CPU Temp: 13C

[Iain Dingsdale]

Anything like this for a none 64 Bit Xp chip, I'd love to get mine
down a bit in the shuttle ?

Nope, its a feature of the processor/mobo

 

[kony]

Anything like this for a none 64 Bit Xp chip, I'd love to get mine
down a bit in the shuttle ?

Athlon XP can use HALT idle cooling with an OS that supports
ACPI. Some boards allow enabling it in the bios (might be
called "S2K" but others require software to program chipset
registers (like WPCREdit, WPCRSet). Google search for it
including phrases like "<name of your shuttle board's
chipset> HLT cooling"

 

[Mxsmanic]

Then what solution to you propose?

Fans, and several of them. On my homebuilt server, there are seven
fans. Three in the power supply (supplied with the PSU), one on the
processor (supplied with the processor), plus three 8-cm fans I added:
two in the front blowing air into the case and drawing it through a
filter, arranged to blow past the disk drives, and one on the side of
the case over the processor, blowing air down over the processor and
drawing it through a filter. So three fans take air in, and one takes
it out, which should create slight positive pressure (the remaining air
just flows out the top and back of the case).

I figure that the machine should still be safe with any single fan
failure, and possibly with two fan failures. The Intel processor
protects itself if the fan fails; the rest of the machine has multiple
fans so that the PSU and MB should be kept acceptably cool even if one
of them drops out.

Fans have far shorter lifespan than the ICs.

That's why I have more fans than I need. If one fails, the machine
won't cook (unlike its predecessor, whose AMD processor did exactly that
when the CPU fan failed). I'm very aware of the sound they make and any
change alerts me; I also explicitly check to see if air is flowing
through them periodically.

Needless to say, I don't care about noise. I actually like hearing
components running, because it makes it possible to detect when they
stop running. Over all three machines, I have some 13 fans running.

 

[kony]

Fans, and several of them. On my homebuilt server, there are seven
fans. Three in the power supply (supplied with the PSU), one on the
processor (supplied with the processor), plus three 8-cm fans I added:
two in the front blowing air into the case and drawing it through a
filter, arranged to blow past the disk drives, and one on the side of
the case over the processor, blowing air down over the processor and
drawing it through a filter. So three fans take air in, and one takes
it out, which should create slight positive pressure (the remaining air
just flows out the top and back of the case).

Ok, if the heat (or RPM reduction for reduced noise)
warrants that many fans, but those fans will not reduce the
northbridge temp much, if we're going under the presumption
that the case was already properly ventilated. Even with
the fans, most likely each and every fan would fail long
before the northbridge. Perhaps not while machine is newer
and valuable enough to be well-kept, but given enough time,
it's not likely anyone would find use or value in the system
by the time it's THAT old.

I figure that the machine should still be safe with any single fan
failure, and possibly with two fan failures. The Intel processor
protects itself if the fan fails; the rest of the machine has multiple
fans so that the PSU and MB should be kept acceptably cool even if one
of them drops out.

That is a good idea, so long as it suits your needs. I buy
fans by the case so I too can do this at reasonable cost
without sacrificing fan quality, but I regularly see people
paying upwards of $10 per fan - buying fewer because they
have not the budget or inclination to add so many fans.

That's why I have more fans than I need. If one fails, the machine
won't cook (unlike its predecessor, whose AMD processor did exactly that
when the CPU fan failed). I'm very aware of the sound they make and any
change alerts me; I also explicitly check to see if air is flowing
through them periodically.

Needless to say, I don't care about noise. I actually like hearing
components running, because it makes it possible to detect when they
stop running. Over all three machines, I have some 13 fans running.

This may be where we differ somewhat, I don't have a single
fan running at over (roughly) 3000 RPM, and only then on a
couple of overclocked gaming video cards where there is more
of a space constraint. Usually I'll want an optimized fan
RPM such that it reduces dust builtup and bearing wear even
on remote systems outside of audible range. The noisiest
things in some/most of my audible systems are WD
ball-bearing based hard drives, but fortunately WD has now
switched over to all fluid bearing models AFAIK.

