Why does XP reset the size of my virtual memory?

We've observed this on all our XP Pro computers; regardless of the virtual
memory custom size setting, upon reboot the pagefile is reset to 1.5 times my
physical memory.

To change the size of an existing pagefile, I must :
* Set XP to "No paging file" and reboot. After reboot, there is no
pagefile, as expected.
* Set a custom pagefile at 2048 (initial & maximum size boxes) & reboot.
Upon restart, there is a 2048M pagefile, as expected.
* When rebooted again (without doing anything else), low & behold the
pagefile is 1534M, 1.5 times my physical memory (1G of RAM)
* If I open the Virtual Memory dialog it still shows 2048 in the two custom
size text boxes, but the currently allocated size is 1534M, and nothing I can
do will increase the size aside from the procedure listed (which is lost
after the next reboot).


As stated, this is observed on all our XP Pro computers (not our 2000
machines), laptops or desktops, Dells, Toshibas, HPs, etc.

I've seen other people complain about this (thanks google.com), but can't
find a solution.

Any ideas?

Ken

"Ratboy Ken" Ratboy wrote:

We've observed this on all our XP Pro computers; regardless of the virtual
memory custom size setting, upon reboot the pagefile is reset to 1.5 times my
physical memory.

To change the size of an existing pagefile, I must :
* Set XP to "No paging file" and reboot. After reboot, there is no
pagefile, as expected.
* Set a custom pagefile at 2048 (initial & maximum size boxes) & reboot.
Upon restart, there is a 2048M pagefile, as expected.
* When rebooted again (without doing anything else), low & behold the
pagefile is 1534M, 1.5 times my physical memory (1G of RAM)
* If I open the Virtual Memory dialog it still shows 2048 in the two custom
size text boxes, but the currently allocated size is 1534M, and nothing I can
do will increase the size aside from the procedure listed (which is lost
after the next reboot).


As stated, this is observed on all our XP Pro computers (not our 2000
machines), laptops or desktops, Dells, Toshibas, HPs, etc.

I've seen other people complain about this (thanks google.com), but can't
find a solution.

Any ideas?

Ken

See MVP Alex Nichol's article on Virtual Memory in Windows XP at
http://aumha.org/win5/a/xpvm.htm especially the section dealing with
problems. And note the section dealing with problems caused by Norton
Antvirus if your computers happen to be infested with that
virus///product.

Good luck


Ron Martell Duncan B.C. Canada
--
Microsoft MVP
On-Line Help Computer Service
http://onlinehelp.bc.ca

"The reason computer chips are so small is computers don't eat much."

"Ratboy Ken" Ratboy wrote:

We've observed this on all our XP Pro computers; regardless of the virtual
memory custom size setting, upon reboot the pagefile is reset to 1.5 times my
physical memory.

To change the size of an existing pagefile, I must :
* Set XP to "No paging file" and reboot. After reboot, there is no
pagefile, as expected.
* Set a custom pagefile at 2048 (initial & maximum size boxes) & reboot.
Upon restart, there is a 2048M pagefile, as expected.
* When rebooted again (without doing anything else), low & behold the
pagefile is 1534M, 1.5 times my physical memory (1G of RAM)

First Do not try to waste that much disk space on the page file with
1GB RAM

When you change placement to a different drive, you need to leave a
token file - init 2 max 50 is fine - on C *and always click Set before
going on*. For your RAM size I would just set that as initial 100, max
maybe 1000 or 2000. The file is unlikely to grow beyond the 100 unless
you are running programs that make *very* heavy demands on memory. The
nxRAM is plain bad advice. See more at my page
www.aumha.org/win5/a/xpvm.htm


--
Alex Nichol MS MVP (Windows Technologies)
Bournemouth, U.K. (remove the D8 bit)

Thanks Alex,

We are running very heavy memory demanding scientific programs that need
virtual memory in excess of 1G. If a computer has only 512M of RAM, the
FORTRAN program will not load unless the virtual memory is set to 1G.
However, under XP, everytime they reboot their pagefile is reset to ~750M,
and the program will not load. The complaints from our processors in the
field is becoming annoying! This never happened under Win2000.

