In which your humble author learns about CPU temperature measurement. It seems that the two most common ways of measuring CPU temperature are with a thermistor in the CPU socket and with a diode on the actual die.  My motherboard is an ASUS A7V600-X which apparently has both.  Needless to say the on chip diode is the more accurate sensor but the Windows based tools I used during construction did not indicate which sensor was in use.

When I started the project I guessed that there would be more tools readily available for Windows then for Linux for testing the CPU so I installed a copy of Windows 2000 on a disk for testing.  I used Motherboard Monitor and SuperPI to monitor and stress the CPU.

Later I discovered that Linux actually had some better tools, namely lm_sensors and burncpu.  I use Scientific Linux 4.2 so lm_sensors was already installed.  Running sensors-detect and restarting lm_sensors with service lm_sensors restart got me up and running.  the command watch sensors gave me a temperature reading that was updated every 2 seconds.

However, the MB and CPU label were switched and the temp didn't look quite right.  I edited my /etc/sensors.conf and updated the it87 section to read:

# Temperature
# Important - if your temperature readings are completely whacky
# you probably need to change the sensor type.
# Adjust and uncomment the appropriate lines below.
# The old method (modprobe it87 temp_type=0xXX) is no longer supported.
# 2 = thermistor; 3 = thermal diode; 0 = unused
   set sensor1 3
#   set sensor2 3
#   set sensor3 3
# If a given sensor isn't used, you will probably want to ignore it
# (see ignore statement right below).

    label temp1       "CPU Temp"
    set   temp1_over  60
    set   temp1_low   15
    compute temp1     (@+128)/3, (3*@-128)
    label temp2       "M/B Temp"
    set   temp2_over  45
    set   temp2_low   15
   ignore temp3
    label temp3       "Temp3"
    set   temp3_over  45
    set   temp3_low   15

# The A7V8X-X has temperatures inverted, and needs a conversion for
# CPU temp. Thanks to Preben Randhol for the formula.
#   label temp1       "CPU Temp"
#   label temp2       "M/B Temp"
#   compute temp1     (-15.096+1.4893*@), (@+15.096)/1.4893

# The A7V600 also has temperatures inverted, and needs a different
# conversion for CPU temp. Thanks to Dariusz Jaszkowski for the formula.
#   label temp1       "CPU Temp"
#   label temp2       "M/B Temp"
#   compute temp1     (@+128)/3, (3*@-128)

# Fans
    set fan1_min 0
    set fan2_min 3000
   ignore fan3
   ignore fan2
   ignore fan1

    set fan3_min 3000

This gave me more reasonable readings.  When testing with SuperPI on the Win2K disk or encoding with Lame on Linux I was able to get the CPU utilization to 100% but the temp only went up a few degrees.  Running the Athlon optimized version of cpuburn caused the temp to climb much higher and much faster.  The moral is an indicated 100% CPU does not always mean the CPU is maxed out as far as heat production is concerned.

As I write this my temps are:

CPU Temp:    +48°C  (low  =   +15°C, high =   +60°C)   sensor = diode
M/B Temp:    +36°C  (low  =   +15°C, high =   +45°C)   sensor = thermistor

Which a little higher then I'd like but nearly identical to what it would be with the original heat sink installed.  
The first test was to just look at the temperature in the BIOS hardware monitor screen. bios temperature

Next I booted with Win2K and ran Motherboard Monitor.  The temp was about 38 degrees C at idle and went up to 43 degrees when running SuperPI for a half hour.

This was very encouraging but, as it turned out, not really a true test.
motherboard temp
I used Teflon tape to seal the water block fitting and the they leaked!  Both of them!
cpu water block leak
I replaced the Teflon tape with good old pipe dope, bye bye leaks!
pipe dope
The typical motherboard lives in a windy box, as such chips like this one are cooled by the breeze from several sources.
hot chip
This is the "Northbridge" and it got really hot too.
another hot chip

I stuck one of the PowerEdge fans on these two chips to keep them cool during my testing.

Later I thought, "Hmm, I wonder what that fan would be like running on 5 volts?"  It worked very well indeed on only 5 volts.  If you put your ear up to it you could hear a small amount of motor cogging noise but no air noise at all.

Rather then making water blocks for these two chips I decided to just install the "half" a fan to gently blow on the hot bits.
half a fan
<-- Previous Next -->

More Projects . . .

The Steampunk Workshop

See me on PBS Make:TV!





Passive Water Cooled PC Project:

Stripping the case and adding drive bays

Water Cooling the Power Supply

Cooling the Video Card

Water Cooling the CPU


Final Assembly