Pagefile Size

[Robert Heiling]

What is it that you are trying to accomplish by setting a
pagefile extremely large without any reason to think it
should be that large?

By "extremely", you mean approximately double what some are already discussing?
What percent of a 250GB or + or even a 120GB HD is that? Are you so tight on
disk space that 2-3 GB represents a big hit? Sure! on the 10 yr old system out
in the garage with an 8 GB HD shared as dual-boot, I don't have a 4 GB pagefile,
but things have changed a lot in the past 10 years.

If it is merely to be sure you don't run out of virtual
memory space,

No. It's actually to be sure that I won't have to worry like the people are
doing here right now. The only mistake you can make is to make it too small.
There's no downside to making it "too big" other than a little bit of disk space
that isn't going to break the bank.

do you see a lot of people reporting they have
that problem? No, and it's not at all usual for them to
have set a 4GB pagefile

I see people here with a problem in making a decision and attempting to
micro-manage and outguess the OS and being presented with conflicting advice.
That's what I see.

IOW, you have a solution for a problem that doesn't usually
exist.

The existence of this discussion proves that a number of people have a problem
deciding how large (or small) their pagefile should be. It's a problem that I
don't have. I'd call that a solution, Yes.

Bob

 

[kony]

By "extremely", you mean approximately double what some are already discussing?

Let's get more specific... How about one byte (ignoring that
it can't be set to a mere one byte difference)? What
justification do you have to set your pagefile even ONE BYTE
larger than others who have no issues with their pagefile
being undersized?

What percent of a 250GB or + or even a 120GB HD is that?

Doesn't matter.
Overkill for no reason is senseless.
IF you had reason, it'd be an entirely different matter.

Are you so tight on
disk space that 2-3 GB represents a big hit? Sure! on the 10 yr old system out
in the garage with an 8 GB HD shared as dual-boot, I don't have a 4 GB pagefile,
but things have changed a lot in the past 10 years.

So you don't really have a reason, just "I have free HDD
space." OK, if you want to be lazy about it, it's your
machine, but that's hardly a reason to advocate it.

No. It's actually to be sure that I won't have to worry like the people are
doing here right now. The only mistake you can make is to make it too small.
There's no downside to making it "too big" other than a little bit of disk space
that isn't going to break the bank.

I don't have a 4GB pagefile and am not worrying. The vast
majority of people don't either. Bob is just nuts. ;-)

I see people here with a problem in making a decision and attempting to
micro-manage and outguess the OS and being presented with conflicting advice.
That's what I see.

No, the OS has to guess because it isn't designed to
dynamically monitor and readjust. The user isn't
constrained like this, they can make an informed and
targeted decision in retrospect, or consideration of future
needs. Only so much logic can be reasonably built into the
OS, and the rest is provided by a savvy user.

It's similar to MANY different OS defaults, did you really
keep ALL of those defaults?

The existence of this discussion proves that a number of people have a problem
deciding how large (or small) their pagefile should be. It's a problem that I
don't have. I'd call that a solution, Yes.

Actually most people never change the setting - and end up
with a much smaller pagefile than you have, never growing
even close to that size, and never have a problem.

Since it is a discussion quite specifically about what size
to make it, the answer "as big as possible" really doesn't
address anything but the OS build-in limit, which they'd
assumed at the time would be far larger than necessary else
they'd have made it even bigger (max).

 

[Robert Heiling]

Let's get more specific... How about one byte (ignoring that
it can't be set to a mere one byte difference)? What
justification do you have to set your pagefile even ONE BYTE
larger than others who have no issues with their pagefile
being undersized?

Reserve capacity is an easy answer to that question. The problem is that an
"undersized" value hasn't been agreed upon here from what I've seen. Why do you
suppose that MS gives a recommendation for the minimum, but not for the maximum
(your "undersized" if not enough)?

Doesn't matter.
Overkill for no reason is senseless.
IF you had reason, it'd be an entirely different matter.

