Do you have a question? Post it now! No Registration Necessary. Now with pictures!
- Posted on
- Fan speeds
- Texas Yankee
March 2, 2006, 8:56 pm
rate this thread
My machine isn't overclocked, but you guys are the experts on cooling
issues - I need some direction. Just purchased a new P4 3.0 GHz, 2 MB
L2-Cache and 800 MHz FSB in a BTX case. Never had any apparent cooling
problems, but because it was SO FREAKING NOISY I replaced the original Intel
CPU fan with an Intel supplied replacement - then tossed it (still noisy)
and put in a Thermaltake unit specifically designed for the BTX case - it's
a lot quieter, but watching the Intel Desktop Utilities Hardware Monitor
there doesn's seem to be any direct correlation between the CPU temp and the
fan speed. In fact, it looks like the reading for System Zone 2 Temperature
has more to do with the fan speed, and the noise. Early AM the machine is
near silent - but during the day, even when the CPU isn't even near at max,
the fan stays running fast - slower would be quieter, and that'd be a lot
better for me - any suggestions? Thanks!
Re: Fan speeds
Well, first of all, cooling design takes the whole platform into
account. Without a full hardware inventory, it is hard to
determine where the best cooling gains can be had. For example,
if the computer case air temperature is high, then adding a fan
or punching out some space for a vent somewhere, or the placement
of a duct, might be used to improve cooling performance.
The single most important temperature zone in the computer, is
the disk drive area. While in theory some disk drives can
report temperature, via SMART, the result may be hard to
interpret due to the lack of standards. A cheap digital
thermometer (I got a 2 channel one at a local computer store
for about $30), can be used to determine the surface temperature
of your disks. Depending on whether the computer is being used
in a controlled humidity environment (air conditioned room, for
example), will determine how far above 35C the internal case
air temperature can be allowed to go.
For the CPU, the most immediate concern, is avoiding the
throttle temperature when the CPU is running at 100%. If the
CPU die temperature exceeds 65-70C, the processor reduces the
compute rate, by gating the clock. So that is a performance
target for your cooling solution.
The next level of concern, might be the effect on reliability,
of the average temperature the processor is maintained at.
I don't have any good numbers for a modern processor, and
MTBF versus temperature. The Arrhenius relationship (a chemistry
equation) correlates with memory-like silicon designs, and
predict the failure rate doubles, for each 7C temperature rise.
But we don't know the absolute failure rate, and cannot say
whether the MTBF is 100 years or 2 years at some reference
This document contains an example control method for an
"Intel Pentium 4 Processor on 90nm Process Thermal and
Mechanical Design Guidelines" (PDF page 33)
In the diagram, they suggest there are two factors controlling
the fan. The Intel fan has a thermistor in the fan hub. It
senses the air temperature in the case. Between 30C and 38C,
an Intel fan ramps between its min speed and max speed. Above
38C it will run at max. Overlaid on top of that case air
temperature sensitive response, is the PWM control signal
fed to the fourth pin on the fan connector. The hardware
monitor chip on your motherboard, senses the CPU diode
temperature, and adjusts the PWM setting to the fan. As the
Intel document describes, it can be set to ramp the speed
from min to max, when the CPU hits a relatively high temperature.
If there is no software installed, to program the monitor
chip, and create the necessary response, the fan may run
flat out. Speedfan, from almico.com, is an example of a
fan control program.
The effect of the two control methods, is most of the time
the fan should be running at relatively low RPMs. If the
computer case air is 38C or higher, it means the base rate
for the fan is as high as it will go. Depending on the
"depth" of the response of the PWM from the monitor chip,
will determine what percentage of that full rate, will be
applied. (If the PWM only varies from 70% to 100%, you may
think it is not working.)
That leaves us with the cooler solution. The basic CPU temp
is determined as follows. If the ambient temp is 25C, a
well cooled computer case will be 32C inside. A user may add
fans or punch extra ventilation holes, to allow enhanced air
flow, until the 32C target is met (i.e. the target is actually
the difference between computer case air temp and room temp, and
should be 7C in a well cooled case). The CPU temperature is
determined by theta_R (in units of degrees C per watt of
power dissipated by the processor) times the CPU power. An
estimate of the power, is the TDP for the processor, which
can be found on the processorfinder.intel.com web site. You
plug in your processor part number, and you can find a TDP
in watts for the processor.
