on the same ide channel, put the hard drive on a separate
ide port, separate ide adapter card, or scsi adapter card.

FAT16 will have less overhead with larger clusters. Process memory size is
of no consequence, a moot point.
"Ivan Bútora" wrote in message
...
Hi everyone,

I am trying to figure out what is the best possible location for the Windows
98 swap file on this computer, which has one 80 GB hard disk. I was setting
the computer up last summer, and I read through the following stuff on
PCGuide.com:

http://www.pcguide.com/opt/opt/osSwapLocation-c.html
http://www.pcguide.com/ref/hdd/file/partSpecial-c.html

Currently, I have a separate partition for the swap file, geographically
somewhere close to the middle of the disk. The partition has 32k clusters (I
had to make it FAT16 in order to have the clusters that large). I am now
thinking what I did might not have been that smart or might simply have been
unnecessary. I have the following questions:

1. Is it better to have 32k clusters (smaller # of clusters, less
fragmentation - although this wouldn't matter much if we're talking about a
dedicated swap file partition, right?) than smaller ones? I remember Gary
Terhune or someone mentioning that 4k clusters are actually best for the
swap file, since Windows 98 processes memory in 4k sections?

2. Does any of this really matter on this machine, which is a Pentium 4 2.40
Ghz with 256 MB of RAM? I find that the swap file is used very rarely. Would
it thus be smart for me to get rid of the special swap file partition and
put the swap file on C: or on a different partition which contains temporary
files?

3. Does fragmentation really matter that much on a machine this fast? I have
not seen defragmentation cause any great performance gains.

4. How would answers to the above be different for a Pentium III, 450 Mhz
machine (but also 256 MB)?


Thanks in advance for anybody's thoughts on these issues.

Ivan

P.S. Nothing like having Windows 98 on a modern system - shuts down in about
2-3 seconds!

Ivan Bútora wrote in message ...
Hi everyone,

I am trying to figure out what is the best possible location for the Windows 98 swap file on this computer, which has one 80 GB hard
disk.


HC The best possible location would be on another hard drive.
As you don't have one, then it doesn't really matter.

I was setting the computer up last summer, and I read through the following stuff on PCGuide.com:

http://www.pcguide.com/opt/opt/osSwapLocation-c.html
http://www.pcguide.com/ref/hdd/file/partSpecial-c.html

Currently, I have a separate partition for the swap file, geographically somewhere close to the middle of the disk. The partition
has 32k clusters (I had to make it FAT16 in order to have the clusters that large). I am now thinking what I did might not have been
that smart or might simply have been unnecessary.

HC: The cluster size dictates the minimum size that any one file can be.
If you create a file containing just one character, the allocated file size
will be 32,768 characters, as will a file containing any number of
characters up to 32,768. A massive waste of space.

I have the following questions:

1. Is it better to have 32k clusters

HC: No.

(smaller # of clusters, less fragmentation - although this wouldn't matter much if we're talking about a dedicated swap file
partition, right?) than smaller ones? I remember Gary Terhune or someone mentioning that 4k clusters are actually best for the swap
file, since Windows 98 processes memory in 4k sections?

HC: Mr Terhune, as usual, knoweth of what he speaketh.

2. Does any of this really matter on this machine, which is a Pentium 4 2.40 Ghz with 256 MB of RAM? I find that the swap file is
used very rarely. Would it thus be smart for me to get rid of the special swap file partition and put the swap file on C: or on a
different partition which contains temporary files?

HC: Sure, it matters. The swap file is really inconsequential to the debate.
EVERY file on your hard disk is impacted by the cluster size.
In a WORST case scenario, you have reduced your hard drive capacity
to the equivalent of a 10 Gig drive.

3. Does fragmentation really matter that much on a machine this fast?
I have not seen defragmentation cause any great performance gains.

HC: You just answered your own question.

4. How would answers to the above be different for a Pentium III, 450 Mhz machine (but also 256 MB)?

HC: Any speed increase would be a larger percentage gain,
and therefore would be more noticeable.


Thanks in advance for anybody's thoughts on these issues.

Ivan

P.S. Nothing like having Windows 98 on a modern system - shuts down in about 2-3 seconds!

I do not have 32k clusters on any partition except the dedicated swap =
file partition. I understand I would be wasting space, but this is not =
an issue on a dedicated swap file partition, since the swap file is the =
only file there. The advice at http://www.aumha.org/a/parts.php also =
points to setting 32k clusters on the swap file partition, but I =
remember somebody said that 4k is best for some reason (and I am not =
entirely sure it was Gary Terhune or somebody else).