Not overkill. A different sense of priorities that I prefer. But Ok, Then let's
hear what you think it should be for my system. Let's hear what you think will
never be too little under any circumstances of use.

So you don't really have a reason, just "I have free HDD
space." OK, if you want to be lazy about it, it's your
machine, but that's hardly a reason to advocate it.

I'm no different than anyone else who advocates to others what he had found to
work for himself.

I don't have a 4GB pagefile and am not worrying. The vast
majority of people don't either.

The vast majority don't even know they have a pagefile, in fact, they aren't
even permitted to look at system files.

Bob is just nuts. ;-)

That's old news. :^}

No, the OS has to guess because it isn't designed to
dynamically monitor and readjust.

That's strange. Then why do they offer a minimum and a maximum size and how does
the OS manage to readjust between those 2 limits?

The user isn't
constrained like this, they can make an informed and
targeted decision in retrospect, or consideration of future
needs. Only so much logic can be reasonably built into the
OS, and the rest is provided by a savvy user.

Is that why I've seen so many different informed and targeted decisions here?

It's similar to MANY different OS defaults, did you really
keep ALL of those defaults?

An apples & oranges smoke screen!

Actually most people never change the setting - and end up
with a much smaller pagefile than you have, never growing
even close to that size, and never have a problem.

For me, this isn't a contest to see who has the smallest pagefile. Would you
spend $1000 in labor to save on $900 in parts?

Since it is a discussion quite specifically about what size
to make it, the answer "as big as possible" really doesn't
address anything but the OS build-in limit, which they'd
assumed at the time would be far larger than necessary else
they'd have made it even bigger (max).

If they had done that, then a different decision would need to be made. But they
didn't!

Bob

 

[Shep©]

I was advised to make the maximum 1.5GB even though the initial was
512MB. I have more than ample disk space to that is not an issue.

The easiest way is to monitor your swapfile usage.I used Cacheman,
http://www.outertech.com/index.php?_charisma_page=product&id=7
I play some big graphics games and my swapfile was being used to
around 350 meg with 512 meg of ram installed.Set minimum to 400 meg
max to 765 meg after defragging pagefile drive in Safemode with
Auslogic defragger,
http://www.auslogics.com/disk-defrag/index.php
which I think also supports your O/S.

Now when I end games the swapfile doesn't have to resize as much and
saves time on the system coming back to normal.
As always with the swapfile/pagefile.sys it's the re sizing that eats
up time.

The reason for the use of Safemode and defrag once I determined the
correct settings for me is so that the space allocated to the swapfile
is not fragmented when formed.

HTH

 

[Citizen Bob]

With 780-odd MB memory allocated as you reported earlier
(but really, you should consider the peak value not the
momentary as you did), 512MB pagefile should work. You are
continually overlooking that there is no one generic answer
that fits all systems as well as actually looking at YOUR
system usage. Don't tell us what you have the pagefile set
to, tell us what your PEAK Commit Charge is.

Currently it is 817556 MB.

Remember that if you have too small a pagefile set, it won't
just slow down your use, you will see a warning message.
You could continue to have the system set to something like
a 512MB minimum (which minimizes fragmentation, contiguous
file if the disk space is available but even then, the file
itself may have fragmented access because that's how paging
works- only what's needed is read back), and a larger
maximum. You'd want your minimum large enough that your big
jobs don't exceed it, and the larger maximum is just a
failsafe should you do something very unusual. Keep in mind
that if you did such an unusual task and suddenly needed
another GB of virtual memory, you'd be sitting around for
ages waiting for the system to stop thrashing the HDD
swapping it all back and forth from disk to real memory.

I am currently using what you recommended earlier when I was having
the problem with corrupt NTFS volumes. That is 512MB/1.5GB. I have not
experienced any thrashing or messages.