 

[ASI Industries]

Hi Kony, sorry i forgot to mention when the chipset got to 65%C, (3 hours
playing heavy 3d games), the mainboard heat protection system cut in and
switched the system off, even though i disabled the heat sensors in the
bios!

The board is not in a case, just sitting on a worktop, i have replaced the
thermal compound, that made no difference, added a chipset fan, that
doubled the time before it crashes (6 hours playing heavy 3d games),
replaced the stock hs with a Zalman Zm-nb47j hs (its 4x the size of stock
hs and wieghs 54g) it reduced the temperature overall by 6%C.

Relocated the 12cm fan to cover both cpu hs and Zalman chipset hs brings
the
temperature down around 22 - 26%C and the stock hs brings the temperature
down around 28 - 35%C, currently using Zalman Cnps6000b-cu cpu hsf.

65C is not "overheating", it is within the specs for the chip.

Can you provide any urls for the specifications of the Via Kt600
northbridge chipset?

 

[Mxsmanic]

Ok, if the heat (or RPM reduction for reduced noise)
warrants that many fans, but those fans will not reduce the
northbridge temp much, if we're going under the presumption
that the case was already properly ventilated. Even with
the fans, most likely each and every fan would fail long
before the northbridge. Perhaps not while machine is newer
and valuable enough to be well-kept, but given enough time,
it's not likely anyone would find use or value in the system
by the time it's THAT old.

My idea is that any single fan failure will not damage the machine, even
a CPU or PSU fan failure. If you have a configuration in which the
failure of a certain fan will damage the system, then you risk your
entire investment on the reliability of a component that will inevitably
fail.

And systems can retain use and value indefinitely. If they do the job
they are intended to do, there's no reason why they can't remain in use
for twenty years or more. I have a very well designed and built HP
Vectra on my desk that has been running continuously for eight years
without a hiccup.

That is a good idea, so long as it suits your needs. I buy
fans by the case so I too can do this at reasonable cost
without sacrificing fan quality, but I regularly see people
paying upwards of $10 per fan - buying fewer because they
have not the budget or inclination to add so many fans.

Ten dollars isn't much to save a $200 motherboard or a $200
microprocessor (or multiple components).

This may be where we differ somewhat, I don't have a single
fan running at over (roughly) 3000 RPM, and only then on a
couple of overclocked gaming video cards where there is more
of a space constraint. Usually I'll want an optimized fan
RPM such that it reduces dust builtup and bearing wear even
on remote systems outside of audible range.

Dust builds up in areas where the air flow is slowing, not in areas of
high flow. The places that receive the strongest air currents are never
the places that contain the dust--it's always hiding in crannies where
the air slows down.

Also, if you put filters on the incoming air, you can greatly reduce
dust (although I'm still looking for ideal filters).

The noisiest
things in some/most of my audible systems are WD
ball-bearing based hard drives, but fortunately WD has now
switched over to all fluid bearing models AFAIK.

Will those bearings last as long as ball bearings?

 

[Mxsmanic]

Hi Kony, sorry i forgot to mention when the chipset got to 65%C, (3 hours

playing heavy 3d games) ...

The entire chipset on the MB, or just the processor?

 

[Kevin]

I'm seeing the same behavior on my MSI K8T Neo2-FIR mobo. The board
shipped with BIOS version 3.0. The CPU and system temperatures reported
by the BIOS in PC Health Status and by Core Center were in agreement at
around 39 and 34 deg. C, respectively, when the CPU is idle. The CPU
temp. under 100% CPU load was 47 deg. C.

Then I upgraded to BIOS 3.2 which has a fix for "BIOS reporting
incorrect CPU temperature". After that, the CPU temperature reported by
Core Center varies wildly from 11 to 29 deg. C while the system
temperature is around 30 deg. C. Meanwhile, PC Health Status in the
BIOS reports 34 for CPU and 30 for system. The ambient air temperature
outside the PC is 19 deg. C.

I upgraded to BIOS 3.3, and the behavior is the same 3.2. I have been
e-mailing MSI Tech Support, but I had a hard time convincing the rep.
that there is a bug in the BIOS or Core Center as demonstrated by the
discrepancy in CPU temp. reported by the two. After many attempts, he
sent me a newer version of the BIOS today. I'll try it tonight, but I'm
not holding my breath.