We only have one drive per computer, so the pagefile must sit on C drive.
And I can change the size once without problem, but then a subsequent reboot
and it goes to 1.5xRAM... always.

As for Norton Antivirus, I am running SystemWorks 2004 Pro with all the
updates as posted by Symantec. Everybody in this newsgroup seems to be
pointing the finger of blame on them... maybe so. But searching Symantec's
website, I couldn't find anything related to this problem. Fortunately I will
be taking receipt of a brand new virgin laptop from Dell in the next few days
(no antivirus solution preinstalled) and I am going to test this hypothesis.

The only way to know is to experiment.

Ken

"Alex Nichol" wrote:

"Ratboy Ken" Ratboy wrote:

We've observed this on all our XP Pro computers; regardless of the virtual
memory custom size setting, upon reboot the pagefile is reset to 1.5 times my
physical memory.

To change the size of an existing pagefile, I must :
* Set XP to "No paging file" and reboot. After reboot, there is no
pagefile, as expected.
* Set a custom pagefile at 2048 (initial & maximum size boxes) & reboot.
Upon restart, there is a 2048M pagefile, as expected.
* When rebooted again (without doing anything else), low & behold the
pagefile is 1534M, 1.5 times my physical memory (1G of RAM)

First Do not try to waste that much disk space on the page file with
1GB RAM

When you change placement to a different drive, you need to leave a
token file - init 2 max 50 is fine - on C *and always click Set before
going on*. For your RAM size I would just set that as initial 100, max
maybe 1000 or 2000. The file is unlikely to grow beyond the 100 unless
you are running programs that make *very* heavy demands on memory. The
nxRAM is plain bad advice. See more at my page
www.aumha.org/win5/a/xpvm.htm


--
Alex Nichol MS MVP (Windows Technologies)
Bournemouth, U.K. (remove the D8 bit)

Ratboy Ken Wrote:
Thanks Alex,

We are running very heavy memory demanding scientific programs that
need
virtual memory in excess of 1G. If a computer has only 512M of RAM,
the
FORTRAN program will not load unless the virtual memory is set to 1G.
However, under XP, everytime they reboot their pagefile is reset to
~750M,
and the program will not load. The complaints from our processors in
the
field is becoming annoying! This never happened under Win2000.

We only have one drive per computer, so the pagefile must sit on C
drive.
And I can change the size once without problem, but then a subsequent
reboot

Ken

"Alex Nichol" wrote:

"Ratboy Ken" Ratboy wrote:

We've observed this on all our XP Pro computers; regardless of the
virtual
memory custom size setting, upon reboot the pagefile is reset to 1.5
times my
physical memory.

To change the size of an existing pagefile, I must :
* Set XP to "No paging file" and reboot. After reboot, there is no
pagefile, as expected.
* Set a custom pagefile at 2048 (initial & maximum size boxes) &
reboot.
Upon restart, there is a 2048M pagefile, as expected.
* When rebooted again (without doing anything else), low & behold
the
pagefile is 1534M, 1.5 times my physical memory (1G of RAM)

First Do not try to waste that much disk space on the page file
with
1GB RAM

When you change placement to a different drive, you need to leave a
token file - init 2 max 50 is fine - on C *and always click Set
before
going on*. For your RAM size I would just set that as initial 100,
max
maybe 1000 or 2000. The file is unlikely to grow beyond the 100
unless
you are running programs that make *very* heavy demands on memory.
The
nxRAM is plain bad advice. See more at my page
www.aumha.org/win5/a/xpvm.htm


--
Alex Nichol MS MVP (Windows Technologies)
Bournemouth, U.K. (remove the D8 bit)


Ken, of course you need a bigger pagefile then what Alex is suggesting,
everybody does, and the settings he suggests will always cause
performance hits.

in your situtuation, you obviously need even more then the default.

have you tried changing the settings as the administrator?..not an
administrator but THE administrator.