You don't need to set initial same as max, just set initial
large enough that you don't "expect" it to ever be
exceeded... for example you could set a 1GB min and 2GB max.

I will try that because with this particular video conversion program,
I am now for the first time I can remember, substantially exceeding
512NB RAM size. Before it was at most 300MB.

If it ain't broke don't fix it.

But as you said, I should set the minimum to a number larger than the
Peak Commit Charge. How about a compromise: 1GB/1.5GB? If the
corruption problem returns I can always go back.


--

"All men seek to be enlightened. Religion is but the most ancient
and honorable way in which men have striven to make sense out of
God's universe. Scientists seek the lawfulness of events. It is
the task of religion to fit man into this lawfulness. Religion
must remain an outlet for people who say to themsleves, 'I am
not the kind of person I want to be'."
--Frank Herbert, "Dune"

 

[Citizen Bob]

The easiest way is to monitor your swapfile usage.I used Cacheman,
http://www.outertech.com/index.php?_charisma_page=product&id=7

I am running Win2K. Which one of the two versions should I use?


--

"All men seek to be enlightened. Religion is but the most ancient
and honorable way in which men have striven to make sense out of
God's universe. Scientists seek the lawfulness of events. It is
the task of religion to fit man into this lawfulness. Religion
must remain an outlet for people who say to themsleves, 'I am
not the kind of person I want to be'."
--Frank Herbert, "Dune"

 

[Shep©]

I am running Win2K. Which one of the two versions should I use?

It states it's compatible with win2k/XP/WinNT.

 

[Citizen Bob]

It states it's compatible with win2k/XP/WinNT.

I installed the XP version. The other one is a command line variant.

Here's the data it reports:

RAM: Peak Usage 500MB. Recall that I have 512MB RAM installed.

So it would appear that I am pushing the upper limit and could profit
from 1GB of RAM. However, this usage is caused by one particular video
application which I never use except this one instance. Ordinarily I
run at most 300MB RAM maximum.

Pafefile: Allocation 732MB. Recall that I have it set to 512MB/1.5GB.

It would seem that I am within the maximum pagefile setting. Maybe I
should set it to 1.0GB/1.5GB. I would rather not increase the maximum
beyond what it is now because in the past when it was larger I ran
into corrupt NTFS volume problems, presumably because of excess
fragmentation.



--

"All men seek to be enlightened. Religion is but the most ancient
and honorable way in which men have striven to make sense out of
God's universe. Scientists seek the lawfulness of events. It is
the task of religion to fit man into this lawfulness. Religion
must remain an outlet for people who say to themsleves, 'I am
not the kind of person I want to be'."
--Frank Herbert, "Dune"

 

[Rod Speed]

I installed the XP version. The other one is a command line variant.

Here's the data it reports:

RAM: Peak Usage 500MB. Recall that I have 512MB RAM installed.

So it would appear that I am pushing the upper limit and could profit
from 1GB of RAM. However, this usage is caused by one particular video
application which I never use except this one instance. Ordinarily I
run at most 300MB RAM maximum.

Pafefile: Allocation 732MB. Recall that I have it set to 512MB/1.5GB.

It would seem that I am within the maximum pagefile setting. Maybe I
should set it to 1.0GB/1.5GB. I would rather not increase the maximum
beyond what it is now because in the past when it was larger I ran
into corrupt NTFS volume problems, presumably because of excess
fragmentation.

It wont have anything to do with fragmentation.

 

[Shep©]

I installed the XP version. The other one is a command line variant.

Here's the data it reports:

RAM: Peak Usage 500MB. Recall that I have 512MB RAM installed.

So it would appear that I am pushing the upper limit and could profit
from 1GB of RAM. However, this usage is caused by one particular video
application which I never use except this one instance. Ordinarily I
run at most 300MB RAM maximum.

Pafefile: Allocation 732MB. Recall that I have it set to 512MB/1.5GB.