I am wondering if canceling the separate swap file partition and putting =
the swap file somewhere else would not lead to fragmentation of the swap =
file (and is that a bad thing even on a fast machine, or does that also =
not matter much?)

Thanks,

Ivan



"Hugh Candlin" wrote in message =
...
=20
Ivan B=FAtora wrote in message =
...
Hi everyone,
=20
I am trying to figure out what is the best possible location for the =
Windows 98 swap file on this computer, which has one 80 GB hard
disk.
=20
=20
HC The best possible location would be on another hard drive.
As you don't have one, then it doesn't really matter.
=20
I was setting the computer up last summer, and I read through the =
following stuff on PCGuide.com:
=20
http://www.pcguide.com/opt/opt/osSwapLocation-c.html
http://www.pcguide.com/ref/hdd/file/partSpecial-c.html
=20
Currently, I have a separate partition for the swap file, =
geographically somewhere close to the middle of the disk. The partition
has 32k clusters (I had to make it FAT16 in order to have the clusters =
that large). I am now thinking what I did might not have been
that smart or might simply have been unnecessary.
=20
HC: The cluster size dictates the minimum size that any one file can =
be.
If you create a file containing just one character, the =
allocated file size
will be 32,768 characters, as will a file containing any =
number of
characters up to 32,768. A massive waste of space.
=20
I have the following questions:
=20
1. Is it better to have 32k clusters
=20
HC: No.
=20
(smaller # of clusters, less fragmentation - although this wouldn't =
matter much if we're talking about a dedicated swap file
partition, right?) than smaller ones? I remember Gary Terhune or =
someone mentioning that 4k clusters are actually best for the swap
file, since Windows 98 processes memory in 4k sections?
=20
HC: Mr Terhune, as usual, knoweth of what he speaketh.
=20
2. Does any of this really matter on this machine, which is a Pentium =
4 2.40 Ghz with 256 MB of RAM? I find that the swap file is
used very rarely. Would it thus be smart for me to get rid of the =
special swap file partition and put the swap file on C: or on a
different partition which contains temporary files?
=20
HC: Sure, it matters. The swap file is really inconsequential to the =
debate.
EVERY file on your hard disk is impacted by the cluster size.
In a WORST case scenario, you have reduced your hard drive =
capacity
to the equivalent of a 10 Gig drive.
=20
3. Does fragmentation really matter that much on a machine this fast?
I have not seen defragmentation cause any great performance gains.
=20
HC: You just answered your own question.
=20
4. How would answers to the above be different for a Pentium III, 450 =
Mhz machine (but also 256 MB)?
=20
HC: Any speed increase would be a larger percentage gain,
and therefore would be more noticeable.
=20
=20
Thanks in advance for anybody's thoughts on these issues.
=20
Ivan
=20
P.S. Nothing like having Windows 98 on a modern system - shuts down in =
about 2-3 seconds!
=20

I don't think you'll see any difference.
Try moving your swap file back to the C:\ drive (don't change the
dedicated partition yet) and try it in some of your most 'ram
intensive' programs and see.
If you can't tell the difference, what does it matter.

"Ivan Bútora" wrote in message
...
I do not have 32k clusters on any partition except the dedicated swap
file partition. I understand I would be wasting space, but this is not
an issue on a dedicated swap file partition, since the swap file is
the only file there. The advice at http://www.aumha.org/a/parts.php
also points to setting 32k clusters on the swap file partition, but I
remember somebody said that 4k is best for some reason (and I am not
entirely sure it was Gary Terhune or somebody else).

I am wondering if canceling the separate swap file partition and
putting the swap file somewhere else would not lead to fragmentation
of the swap file (and is that a bad thing even on a fast machine, or
does that also not matter much?)

Thanks,

Ivan



"Hugh Candlin" wrote in message
...

Ivan Bútora wrote in message
...
Hi everyone,

I am trying to figure out what is the best possible location for the
Windows 98 swap file on this computer, which has one 80 GB hard
disk.


HC The best possible location would be on another hard drive.
As you don't have one, then it doesn't really matter.

I was setting the computer up last summer, and I read through the
following stuff on PCGuide.com:

http://www.pcguide.com/opt/opt/osSwapLocation-c.html
http://www.pcguide.com/ref/hdd/file/partSpecial-c.html

Currently, I have a separate partition for the swap file,
geographically somewhere close to the middle of the disk. The
partition
has 32k clusters (I had to make it FAT16 in order to have the
clusters that large). I am now thinking what I did might not have been
that smart or might simply have been unnecessary.