--
perris
------------------------------------------------------------------------
perris's Profile: http://forum.osnn.net/member.php?userid=17
View this thread: http://forum.osnn.net/showthread.php?t=50124

perris Wrote:
Ken, of course you need a bigger pagefile then what Alex is suggesting,
everybody does, and the settings he suggests will always cause
performance hits.

in your situtuation, you obviously need even more then the default.

have you tried changing the settings as the administrator?..not an
administrator but THE administrator.
if that doesn't do the trick, there are a few common causes for this
problem, norton being one of them, or:

1. You set some nonstandard permissions on the root directory of
whatever partition you're using. 'system' must have 'full control'.

2. The registry value where the pagefile settings get stored might have
become corrupted. It's in

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Contro l\Session
Manager\Memory Management

the value is called "PagingFiles". It should be a REG_MULTI_SZ and for
the setting you're trying there, should look like this:

D:\pagefile.sys 800 800, those two numbers representing your settings

Try deleting it, then go into the System applet again and set it to
what you want it to be from there.

I hope that helps


--
perris
------------------------------------------------------------------------
perris's Profile: http://forum.osnn.net/member.php?userid=17
View this thread: http://forum.osnn.net/showthread.php?t=50124

perris Wrote:
if that doesn't do the trick, there are a few common causes for this
problem, norton being one of them, or:

1. You set some nonstandard permissions on the root directory of
whatever partition you're using. 'system' must have 'full control'.

2. The registry value where the pagefile settings get stored might have
become corrupted. It's in

HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Contro l\Session
Manager\Memory Management

the value is called "PagingFiles". It should be a REG_MULTI_SZ and for
the setting you're trying there, should look like this:

D:\pagefile.sys 800 800, those two numbers representing your settings

Try deleting it, then go into the System applet again and set it to
what you want it to be from there.

I hope that helps
and finally, if none of that does the trick, try the solution here;

http://tinyurl.com/52abt


--
perris
------------------------------------------------------------------------
perris's Profile: http://forum.osnn.net/member.php?userid=17
View this thread: http://forum.osnn.net/showthread.php?t=50124

perris wrote:


Ken, of course you need a bigger pagefile then what Alex is suggesting,
everybody does, and the settings he suggests will always cause
performance hits.

That I flatly deny. *Everybody* does not, though This particular
workload probably does.

--
Alex Nichol MS MVP (Windows Technologies)
Bournemouth, U.K. (remove the D8 bit)

Alex Nichol Wrote:
perris wrote:


Ken, of course you need a bigger pagefile then what Alex is
suggesting,
everybody does, and the settings he suggests will always cause
performance hits.

That I flatly deny. *Everybody* does not, though This particular
workload probably does.

--
Alex Nichol MS MVP (Windows Technologies)
Bournemouth, U.K. (remove the D8 bit)
deny it all you like, it is a fact...data in memory has to be charged
against the hardrive where the memory manager will release and reclaim
as a users need indicates, private writeable address is charged against
the pagefile, your settings eliminate the area on the hardrive that's
neccessary for the memory manger to be efficient.

you seem to think that a the pagefile only needs to be as big as the
amount of info that's likey to be written to it.

no, all datat in memory NEED it's OWN area on the hardrive for the
memory manager

you also seem quite willing to invite expansion, and more then willing
to have users invite expansion

for this, I am amazed


--
perris
------------------------------------------------------------------------
perris's Profile: http://forum.osnn.net/member.php?userid=17
View this thread: http://forum.osnn.net/showthread.php?t=50124

Alex Nichol Wrote:
perris wrote:


Ken, of course you need a bigger pagefile then what Alex is
suggesting,
everybody does, and the settings he suggests will always cause
performance hits.

That I flatly deny. *Everybody* does not, though This particular
workload probably does.

--
Alex Nichol MS MVP (Windows Technologies)
Bournemouth, U.K. (remove the D8 bit)
here's an important point for you to consider Alex, since you seem
determined to "save" 4 or 5 hundred mbs of hardrive area for some
purpose I can't figure out.

left to the default, xp uses LESS hardrive not more hardrive then your
settings.

this is because at the default, the ow can actually make the pf smaller
then your settings if storage are becomes more of an issue then
performance

your settings, this is NOT the case.

your settings accomplish NOTHING in saving hardrive area, and COST
hardrive area...in addition, your settings actually INVITE expansion.