It would seem that I am within the maximum pagefile setting. Maybe I
should set it to 1.0GB/1.5GB. I would rather not increase the maximum
beyond what it is now because in the past when it was larger I ran
into corrupt NTFS volume problems, presumably because of excess
fragmentation.

I'd set it at say 600 meg min and 765 max
Unless you use another ram hungry program sometime that should do you
fine and if you did run out you know what to do anyway e.g just
increase the max size a bit after monitoring it again
PS
I've just added another 512 meg of RAM to bring mine to 1 gig but left
my 300/765 as is.My big games now shutdown instantly as the
pagefile.sys is barely if ever used however some pagefile.sys is
always used if there as it's tied to the RAM allocation but also in
Cacheman I disable the,"Executive Paging" as that forces windows to
use all the physical ram before going to the swapfile/pagefile.sys

HTH

 

[Citizen Bob]

in Cacheman I disable the,"Executive Paging" as that forces windows to
use all the physical ram before going to the swapfile/pagefile.sys

Where is this "executive paging". I looked in CachemanXP but did not
find it.


--

"All men seek to be enlightened. Religion is but the most ancient
and honorable way in which men have striven to make sense out of
God's universe. Scientists seek the lawfulness of events. It is
the task of religion to fit man into this lawfulness. Religion
must remain an outlet for people who say to themsleves, 'I am
not the kind of person I want to be'."
--Frank Herbert, "Dune"

 

[Shep©]

Where is this "executive paging". I looked in CachemanXP but did not
find it.

Tweaks Tab.Untick it.

You must understand that no matter what MS say there are core
fundamental rules that never change in their O/S.
I've been tweaking this crap since win 3.11 let alone DOS.MS always
allow for the none intervention of user interaction,which is fine.It
doesn't matter if it's Win95/Win98/ME/Win2K/NT 1/2/3/4/WinXP.

I've said it before and I'll say it again.It was only a
genius,"Slight-of-Hand" by BG that led us to these badly programmed
operating system running on out-dated before they were born updated
2086 processors.

NAND
Only gamers know what PCs can really do and can't do ;-)

 

[Citizen Bob]

Tweaks Tab.Untick it.

Found it.

You must understand that no matter what MS say there are core
fundamental rules that never change in their O/S.
I've been tweaking this crap since win 3.11 let alone DOS.MS always
allow for the none intervention of user interaction,which is fine.It
doesn't matter if it's Win95/Win98/ME/Win2K/NT 1/2/3/4/WinXP.

How does this Executive Paging help performance? What is the tradeoff?

I've said it before and I'll say it again.It was only a
genius,"Slight-of-Hand" by BG that led us to these badly programmed
operating system running on out-dated before they were born updated
2086 processors.

I would like to catch the idiot who didn't make DOS re-entrant.
Everyone worth a crap who writes real time embedded operating systems
knows to make them re-entrant.

NAND
Only gamers know what PCs can really do and can't do ;-)

They have a way to go - they haven't had to use liquid nitrogen
coolers yet.


--

"All men seek to be enlightened. Religion is but the most ancient
and honorable way in which men have striven to make sense out of
God's universe. Scientists seek the lawfulness of events. It is
the task of religion to fit man into this lawfulness. Religion
must remain an outlet for people who say to themsleves, 'I am
not the kind of person I want to be'."
--Frank Herbert, "Dune"

 

[kony]

How does this Executive Paging help performance? What is the tradeoff?

Some driver and kernel code that was flagged as pageable
won't be, thus keeping in main memory instead of having OS
page it out more aggressively. The result is you may not
have the wait times to page this back into main memory but
since it uses more memory you must have more memory
installed.

Considering your prior memory utilization figures, I would
not disable this until after you had increased your
installed memory past 512MB you presently have... doing that
should be priority #1. Even so you can try it both ways
now, keeing in mind what changes you've made so they're
easily reverted back to the prior state if necessary.