Kevin

I have just updated the drivers for my MSI K8T Neo mobo, updated my BIOS and
updated the MSI "CoreCenter" monitoring software.

CoreCenter and the BIOS both give me temperature readings for my CPU. Before
the updates the BIOS and CoreCenter reported my CPU as having temperatures
of between 39C and 51C depending on the room temp and what I was doing. This
seemed a fairly normal temperature and I was satisfied with it.

However since the updates my CPU temperature has went way down according to
this software. When I go into the BIOS it's reporting the temp as 19C. When
I'm sitting idle in Windows XP - as I am now - it's sitting at 13C! I tested
it by encoding a large MP3 and it went up to 27C after 5 mins of encoding.

I have CoolNQuiet (http://www.amdboard.com/coolnquiet.html) switched on and
an Athlon-64 3400 Clawhammer (2.2GHz) with a Arctic Cooling Silencer 64 fan.
Can this be real? I'm concerned that the temps are being

misreported...

 

[kony]

My idea is that any single fan failure will not damage the machine, even
a CPU or PSU fan failure. If you have a configuration in which the
failure of a certain fan will damage the system, then you risk your
entire investment on the reliability of a component that will inevitably
fail.

Possibly but many components that would be susceptible to
this also have a thermal shutdown mechanism.

And systems can retain use and value indefinitely. If they do the job
they are intended to do, there's no reason why they can't remain in use
for twenty years or more. I have a very well designed and built HP
Vectra on my desk that has been running continuously for eight years
without a hiccup.

That's not really true, they become a liability after about
5 years due to mechanical connections and capacitor rot.
Additionally I don't expect today's systems to last nearly
as long as (say an old 486 did). Even with quite a few
fans the mechanical and cap issues may be higher.

Ten dollars isn't much to save a $200 motherboard or a $200
microprocessor (or multiple components).

But per your example, it'd be 7 x $10. I have a system or
two or three (lost count) with 7 or more fans but dont' see
this too often due to case manufacturers limiting fan
numbers as a cost-reduction. Many people would not pay $70
more for same case with more fans already in it, even if it
saved them the cost of adding the fans. While it can be a
good idea, it may not be so necessary with good fans and
thermal managment. It is pretty rare for a good
sleeve-bearing fan to fail within the first few years (like
Panaflo) unless driven at very high RPM. After the first
few years, each person may not place same value on their
systems as you do, because again they see some that aren't
so expensive with fewer fans.

Dust builds up in areas where the air flow is slowing, not in areas of
high flow. The places that receive the strongest air currents are never
the places that contain the dust--it's always hiding in crannies where
the air slows down.

While that's a nice theory, in practice the dust builds up
everywhere. Evidence includes dust buildup on the fan
blades themselves, certainly not an area of slowing air
flow. Overall a higher flow through the chassis will always
result in more dust accumulation.

Also, if you put filters on the incoming air, you can greatly reduce
dust (although I'm still looking for ideal filters).

Me too, ideal filters are restrictive enough on flow that
the last case I put one in was using a filter area of
roughly 3-4X the area of the intake fans. Then keeping a
positive pressurization on the case to keep all intake
filtered means either louder intake fans, more frequent
filter exchange, or exhaust fan reduction at certain
interval (or sealing up ever other nook and cranny on the
case).

Will those bearings last as long as ball bearings?

Only time will tell? They seem to imply they're more
reliable, but when would a manufacturer not claim this about
product changes?

 

[kony]

Hi Kony, sorry i forgot to mention when the chipset got to 65%C, (3 hours
playing heavy 3d games), the mainboard heat protection system cut in and
switched the system off, even though i disabled the heat sensors in the
bios!

Hmm, I didn't remember there being a way to disable the heat
sensors. Perhaps you mean the thermal shutoff for the CPU?
Are you running the PCProbe software? Maybe that is the
reason for the shutdown?

The board is not in a case, just sitting on a worktop,

Is this how it was all along? Perhaps that 's it, your CPU
heatsink is simply recirculating same pre-heated air around
the northbridge.

i have replaced the
thermal compound, that made no difference, added a chipset fan, that
doubled the time before it crashes (6 hours playing heavy 3d games),
replaced the stock hs with a Zalman Zm-nb47j hs (its 4x the size of stock
hs and wieghs 54g) it reduced the temperature overall by 6%C.