I'm waiting for you to give SOME reason you think your settings are
better then default for people that don't have a storage issue

any reason at all will do


--
perris
------------------------------------------------------------------------
perris's Profile: http://forum.osnn.net/member.php?userid=17
View this thread: http://forum.osnn.net/showthread.php?t=50124

on your benchmark I stand corrected ming, though I'd like to see that
happen myself


--
perris
------------------------------------------------------------------------
perris's Profile: http://forum.osnn.net/member.php?userid=17
View this thread: http://forum.osnn.net/showthread.php?t=50124

perris Wrote:
on your benchmark I stand corrected ming, though I'd like to see that
happen myself
oops, wrong thread, dissregard my previous post


--
perris
------------------------------------------------------------------------
perris's Profile: http://forum.osnn.net/member.php?userid=17
View this thread: http://forum.osnn.net/showthread.php?t=50124

perris Wrote:
on your benchmark I stand corrected ming, though I'd like to see that
happen myself
sorry, this response is intended for a thread regarding hibernate,
dissregard this post


--
perris
------------------------------------------------------------------------
perris's Profile: http://forum.osnn.net/member.php?userid=17
View this thread: http://forum.osnn.net/showthread.php?t=50124

perris wrote:


deny it all you like, it is a fact...data in memory has to be charged
against the hardrive where the memory manager will release and reclaim
as a users need indicates, private writeable address is charged against
the pagefile, your settings eliminate the area on the hardrive that's
neccessary for the memory manger to be efficient.

Balderdash. Hogwash. Malarkey.

The pagefile in Windows XP is used for the following specific
functions in Windows XP:

1. To compensate for the lack of sufficient physical RAM in the
computer to meet the total memory load requirements.

2. To fulfill the memory address space requirements for the unused
portions of memory allocation requests. And all that is required in
Windows XP is that the maximum size limit be large enough so that the
page file could be increased if these memory items were to be actually
used.

3. To hold the memory content for other users if multiple users are
configured on the computer and if the "fast user switching" option is
in effect.

4. To received the contents of the "system failure memory dumps" if a
memory dump option has been configured. This requires that the
pagefile be located on the boot drive.


you seem to think that a the pagefile only needs to be as big as the
amount of info that's likey to be written to it.

no, all datat in memory NEED it's OWN area on the hardrive for the
memory manager

Wrong. All requested memory must be allocated memory address space.
These addresses may be either in RAM or in the page file. There is no
requirement for the same items to be allocated space in both. The
memory manager decides which items will be in RAM and which will be in
the pagefile on a dynamic basis and swaps them back and forth as
requirements change.


you also seem quite willing to invite expansion, and more then willing
to have users invite expansion

for this, I am amazed

For meeting the memory address requirements of the unused portion of
memory allocation requests all that is requires is that the potential
to enlarge the pagefile exist. It does not have to actually be
enlarged for these items. The unused portions of requested memory
can easily aggregate to several hundred megabytes even on a lightly
used system. For example on my own system these items currently total
208 mb. Task Monitor tells me that the Page File Usage is 308 mb
while another utility tells me that there is only 94 mb of active
memory content residing in the page file. And the actual size of the
pagefile is 160 mb, which is the minimum that I have set for it.


Ron Martell Duncan B.C. Canada
--
Microsoft MVP
On-Line Help Computer Service
http://onlinehelp.bc.ca

"The reason computer chips are so small is computers don't eat much."

Ron Martell Wrote:
perris wrote:


deny it all you like, it is a fact...data in memory has to be charged
against the hardrive where the memory manager will release and
reclaim
as a users need indicates, private writeable address is charged
against
the pagefile, your settings eliminate the area on the hardrive that's
neccessary for the memory manger to be efficient.

Balderdash. Hogwash. Malarkey.

The pagefile in Windows XP is used for the following specific
functions in Windows XP:

1. To compensate for the lack of sufficient physical RAM in the
computer to meet the total memory load requirements.

2. To fulfill the memory address space requirements for the unused
portions of memory allocation requests. And all that is required in
Windows XP is that the maximum size limit be large enough so that the
page file could be increased if these memory items were to be actually
used.