 

[Citizen Bob]

Some driver and kernel code that was flagged as pageable
won't be, thus keeping in main memory instead of having OS
page it out more aggressively.

CachemanXP reports that memory usage is down to 380MB and the pagefile
is down to 80MB. Apparently I need to reboot more often to clear the
memory leaks.

since it uses more memory you must have more memory
installed.

I would if I planned to run this video application a lot. But this is
hopefully a one-time event so I will sweat it out. Normally I run
under 300MB fully loaded.

On another matter in a thread I started about temperature and CPU
throttling, I just cleaned the heat sink and case and now the CPU temp
as measured by MBM5 Winbond Diode 1 shows a temp of 45C compared to
58C before cleaning. Therefore the CPU is running at full speed which
I can tell by that application which is hogging the CPU.

--

"All men seek to be enlightened. Religion is but the most ancient
and honorable way in which men have striven to make sense out of
God's universe. Scientists seek the lawfulness of events. It is
the task of religion to fit man into this lawfulness. Religion
must remain an outlet for people who say to themsleves, 'I am
not the kind of person I want to be'."
--Frank Herbert, "Dune"

 

[kony]

CachemanXP reports that memory usage is down to 380MB and the pagefile
is down to 80MB. Apparently I need to reboot more often to clear the
memory leaks.

I'm not about to guess what might be wrong with an aged
Win2k upgrade kept several years, but in general it is not
necessary to reboot but on rare occasions. You should be
able to see in Task Manager, Processes tab, what's eating up
memory. If you have particular apps leaking you might
consider replacing them. If your prior figures were with
more apps running, of course they'll consume more, but it is
not the just-rebooted figures you need to be concerned
about, it is the peak figures you cause when running the
larger jobs... and IF you have leaks and don't want to
replace those apps, OR if you don't have the leaks and it's
just a regular usage of more memory between reboots, you can
still factor for the peak figures.

I would if I planned to run this video application a lot. But this is
hopefully a one-time event so I will sweat it out. Normally I run
under 300MB fully loaded.

That would make 512MB closer to appropriate, though I'd
still want more margin, remembering that additional memory
can be used as a filecache. At this moment the box I'm on
is running Win2k and CachemanXP shows 271MB used, but the
Task Manager system cache is around 760MB... a LOT of things
I use don't have to be read from HDD. If I were to reboot
as you are doing (more frequently), I lose the cached files
and there's several hundred MB that has to be reread from
HDD again if the usage pattern was essentially similar. Not
a problem per se, but it is lower performance to keep losing
caching and re-read it from HDD.

On another matter in a thread I started about temperature and CPU
throttling, I just cleaned the heat sink and case and now the CPU temp
as measured by MBM5 Winbond Diode 1 shows a temp of 45C compared to
58C before cleaning. Therefore the CPU is running at full speed which
I can tell by that application which is hogging the CPU.

That's a nice drop, you ought to write "clean heatsink" on
your calendar.

 

[Citizen Bob]

I'm not about to guess what might be wrong with an aged
Win2k upgrade kept several years, but in general it is not
necessary to reboot but on rare occasions. You should be
able to see in Task Manager, Processes tab, what's eating up
memory.

CachemanXP is running high on the list and has 434,000 page faults as
well.

That would make 512MB closer to appropriate, though I'd
still want more margin, remembering that additional memory
can be used as a filecache. At this moment the box I'm on
is running Win2k

That's a nice drop, you ought to write "clean heatsink" on
your calendar.

It's on there but only every 6 months, I just moved it to every 3
months. My son has a 3.2GHz P4 Prescott chip and the retail box cooler
was a big joke, especially when he loaded the machine up with video
applications. So he got a full copper Zalman - the biggest he could
find - and runs it without any noise reduction. And he has to clean it
often or his CPU will start heating up.