HC: The cluster size dictates the minimum size that any one file
can be.
If you create a file containing just one character, the
allocated file size
will be 32,768 characters, as will a file containing any
number of
characters up to 32,768. A massive waste of space.

I have the following questions:

1. Is it better to have 32k clusters

HC: No.

(smaller # of clusters, less fragmentation - although this wouldn't
matter much if we're talking about a dedicated swap file
partition, right?) than smaller ones? I remember Gary Terhune or
someone mentioning that 4k clusters are actually best for the swap
file, since Windows 98 processes memory in 4k sections?

HC: Mr Terhune, as usual, knoweth of what he speaketh.

2. Does any of this really matter on this machine, which is a
Pentium 4 2.40 Ghz with 256 MB of RAM? I find that the swap file is
used very rarely. Would it thus be smart for me to get rid of the
special swap file partition and put the swap file on C: or on a
different partition which contains temporary files?

HC: Sure, it matters. The swap file is really inconsequential to
the debate.
EVERY file on your hard disk is impacted by the cluster
size.
In a WORST case scenario, you have reduced your hard drive
capacity
to the equivalent of a 10 Gig drive.

3. Does fragmentation really matter that much on a machine this
fast?
I have not seen defragmentation cause any great performance
gains.

HC: You just answered your own question.

4. How would answers to the above be different for a Pentium III,
450 Mhz machine (but also 256 MB)?

HC: Any speed increase would be a larger percentage gain,
and therefore would be more noticeable.


Thanks in advance for anybody's thoughts on these issues.

Ivan

P.S. Nothing like having Windows 98 on a modern system - shuts down
in about 2-3 seconds!

Well, I did a search on Google:
http://groups.google.com/groups?hl=3...Dswap+file+4k=
+group%3Amicrosoft.public.win98.gen_discussion&btn G=3DSearch&meta=3Dgroup=
%3Dmicrosoft.public.win98.gen_discussion

and based on the information provided by several MVPs (Terhune, Martell, =
Harper, Clayton, Woodruff), Windows writes to and reads from the swap =
file in 4k blocks. Do I thus understand correctly that if my swap file =
partition has 32k clusters, each cluster will only have 4k of data, and =
thus the number of clusters will be the same if the cluster size is =
equal to or greater than 4k?



"Lil' Dave" wrote in message =
...
Only real gain can be seen by putting the swapfile on a separate hard
drive's partition at the beginning of that hard drive. And that hard =
drive
should be as fast as the hard drive containing the OS. Since there's =
never
actual dual access on the same ide channel, put the hard drive on a =
separate
ide port, separate ide adapter card, or scsi adapter card.
=20
FAT16 will have less overhead with larger clusters. Process memory =
size is
of no consequence, a moot point.
"Ivan B=FAtora" wrote in message
...
Hi everyone,
=20
I am trying to figure out what is the best possible location for the =
Windows
98 swap file on this computer, which has one 80 GB hard disk. I was =
setting
the computer up last summer, and I read through the following stuff on
PCGuide.com:
=20
http://www.pcguide.com/opt/opt/osSwapLocation-c.html
http://www.pcguide.com/ref/hdd/file/partSpecial-c.html
=20
Currently, I have a separate partition for the swap file, =
geographically
somewhere close to the middle of the disk. The partition has 32k =
clusters (I
had to make it FAT16 in order to have the clusters that large). I am =
now
thinking what I did might not have been that smart or might simply =
have been
unnecessary. I have the following questions:
=20
1. Is it better to have 32k clusters (smaller # of clusters, less
fragmentation - although this wouldn't matter much if we're talking =
about a
dedicated swap file partition, right?) than smaller ones? I remember =
Gary
Terhune or someone mentioning that 4k clusters are actually best for =
the
swap file, since Windows 98 processes memory in 4k sections?
=20
2. Does any of this really matter on this machine, which is a Pentium =
4 2.40
Ghz with 256 MB of RAM? I find that the swap file is used very rarely. =
Would
it thus be smart for me to get rid of the special swap file partition =
and
put the swap file on C: or on a different partition which contains =
temporary
files?
=20
3. Does fragmentation really matter that much on a machine this fast? =
I have
not seen defragmentation cause any great performance gains.
=20
4. How would answers to the above be different for a Pentium III, 450 =
Mhz
machine (but also 256 MB)?
=20
=20
Thanks in advance for anybody's thoughts on these issues.
=20
Ivan
=20
P.S. Nothing like having Windows 98 on a modern system - shuts down in =
about
2-3 seconds!
=20