Let us say your processor is 89W max. The following BTX
cooler is a relatively poor 0.29C/W theta_R, at max fan
speed. 89W * 0.29C/W = 25.8C delta temperature. We add the
25.8C from the calculation, to the case air temperature of
32C, to get the CPU temp of 57.8C. And, you can see, that
if the fan cannot run flat out for some reason, the temp
will rise higher than that. The curve on the thermaltake
web page shows the theta_R versus fan speed. (I personally
shop for aftermarket heatsinks that are lower than 0.25C/W,
but it really depends on how bad the processor is. A 130W
processor needs a damn good heatsink.)
I hope you can see from this, that a lot of info about your
hardware, is needed to figure out how best to fix the problem.
If the case air is too hot, the case must be modified to help
keep the disk drives healthy. Depending on the quality of the
CPU cooler, it is a judgement call as to whether money would
be better spent making the computer case cooler, or the CPU
cooler. If the case air temp was 45C, the worlds best CPU cooler
will not help. It is already too hot in there for the cooler
to do anything.
There are some after market video card coolers, that blow their
exhaust air right out of the case, so the waste heat is not
recirculated inside the computer case. THe BTX design vents the
CPU hot air out of the case, and that should also help keep the
case temp down. If you have a monster video card, you might
consider a product from this page (shop carefully, to find
something that will actually fit, and the holes line up).
Re: Fan speeds
Thanks very much for the detailed response. After reading your post, and
doing some more research, I *think* I understand what's going on. The MB is
a D945GCZ and the layout is pretty much as described in the graphic in
http://www.intel.com/support/motherboards/desktop/sb/cs-012552.htm - the HD
is a 120 GB that's installed to the left of the MB in the front of the BTX
case - yeah, it gets hot, but I don't think that's the root of the problem.
No problems with CPU temperature either, according to Intel Desktop
Utilities - always well within the recommended temperature range.
I should have talked with you before I bought my OEM cooler - yup, it's the
cl-p0191 - but it's a lot quieter than the stock Intel unit and it seems to
cool the CPU just fine.
What I *think* is going on is that the register for System Zone 2
Temperature (look at the diagram in the link) is almost directly in line
behind 1) the air output from the CPU cooler fan, and 2) then there's
another heatsink (I think it's for the onboard video), then it hits the
temperature register - so the hot air from the CPU cooler passes over
another heatsink and then hits the temperature register - and the fan runs
because System Zone 2 show hot.
I'm thinking an additional fan at the back of the case, an exhaust fan,
might help - but that wouldn't do anything to deflect the hot air that's
blowing directly onto the register.
Re: Fan speeds
The exact location of your thermal sensors, is shown on PDF
page 37 of this document. One sensor is before the Northbridge,
the other on the lower left, by the memory.
This picture of a BTX case from Anandtech, and the orientation
of the case (front of computer is at bottom of picture) matches
the orientation in the Intel manual.
The purpose of the sensor near the memory, would appear to be
an attempt to measure air temperature near the disk drives.
The one near the Northbridge kinda has me baffled.
I've never worked with BTX before, and I can never be sure I've
got the direction of flow correct. First of all, there are several
sizes of BTX cases, and the airflow in some of the Dells might
not be done exactly the same way as Intel describes.
Generally, when adding fans to a computer case, you don't want
to disturb the general airflow pattern, but want to enhance it.
In other words, when multiple fans are present, depending on
the airflow of each, it is actually possible to blow air backwards
through the lowest capacity fan. If your computer case is
particularly small, there simply might not be an advantageous
place to put another fan.
I tried the Intel web site, but cannot find a really good
set of drawings of how it is supposed to work. The thermograph
in one of Intel's pictures, doesn't tell me how the fans are
supposed to run, for example.
The fan on the power supply, is usually such a low velocity fan,
as to be "useless" or "negligible" in the whole scheme of things.
That means, it is entirely likely that the processor fan is
providing _all_ the effective cooling in the computer case.
Whatever device or software controls the fans, could be weighing the
temp of the various zones, and cranking the processor fan speed
based on that. With the four pin header on the CPU fan, the
fan speed could be the product of the case air temp and the
voltage response of the monitor chip. Which points to the case
air temperature being a major driver for the response of the
Do you have the URL for a picture of the computer case ?
Adding an intake fan on the back of the case, will be blowing
air in the same direction as the processor fan (towards the
front?). If the intake fan is really powerful, it may end
up blowing air backwards through the PSU. If the PSU gets
hotter, the PSU should speed up its own fan in response.
Adding an exhaust fan on the front of the case, works in
parallel with the CPU fan. But then that exhaust would be
blowing in your face. And there might not be a place to
mount the fan.