I would like to get one of those hood assemblies that look like the
top of a square air cleaner off an car with a carburettor. It screws
onto the top of the fan and a duct runs to the back of the case where
it routes the air flow from outside. I would put a fan filter in front
of the inlet so I could easily clean the dust off periodically from
the outside.

Not only would the air going to the CPU be dirt-free but it would also
be cooler than the air inside the case. In fact I wonder why more case
manufacturers don't providing this because I can't be the only one
whose heat sink clogs with dust. But then look how long it took for
the automotive industry to adopt the 12 volt battery (ever try to
start a car that sat outside overnight in 20F weather with a 6 volt
battery?).


--

"All men seek to be enlightened. Religion is but the most ancient
and honorable way in which men have striven to make sense out of
God's universe. Scientists seek the lawfulness of events. It is
the task of religion to fit man into this lawfulness. Religion
must remain an outlet for people who say to themsleves, 'I am
not the kind of person I want to be'."
--Frank Herbert, "Dune"

 

[kony]

It's on there but only every 6 months, I just moved it to every 3
months. My son has a 3.2GHz P4 Prescott chip and the retail box cooler
was a big joke, especially when he loaded the machine up with video
applications. So he got a full copper Zalman - the biggest he could
find - and runs it without any noise reduction. And he has to clean it
often or his CPU will start heating up.

It can't be THAT hot in TX right now!

Get a room air cleaner... you won't have to dust off
yourself as often, either.

On many of the stamped-out sheeting types of heatsinks, you
can reduce the amount of dust (rate of accumulation) by
taking very fine grit sandpaper and sanding down the leading
stamped edge (facing fan if a pusher fan). I'd go with the
room air cleaner first, but it all adds up.

I would like to get one of those hood assemblies that look like the
top of a square air cleaner off an car with a carburettor. It screws
onto the top of the fan and a duct runs to the back of the case where
it routes the air flow from outside.

Assuming your heatsink fan is blowing into the 'sink, that
is a bad idea. It will draw the heated PSU exhaust back
into the system creating a so-called "short loop" that
recirculates the heated air. If you need airflow changes,
get a higher flow rate into the bottom front of the case.
It will cool all parts better including CPU, while the short
loop robs other parts of some airflow. You can't just flip
the fan over with same (otherwise) results if the heatsink
isn't designed with that in mind, the gains from the duct
would be offset by the losses from flipped fan... and in
most of Intel's sinks you can't even flip the fan as it's
integrated into a frame-holder.

I would put a fan filter in front
of the inlet so I could easily clean the dust off periodically from
the outside.

Put the filter in the room, cleans everything.
If that's not enough, put a filter on the case front intake,
at least it then cleans as much as possible in the case
(with a front intake fan to reduce any air intake though
non-filtered cracks, holes, seams, etc.

Not only would the air going to the CPU be dirt-free but it would also
be cooler than the air inside the case.

See above, plus merely having a filter dense enough to be
very effective will substantially cut down on airflow rate
unless the filter area is MUCH larger than the fan intake
area, plus the duct alone (without filter installed) also
cuts down on airflow rate. Axial fans are poor at
maintaining pressure in a ducted scenario. Squirrel cage
fans do this better, but are usually quite a bit noisier
without any need currently, as mentioned above it can't be
THAT hot in TX right now.

In fact I wonder why more case
manufacturers don't providing this because I can't be the only one
whose heat sink clogs with dust.

1) Cost more than not having it.
2) Aftermarket sales of accesssory filter panel = more
profit
3) Most people don't need it, would just clean the 'sink
every few months _if_ that often, and over life of system
that's not THAT many cleaning cycles when considering the
extra effort to install and test the filtered-duct, AND that
the filter panel on the duct has to be cleaned too... maybe
even more frequently than without the filter as not all dust
gets stuck in the heatsink without a filter, a substantial %
is exhausted out of the sink... while such particles would
probably be trapped in the filter.