It has now dawned on me that your board didn't have the tall
silver Asus heatsink but rather the shorther dark-colored
one? Mine has the silver one and has done fine with it...
but it doesn't explain why you're not seeing good results
with the Zalman.

Relocated the 12cm fan to cover both cpu hs and Zalman chipset hs brings
the
temperature down around 22 - 26%C and the stock hs brings the temperature
down around 28 - 35%C, currently using Zalman Cnps6000b-cu cpu hsf.

Can you provide any urls for the specifications of the Via Kt600
northbridge chipset?

No I don't see one available to the public on Via's website,
but there aren't any that have lower than 80-85C, usually
higher, that I recall. Granted that many not rule out
instability but from time to time I'll drop in on some
motherboard forums and had not noticed any significant rate
of these boards overheating... but then I wasn't looking for
such data since mine isn't overheating? Right now PCProbe
reports it @ 36C. Upon feeling the heatsink that figure
seems close enough to accurate, ie- it's certainly not at ~
65C.

 

[Mxsmanic]

That's not really true, they become a liability after about
5 years due to mechanical connections and capacitor rot.

I've never seen an MB fail in the absence of something that damaged it
(heat, water, etc.). While components do fail, most people never wait
for it to happen, and methodically replace machines while they are still
in perfect working order.

Additionally I don't expect today's systems to last nearly
as long as (say an old 486 did).

Why not?

But per your example, it'd be 7 x $10.

That's $70, compared to over $850 for the entire system.

It is pretty rare for a good
sleeve-bearing fan to fail within the first few years (like
Panaflo) unless driven at very high RPM.

After being burned by sleeve bearings (so to speak), I now use only
ball-bearing fans.

While that's a nice theory, in practice the dust builds up
everywhere.

It's what I've observed from experience, not theory. The air has to
move slowly for the dust to settle out, so the places where there is a
slowing of the air or similar turbulence are where the dust collects.

Evidence includes dust buildup on the fan
blades themselves, certainly not an area of slowing air
flow.

But it is an area of slowing air flow, along trailing edges that
generate vortices.

Overall a higher flow through the chassis will always
result in more dust accumulation.

Whence the interest of using filters. Now if only I could find good
filters.

Only time will tell? They seem to imply they're more
reliable, but when would a manufacturer not claim this about
product changes?

I'll wait and see, then. I know ball bearings last a long time. I'll
let someone else be the pioneer with new types of bearings.

 

[ASI Industries]

Hi Mxsmanic, no just the Via KT600 NB SHS.

Anyway i just borrowed a multi-temperature data logger, the unit measures
16 sensors at the same time.

Two test were done one with the SHS and the othe with the ZHS, with 1 hour
cool-off period between tests.

Sensor: Idle: Avg: Fail: OBI: Avg: OBF:
----------------------------------------------------------------------------
----------------
Cpu core: 36 50 65 26 35 28
Cpu HS: 26 43 60
KT600 NB core: 38 62 86 27 30 32
KT600 NB SHS: 28 47 76
KT600 NB ZHS: 24 47 76
KT600 SB core: 22 22 22
Vga core: 45 50 52 45 48 52
Vga HS: 40 44 48
Vga memory: 35 40 45 39 42 42
HDD: 16 16 16 16 16 16
Memory: 22 22 22
Room: 22 22 22

Notes: The tests began roughly 30 secs after Win XP loaded.
Everest v2.00.254 for onboard sensors for (cpu, system (usually NB) and
HDD).
Everest v2.00.254 Download http://tinyurl.com/4trst
Asus Smartdoctor v4.60 for onboard onboard sensors for (GPU and memory).
HDD is cooled via PSU.
SHS = Stock HeatSink, ZHS = Zalman ZM-NB47J HeatSink
NB = North Bridge, SB = South Bridge
OBI = OnBoard Idle, OBF = OnBoard Fail