If you are using the built-in graphics, then the processor
airflow provides the cooling. If you are using a separate
powerful graphics card (one with a power drain of 70W or
so), then a video cooler like one from this page
could be used.
The way those coolers work, is they use an adjacent PCI
slot hole, and exhaust the hot air from the video card,
out through the PCI slot. That prevents the video card
heat, from heating up the computer case air.
In your BTX case though, I'm not sure the cooler is on
the right side of the card. I suspect the plastic vent
would not line up with a hole on the case.
See if you can dig up a picture of the case. If the
case is really tiny, there might not be too much you can
do to it. Putting an exhaust up near the drive bays,
might reduce the air temperature up there. (If I am
interpreting the picture of the computer case correctly,
the disk drives look to be in a dead zone air-wise.)
Putting an exhaust over the Northbridge, and on the side
of the case, might be another option.
As for cooling fans, I try to avoid the ones with
thermistors built in. The problem with the thermistor, is
it implements a cooling curve of the manufacturer's choice,
and not the end user.
According to Intel's notes, on the four pin fans, Intel
does not recommend old fashion voltage control. The Intel
fan is supposed to get between 10.8V and 12V on its 12V
input, and a PWM signal goes to the fourth pin. Old fashion
voltage control, would not use the PWM signal, and would
deliver maybe 7V to 12V to the 12V input. To use a manual
fan speed control, it might be better to find a fan with
a three pin header, to take the place of the fan with the
four pin header.
The video card in this Dell, gets to add its warm air
to the rest of the computer case air. And it looks to me,
like the hot air from the Processor, flows over a part of
the lower drive bay, before going out through the front
vent ? Its a puzzle, thats for sure.
Re: Fan speeds
Thanks again, Paul.
Here's a link to the case that I'm using:
Sensors B anc C never have any problems - temps are always fine - I believe
that it's Sensor A that's causing my fan to run excessively fast.
Looking at the diagram of the MB, the CPU fan blows air from outside INTO
the box (over the CPU HS) - then warm air from this HS passes over another
HS (the square directly above and slightly to the left of the CPU) and then
hits Sensor A - no wonder it shows hot, huh?
Even if I put another exhaust fan at the back of the box, it's not going to
have a big affect on the reading from Sensor A, in my opinion - the hot air
will have already hit the sensor - besides, the last thing I need is another
noisy fan running.
Interestingly, almost as soon as the CPU goes to 100%, the fan speeds up -
and then slows down noticably when the CPU utilization drops off - but it
never drops off to where it was previously - it stays running faster - it's
gotta be the Sensor A reading that keeps it running faster - although the
Sensor A reading drops back down, too, just not as fast as the Sensor B
reading . . .
I think you're right on "the case air temperature being a major driver for
the response of the cooling system."
I'm using built-in graphics - no additional card - I believe that's what the
second HS is for - to cool the graphics chips - I've already tuned hardware
accelleration off - no big difference as far as I can see.
I tried using a small piece of glass to divert the airflow away from the
Sensor A - that does seem to help.
I'm still wondering though - I can't be the only one that's running into
this - looks like a poor design . . .
Any other suggestions?
Re: Fan speeds
OK. Here are two reference pictures. Correct me if I get the direction
of flow wrong again :-) It's so confusing.
System laying down. Air intake is from the single vent in the top of
the picture, pushed over the CPU by a max 60CFM fan. On the rear of
the case, the PSU exhausts air from the case (and the CFMs are likely
pretty low). Several passive vents on the back of the case, offer a
low resistance passive exhaust path.
Side view, showing drive tray rotated back into place.
As I understand it, this is your current cooling system. The CPU fan
is a wind tunnel design, with a 60CFM fan at one end of the duct.
I don't know if the fan is at the "inside the case" end of the duct,
or the fan is right up against the front of the case. Front of the
computer case is at the top of the picture. 5.6" x 15.3" x 13.1".
| CPU | Original
| Fan | System
| | |
| v |
| | |
| +-- | --|
| | v |
|PCI Passive | PSU |
|Slots Vent | Fan |
In the passive vent area, which I think is over top of the I/O
area, it looks like there may be room for a 70mm fan. A
midrange fan might be 30CFM or so. (A 70mm fan doing the job,
will be noiser than an 80mm. A 60mm even more so.) If you
place an exhaust fan there, that will change the direction
of airflow on the PCI Slot vents, and the tiny vent underneath
the PSU, to intake. That would be your first experiment. It
would tend to mix more cool air through the (reversed) passive
vents, diluting the hot air over the Northbridge heatsink.