IMO, if you have the heatsink intake filtered you should
FIRST filter the whole case, else the heatsink filter gets
clogged that much quicker and you have to open up the case
to swap that filter in any normal arrangement unless you had
the undesirable short-loop intaking the heated PSU exhaust
as mentioned above.

But then look how long it took for
the automotive industry to adopt the 12 volt battery (ever try to
start a car that sat outside overnight in 20F weather with a 6 volt
battery?).

I dunno what to say, systems here that have sat over a year
inbetween heatsink cleanings (overclocked Athlon XPs... not
exactly cool running) aren't running as hot as yours was.
Granted, they aren't primary use systems anymore so it's not
24/7 operation of fans which reduces dust accumulation but
on the other hand I suspect a larger reason is they had
better than stock heatsinks on them... a good heatsink
allows lower fan RPM, with the obvious reduction in airflow
as a result, not only does that account for lower dust
accumlation directly, but the lower the velocity of the air
the more dust particles settle out from gravity instead of
making it up to the 'sink. Plus, very quiet and long lived
fans are nice too... it would be unusual if any of these
fans fail within a decade.

 

[Citizen Bob]

It can't be THAT hot in TX right now!

It was over 80F in Houston. But now it is cooling off.

Get a room air cleaner... you won't have to dust off
yourself as often, either.

LOL. What is a room air cleaner?

On many of the stamped-out sheeting types of heatsinks, you
can reduce the amount of dust (rate of accumulation) by
taking very fine grit sandpaper and sanding down the leading
stamped edge (facing fan if a pusher fan).

The fins on this Intel retail box heat sink are extemely close spaced.

Everything is just fine now. The highest temp has been 45C according
to MBM5. That's compared to 58C under the same conditions.

I set my calendar to every two months. It's no big deal to open the
case and squirt some canned gas onto the heat sink. However next time
I am down at Directron I am going to ask Juan to show me that
hood-duct gizmo and compare to measurements I have taken off my
geometry. Add an external fan filter in front of the inlet and no more
dust in the heat sink - and much cooler air blowing over it. Then I
can easily clean the dust off the filter from the outside.

I'd go with the
room air cleaner first, but it all adds up.

I am afraid that would only delay the dust buildup. The
hood-duct-filter would completely eliminate it.

Assuming your heatsink fan is blowing into the 'sink,

The fan is blowing down onto the heat sink.

that is a bad idea. It will draw the heated PSU exhaust back
into the system creating a so-called "short loop" that
recirculates the heated air.

The hood bolts down to the fan so there is no opening for the air to
recirculate. The duct goes to the back of the case so any air getting
to the heat sink must come from the outside.

If you need airflow changes,
get a higher flow rate into the bottom front of the case.
It will cool all parts better including CPU, while the short
loop robs other parts of some airflow. You can't just flip
the fan over with same (otherwise) results if the heatsink
isn't designed with that in mind, the gains from the duct
would be offset by the losses from flipped fan... and in
most of Intel's sinks you can't even flip the fan as it's
integrated into a frame-holder.

I believe all this is moot based on my comment above. But I will keep
it in mind.

put a filter on the case front intake,

OK, so the front fan needs to be blowing into the case and should have
a filter. I can try that quite easily.

How should the side fan be blowing? How about the rear fan? I believe
the PSU fan is blowing out so it would seem that if the front is
blowing in, the extra rear fan I have should be blowing out. But what
about the side fan?

BTW this case has airholes in the side and they get dust on them
indicating that air is being sucked in. I suppose I should seal that
so no air can circulate thru it or maybe put some kind of filter on
it.

as mentioned above it can't be THAT hot in TX right now.

It makes no difference what the temp is outside - we have air
conditioning.

AND that the filter panel on the duct has to be cleaned too.

That would be easy since it would be mounted on the outside of the
case where the duct is sucking air.