System details:
ASUSTEK A7V600-X, Rev 1.0, Bios 1007. http://tinyurl.com/6v3xd
Amd XP3200+ Barton (2200mhz 11x 200).
Asus V9999GT /TD128 (Nvidia 6800GT) driver 71.24.
http://tinyurl.com/6gvkc
1024mb 2x Corsair value ddr400 512mb, matched pair Cas 2.5:3-3-8.
Zalman Cnps6000-cu.
http://tinyurl.com/4fxu4
Zalman ZM-NB47J.
http://tinyurl.com/2x2qf
Windows XP.SP1 DX9.0c
Qtec 400wt dual fan psu (+3+5=180 / +12=192 / 372).
+3.3v = 3.34, +5v = 5.12, +12v = 12.12.
Floppy, Nec ND3500A dual dvd rw, Samsung SP0411N 40gb HDD pata 100.

 

[kony]

I've never seen an MB fail in the absence of something that damaged it
(heat, water, etc.). While components do fail, most people never wait
for it to happen, and methodically replace machines while they are still
in perfect working order.

Well that has to be weighed against all common failures
though... To a certain extent we could always attribute
failure to 'something', whether that be electrical surge,
manufacturing defect, oxide buildup, epoxy deterioration,
heat, water, physical shock, etc, or just gradual
evaporation of capacitor electrolyte. Often I do see boards
that have failed for no obvious reason, and for cost and
time constraints it is more feasable to replace rather than
diagnose.

Why not?

Power requirements elevated relative to the actual capacity
of the average power supply, higher current swings, and with
the relatively similar PCB real-estate, a limitation in
spacd and expense is reached where the margin for capacitors
is lower. Even a gale force wind will not keep a cap as
cool on a (for example, Prescott 3.6GHz) VRM circuit as on
the typical (Celeron 600) board. All capacitor
manufacturers are in agreement that higher cap temp will
directly effect their viable lifespan. In some situations I
feel it would be preferrible to have socketed caps else at
least a higher-tiered board for "PC" rather than server
which has more optimal cap such as organic compositions.

That's $70, compared to over $850 for the entire system.

True, I agree that extra expense is worthwhile, but since I
do buy fans in bulk I can't begin to rationalize it for
someone else. Pure performance is too often emphasized
rather than lifespan, such that a customer seeing two
systems with *same* price-point may easily choose one that
is quieter, faster CPU, and hotter running, rather than one
which is cooler but slower CPU. With every $70 there is
some kind of tradeoff to be made in a modular PC.

After being burned by sleeve bearings (so to speak), I now use only
ball-bearing fans.

Were they good makes though? I would never put
median-to-low quality sleeve bearings in a system and remove
(replace) such fans from cases with them already installed,
but find Panaflo and Papst in particular to be very good
quality and lifespans exceeding 10 years in most cases,
perhaps far longer if a lubrication interval is maintained
after the first 5-8 years. It would be a personal
preference, a drop of oil every 3 years or so seems like a
fair trade for great reduction in higher frequency bearing
noise, moreso in often-used, "PC" systems rather than remote
systems outside of audible range.

Junk sleeve-bearing fans are a real liability. Placing one
on a power supply exhaust, video card or CPU heatsink is
quite often the primary reason I see old systems replaced.
Not that there aren't other reasons some will replace an old
system, but in practice I see people who would've kept using
a system had it not been for such a fan failure, then faced
with a (power supply loss or maybe just a suddenly loud CPU
fan), they self-rationalize that they'd rather spend another
$700 for a modern system to do web-surfing, email, and
office tasks. I don't argue with them, as I'm quite happy
to receive any OEM systems that only need minor fixes and
include a valid Windows license.

It's what I've observed from experience, not theory. The air has to
move slowly for the dust to settle out, so the places where there is a
slowing of the air or similar turbulence are where the dust collects.

True, it will collect more where there is slowing, BUT the
origin of the dust, it's entry into the system is increased
by this higher flow rate. Even though dust settles in the
lower flow areas, it also settles elsewhere.

But it is an area of slowing air flow, along trailing edges that
generate vortices.

Relatively speaking it's quite higher air flow than anywhere
else in the system except another area of the fan blade.
Further it is not only trailing edges that collect dust, the
entire blade does.

Whence the interest of using filters. Now if only I could find good
filters.