(Note - for this to work, when the fan is fitted to the large
passive vent over the I/O area, the "spare space" next to the
fan must be closed off with sheet metal. That is to prevent
air from "cheating" and flowing around the edge of the fan.)
To make better use of the vent area, you could fit a large
fan to the outside back of the case, and build a plenum to
fit the larger fan, to the rectangular vent space. I've done
a mod like that on one computer case here.
| CPU | Plan "A"
| Fan | Add exhaust to back...
| | |
| v |
| | |
| +-- | --|
| | v |
|PCI 70mm | PSU |
|Slots Fan | Fan |
^ | | | ^
| v v v |
* (tiny vent under PSU,
becomes an intake)
If that doesn't help (and it may not be enough), the second
step would be tougher. You need to find a place to add vents
to the front of the computer case (i.e. to the top of my figure
above). Those vents would augment the bit of ventilation coming
from the PCI slots and underneath the PSU. The air would flow
past the sides of the CPU duct, and the airstream would mix
with the CPU air at the Northbridge. The CPU fan is pretty
powerful, and perhaps the side stream air from the new front
vents, wouldn't have an impact on the sensor. Due to the drive
bays on the hinged piece of metal, the added vents would have
to be located below the drives. The pictures on Newegg,
of your computer case, don't show enough detail as to how
difficult that would be. This plan helps the drive bay
temperature, but may not be enough to reduce the temps
at the Northbridge. (Call this Plan "A+B".)
The side of the case, and the drive bay hinged metal piece,
would make it hard to add a vent to the side of the computer.
The vent would have to be closer to the back of the case,
and by getting close to the newly added fan, the air path
to the fan would be so short, no components would get the
benefit from the rear fan.
An alternative, would be to fit a fan to the side of
the case, and arrange it to be over top of the Northbridge.
A well placed piece of cardboard, might be used to deflect
the CPU air to either side of the Northbridge. (The piece
of card doesn't have to be a full duct, and a half duct
might be sufficient to keep the hot air away.) The fan
attached to the side of the case could be a large, low
RPM model, which would be better from a noise perspective.
With the deflector in place, the air coming into the case
via the new side fan, gets to the Northbridge area without
mixing with the CPU exhaust. This plan does nothing for the
disk drives, and besides trying to get a quieter solution,
you do want to keep the drive surface temp down to 35C
if you can manage it. I don't see how Intel's grand plan
is doing that in this case.
| CPU | Plan "C"
| Fan |
| | | Intake fan above
| v | the Northbridge.
| ______ deflector | Top to bottom deflector
| / \ | cardboard, to prevent
| / \ / \ | CPU exhaust from hitting
| \/ | | the Northbridge.
| In /\ Take +-- | --|
| / \ | v |
|PCI | PSU |
|Slots Passive | Fan |
Due to the 5.6" dimension of the case, and the possibility
that the Northbridge headsink is a tall one, you may have
to shop for a reduced thickness fan. I'm having trouble
imagining what that looks like in 3D, and how much room
is available. The fan should be a glove fit to the side
of the case, to prevent air leakage around the edges of
the fan body. (I use thin sheet aluminum available at
the hardware store, and sometimes in the auto body repair
section of the store. I have a nibbling tool for cutting
detailed shapes in the metal. Electrical tape or rubber
inner tube material, might be used as a gasket.) The
hole in the case side panel, may require a pretty brutal
saw to get through the material - one reason for trying
plan "A" first. It may be easier to cut a square hole, and
then fill in the differences with sheet aluminum which is
easier to work into the required shape.
If you need inspiration, this in Intel's effort to dazzle us
with equations. I'm afraid I'm still not a big fan of the
BTX design concept:
In this doc, one of the tables shows the design target as
55C max at the disk drive. They must be nuts. The allowed
temp curve in a couple of disk drive docs allows nowhere
near that high a temp.
There is no shame in getting a larger computer case, with
room for more cooling devices. They don't have to be loud
to do the job. A Coolermaster CM stacker is a dual use,
ATX/BTX case, that can be reassembled to suit either
application. A new version has come out, and the older
version is visually more appealing. Don't ask me which type
of CPU cooler you use with this thing...
Re: Fan speeds
Just curious - the fan runs faster than it should (IMO) just because Zone 2
gets hot, as expected with the air from the two heatsinks hitting it
directly - I'm going to try removing the Zone 2 sensor and see what happens.
- » amazing retro FSB overclock on old Prescott P4 3400MHz 478 Pin uPGA
- — Newest thread in » Overclocking PC Components