I dunno what to say, systems here that have sat over a year
inbetween heatsink cleanings (overclocked Athlon XPs... not
exactly cool running) aren't running as hot as yours was.
Granted, they aren't primary use systems anymore so it's not
24/7 operation of fans which reduces dust accumulation but
on the other hand I suspect a larger reason is they had
better than stock heatsinks on them... a good heatsink
allows lower fan RPM, with the obvious reduction in airflow
as a result, not only does that account for lower dust
accumlation directly, but the lower the velocity of the air
the more dust particles settle out from gravity instead of
making it up to the 'sink. Plus, very quiet and long lived
fans are nice too... it would be unusual if any of these
fans fail within a decade.

I am going to try the filter in front idea and try to seal the case. I
suspect the correct direction for the flow of the side fan would be to
the outside, which is the way it is blowing now.

Come to think of it, it's kinda dumb to have air holes next to that
side fan. Talk about a "short loop". I need to seal them and then
maybe more air will come in the front.

 

[kony]

It was over 80F in Houston. But now it is cooling off.

So yes, it's not THAT hot.

LOL. What is a room air cleaner?

Well if you're going to have a filter somewhere, it could
have maximum effect on everything...

I am afraid that would only delay the dust buildup. The
hood-duct-filter would completely eliminate it.

No, it will not. A filter medium effective enough to
eliminate dust rather than slow down accumulation will
substantially decrease airflow, as much or more than the
dust level you had just before you cleaned it. You will
also have to clean out the filter panel, so it's just a
similar maintenance task, different buildup location. At
least if you filter the entire chassis intake you end up
with a clean system (if adequately sealed around the filter
and case is positively pressurized). This also presumes a
larger filter area more easily implemented on the case front
than a duct intake.

People have already gone where you think of going, done what
you are speculating. If you have to learn it all for
yourself from the start, so be it, but keep in mind what
I've written when you encounter these issues.

The hood bolts down to the fan so there is no opening for the air to
recirculate. The duct goes to the back of the case so any air getting
to the heat sink must come from the outside.

It will recirculate, it will pull the air from the rear
exhaust back in through the duct. If the duct is in the
side panel instead of the rear, this preheated air will at
least mix with the room air far more and be cooler, though
not at the 80F room ambient. There is no need though, and
drawbacks to the plan.

How should the side fan be blowing?

Nowhere. You have zero need for a side fan and it will
probably have negative impact. Blowing in it woud require
more front fan speed and noise to maintain same flow rate
over the drive rack. Blowing out it goes against the CPU
fan and robs the PSU of some flow rate. Side fans were only
an intel attempt to get away with shipping cheaper heatsinks
for P4 Prescott, or case manufacturers looking to appeal to
the modder crowd by adding a line item feature. They didn't
care that your CPU (as with most moderate heat CPUs) don't
need it.

How about the rear fan? I believe
the PSU fan is blowing out so it would seem that if the front is
blowing in, the extra rear fan I have should be blowing out. But what
about the side fan?

There is no "extra" rear fan. The PSU exhaust is required.
The rear chassis fan under the PSU is also required. They
are not optional for a well designed system, doing without
either will have significant drawbacks of one sort or
another.

If you want a filtered intake, the next fans are on the
chassis front, behind the sealed filter intake panel.

Your system has no need for a side panel fan. If you
install it you will have to change the optimal arrangement
of front and rear fans to offset it. A Celeron 2.4 needs no
elaborate measures.

BTW this case has airholes in the side and they get dust on them
indicating that air is being sucked in. I suppose I should seal that
so no air can circulate thru it or maybe put some kind of filter on
it.

You should install front case fans behind a filter panel.
Change the case pressurization from slightly negative
(sucking in through holes and gaps) to slightly positive
(blowing out holes and gaps). All intake in the front is
filtered, one easily accessible removable filter panel.
That is if you won't just clean out the heatsink as most end
up doing... but if 3 months is all it take for dust buildup
I think the room air cleaner is a nice step, your lungs may
thank you as well.