I'm slowly beginning to abandon the idea of a single-density
medium in favor of a double or triple layer filter. The
primary problem is implementation in a typical case. While
it's not so bad to do a couple cases for personal use
(particularly when forethought allows setting up a new
system thoroughly prior to migration from an older one), the
issue of having a larger filter area than fan intake area
makes it difficult to implement without significant case
front modifications- something many people seem to shy away
from when they've just received a new case. I sometimes
argue that it's easier to do modification work prior to
building into a case rather than later, but maybe it's the
excitement of a new build, maybe the perceived value of a
"new" case, for whatever the reason most often people seem
to be building a system and THEN thinking about cooling
rather than the other way around.

I'll wait and see, then. I know ball bearings last a long time. I'll
let someone else be the pioneer with new types of bearings.

It doesn't necessarily matter, begin a trial use with only a
few of them, continuing to make regular backups of data as
always, then if they're viable for the purpose you'll have
demonstrated it without significant disruption.

 

[ASI Industries]

I forgot to add i was running a 5cm x1cm 1000rpm fan instead of the 12cm
Zalman fan.

Also CpuIdle was running. http://www.cpuidle.de

The purpose of the test was how long before failure and with the stock hs
was only 35 minutes and 2 hours 35 minutes with the Zalman hs, before the
agp bus overheated and i lost video.

And i found out how inaccurate the onboard sensors were.

 

[ASI Industries]

Hi Kony,

Is this how it was all along? Perhaps that 's it, your CPU
heatsink is simply recirculating same pre-heated air around
the northbridge

No, 4 days ago it was in a case, maybe.

Hmm, I didn't remember there being a way to disable the heat
sensors. Perhaps you mean the thermal shutoff for the CPU?

Yes you are right, sorry got mixed-up with my Abit NF7-S V2 it does, but it
did shutoff and cpu hs was warm (30-40%C), NB hs was so hot i burnt my
finger, 65%C on my dmm.

Are you running the PCProbe software? Maybe that is the
reason for the shutdown

No i have found PCProbe to be inaccurate, Everest (Aida32) is better.

It has now dawned on me that your board didn't have the tall
silver Asus heatsink but rather the shorther dark-colored
one? Mine has the silver one and has done fine with it...
but it doesn't explain why you're not seeing good results
with the Zalman.

Yes the little black one about 10g, the Zalman NB hs is Aluminum
and its quite poor with heat transfer.

No I don't see one available to the public on Via's website,
but there aren't any that have lower than 80-85C, usually
higher, that I recall.

Same here, but i did see the specifications of the Via KT133A and
KT400 northbridge chipset and they were only about 10%C above
room / case temperature and with a hs / hsf it reduced the temps by
at least 50%.

Oh i was running Win98se at the time of the 65%C shutdown.

Anyway i have new data you will find it under `Mxsmanic' post.

 

[Mxsmanic]

Were they good makes though?

The sleeve or the ball bearings? I don't know enough about makes to
know which is best, and often I'm constrained to what I can easily
obtain, anyway (rather than the best of the best, which may simply not
be available).

True, it will collect more where there is slowing, BUT the
origin of the dust, it's entry into the system is increased
by this higher flow rate.

That's where filters come in.

Now, if you can recommend a filter (commercial or home-made) that will
effectively screen out housedust without too greatly impairing airflow
when placed in front of the fan, I'd be very interested. My current
fans have washable coarse plastic filters, but I think dust is getting
past those. I'd also prefer disposable filters to washable filters.

I'm slowly beginning to abandon the idea of a single-density
medium in favor of a double or triple layer filter. The
primary problem is implementation in a typical case. While
it's not so bad to do a couple cases for personal use
(particularly when forethought allows setting up a new
system thoroughly prior to migration from an older one), the
issue of having a larger filter area than fan intake area
makes it difficult to implement without significant case
front modifications- something many people seem to shy away
from when they've just received a new case. I sometimes
argue that it's easier to do modification work prior to
building into a case rather than later, but maybe it's the
excitement of a new build, maybe the perceived value of a
"new" case, for whatever the reason most often people seem
to be building a system and THEN thinking about cooling
rather than the other way around.