3. To hold the memory content for other users if multiple users are
configured on the computer and if the "fast user switching" option is
in effect.

4. To received the contents of the "system failure memory dumps" if a
memory dump option has been configured. This requires that the
pagefile be located on the boot drive.


you seem to think that a the pagefile only needs to be as big as the
amount of info that's likey to be written to it.

no, all data in memory NEED it's OWN area on the hardrive for the
memory manager

Wrong. All requested memory must be allocated memory address space.
These addresses may be either in RAM or in the page file. There is no
requirement for the same items to be allocated space in both. The
memory manager decides which items will be in RAM and which will be in
the pagefile on a dynamic basis and swaps them back and forth as
requirements change.


you also seem quite willing to invite expansion, and more then
willing
to have users invite expansion

for this, I am amazed

For meeting the memory address requirements of the unused portion of
memory allocation requests all that is requires is that the potential
to enlarge the pagefile exist. It does not have to actually be
enlarged for these items. The unused portions of requested memory
can easily aggregate to several hundred megabytes even on a lightly
used system. For example on my own system these items currently total
208 mb. Task Monitor tells me that the Page File Usage is 308 mb
while another utility tells me that there is only 94 mb of active
memory content residing in the page file. And the actual size of the
pagefile is 160 mb, which is the minimum that I have set for it.


Ron Martell Duncan B.C. Canada
--
Microsoft MVP
On-Line Help Computer Service
http://onlinehelp.bc.ca

"The reason computer chips are so small is computers don't eat much."

incorrect, just about your entire response to my post

quote;

The pagefile in Windows XP is used for the following specific
functions in Windows XP:

1. To compensate for the lack of sufficient physical RAM in the
computer to meet the total memory load requirements. unquote

incorrect, the pagefile is only to provide backing store for modified
pages so they can be considered by the memory manager

everything that's not modified gets backed to the hardrive from whence
it came...the exe, dll, whatever

you would need over 2 gigs to run xp without backing store, and
everything in memory needs a place, it's OWN place on the hardrive so
the memory manager will be able to conceder it in the memory management
model

as far as memory dumps, ya, that's a good purpose of it too...you do
have that one right

you also got the following right;

quote;

The
memory manager decides which items will be in RAM and which will be in
the pagefile on a dynamic basis and swaps them back and forth as
requirements change. unquote

memory is addressed first and allocated second, the memory manager
needs an area to perform these "swaps" you're speaking about, the
"swap" space is not shared

and your claim

quote


For meeting the memory address requirements of the unused portion of
memory allocation requests all that is requires is that the potential
to enlarge the pagefile exist. It does not have to actually be
enlarged for these items. The unused portions of requested memory
can easily aggregate to several hundred megabytes even on a lightly
used system. For example on my own system these items currently total
208 mb. Task Monitor tells me that the Page File Usage is 308 mb
while another utility tells me that there is only 94 mb of active
memory content residing in the page file. And the actual size of the
pagefile is 160 mb, which is the minimum that I have set for it.

unquote

rediculous...you think that just because only 94 mbs of information is
actually in the pagefile, that's all that the memory manager is
charging to it?

obsurd...taskmanager is exactly correct in what is charged to the
pagefile, yet you want to circumvent this strategy.

the kernel team IS EXTREMELY happy with the memory management model of
the NT kernel, and yes, they do know how much memory is available on
modern systems

they've continued to raise, not lower the recommendations for pagefile,
the continue their recommendations in server 2003, and in longhorn

how you can defend circumventing the recommendation of the kernel team
when as a fact you KNOW there is no performance to gain for the effort,
and wuite a bit to loose for some users, (as the very poster of this
thread clearly demonstrates) is irresponsible in every sense

whether or not YOU put your memory under pressure doesn't mean I
don't, or my customers, or the people that work for me, and those that
mess with these machines because of the irresponsible papers that
"recommend" lowering the default for absolutely NO reason whatsoever

in case you didn't know it, Microsoft even wrote hacks for users to
overcome the 4 gig threshold for page files


--
perris
------------------------------------------------------------------------
perris's Profile: http://forum.osnn.net/member.php?userid=17
View this thread: http://forum.osnn.net/showthread.php?t=50124