So what filters do you use? I should think a very thin, fine, diposable
filter would work pretty well. It should be possible to keep airflow
relatively high and filtering efficiency also very high if the filters
are designed correctly and replaced frequently.

It doesn't necessarily matter, begin a trial use with only a
few of them, continuing to make regular backups of data as
always, then if they're viable for the purpose you'll have
demonstrated it without significant disruption.

But is there a reason to buy sleeve bearings other than ball bearings?
I don't care if they are cheaper. And noise doesn't bother me. I just
want airflow and reliability.

 

[kony]

The sleeve or the ball bearings? I don't know enough about makes to
know which is best, and often I'm constrained to what I can easily
obtain, anyway (rather than the best of the best, which may simply not
be available).


That's where filters come in.

Now, if you can recommend a filter (commercial or home-made) that will
effectively screen out housedust without too greatly impairing airflow
when placed in front of the fan, I'd be very interested. My current
fans have washable coarse plastic filters, but I think dust is getting
past those. I'd also prefer disposable filters to washable filters.

Therein lies the problem, even with a filter that's not very
effective at filtering out smaller particules, the flow rate
will still be drastically cut. I don't think there's any
way around that except to have the largest filter surface
area possible. The way I did it last time was to mount the
fans flush with the case wall and put a large filter panel
about 20mm away from the fan intake, that filter panel
suspected with a thick sealed-foam tape surround such that
it kept it sealed.

The filter material I used was cheap enough to be considered
disposable but was washable, a panel meant for air
conditioners that was cut down to smaller pieces. Next time
I'll probably use accordian-folded paper fiber filters such
as 3M filtrete,
http://www.3m.com/us/home_leisure/filtrete/411_ultra.jhtml ,
or more likely a generic somewhat-lower performance type
(but overall same construction) as it will be desirable to
have a little better airflow at the expense of not capturing
the smallest particles, not to mention a lower price. I
noted a local hardware store had some off-brand for less
than 1/4 the price and it still looked far better than any
other type of filtering medium. While that might work for a
bezel-internal filter, the pleating wouldn't be so welcome
over any other (like side-panel) fans.

So what filters do you use? I should think a very thin, fine, diposable
filter would work pretty well. It should be possible to keep airflow
relatively high and filtering efficiency also very high if the filters
are designed correctly and replaced frequently.

Depends on what you consider "relatively high" to mean.
Generally if they're worth installing at all the resultant
flow is 40% or lower than previously, and even worse with
thin fans, I try to use 38mm thick when possible... which
often isn't possible if the case doesn't have a particularly
spacious bezel or recessed drive bays. I'm fairly handy
with sheet-metal and could fab a different bay or move
existing bays but beyond a certain point it's a lot of work
compared to the benefit of leser filtration, which still may
take quite a while to build up. Plus I still try to keep
noise levels as low as possible, would prefer a quiet
slightly dusty system over an audible, spotless one.

But is there a reason to buy sleeve bearings other than ball bearings?
I don't care if they are cheaper. And noise doesn't bother me. I just
want airflow and reliability.

Sleeve bearings are quieter, and what noise they do make is
that of turbulence, little to no bearing noise. Most
people, myself included, fine the bearing whine more
distracting than lower pitched turbulent sounds.

They aren't cheaper, not the decent ones anyway. The Papst
are among the most expensive one can find, ignoring spot
pricing of the rare fans at electronics warehouses which
often tend to charge more. A good sleeve-bearing fan in a
moderate temp environment (preferribly not on the power
supply exhaust) and vertical mounting orientation can last
quite long. I have some well over 10 years old that are
still quieter than the average new ball-bearing fan.

If you absolutely dont' care about the noise there is no
reason to buy sleeve-bearing fans, except for some unusual
environement with a lot of shock-stress, as the ball-bearing
fans are not so good for shock resistance. Some of the
larger fan manufacturers might provide more data on this, as
offhand I don't recall the g-force tolernace of common
ball-bearing types... it's not ever an issue in a computer
for sure, even a laptop couldn't be treated that badly and
be expected to work beyond which fan was installed.

 

[mcl]

13C is 55.4F.
Either he keeps his room very cold, or has a refrigerator for a cooler.
If neither of those two, the reading is out to lunch.