Considering the difficulty in finding a system health monitor utility that
works well with my EVGA 680i SLI motherboard, my Centech non contact
infrared digital thermometer is the best workaround I have found. And at
only $7 US, quite a bargain.
The Centech (item 93983-2VGA at Harbor Freight
http://ww2.harborfreight.com/cpi/ctaf/Category.taf?f=bylogo&logourl=ce...ch.gif& )
seems to be accurate to within ~ one degree C. For your $7 US you get, in a
device about the size and shape of a fat USB thumb drive that offers
temperature measurement range: -33 C to 110 C, Fahrenheit or Celsius
selectable
accuracy: +/- 2%
resolution: 0.1 degree
distance to spot diameter ratio: 1:1
fixed emissivity: 0.95 fixed
measurement modes: Hold / Minimum / Maximum
PLUS digital clock and stopwatch.
If you get easily board by details, just resolve to purchase this
inexpensive tool and skip the rest of the post. But I'll bet you get hooked
by this new way of examining PC system temperatures.
With my EVGA 680i SLI motherboard, E4300 @ 2.7 GHz, 1.275v core, Patriot
DDR2 memory @ 1200 MHz, 2.150v, nForce (chipset) @ 1.200v, nVidia 8800 GTS
320 MByte at stock speeds and voltage
and an ambient temperature of 26.8 C (80 F) [ central heating thermostat
stuck ]
system at idle TAT reports Core0 and Core1 temperatures of 39 C and 37 C
respectively
nVidia Monitor SpeedFan 4.33 Centech
CPU: 34 C CPU: 34 C ThermalTake 7i heatsink
fins: 34 C
System: 40 C AUX: 40 C Chipset heatsink fins: 44.7
C
GPU1: 54 C Core: 54 C Back of video card opposite
GPU: 44.8 C
HD0: 32 C Hard drive pc
board: 33.4 C
HD1: 30 C Hard drive pc
board: 29.1 C
Core0: 22 C Sound control foam,
inside case wall: 27.6 C
Core1: 22 C Antec NEO 550 HE PS
box exterior: 29.7 C
Ambient: 0 C ????
System: -65 C ????
After 15 minutes with 100% TAT load on both cores
ambient air temperature 26.8 C (80 F)
TAT reports Core0: 68 C, Core1: 67 C
nVidia Monitor SpeedFan 4.33 Centech
CPU: 63 C CPU: 63 C ThermalTake 7i heatsink fins:
35.2 C
System: 40 C AUX: 40 C Chipset heatsink fins: 47.3 C
GPU1: 55 C Core: 55 C Back of video card opposite
GPU: 44.6 C
HD0: 32 C Hard drive 0 pc
board: 33.9 C
HD1: 30 C Hard drive 1 pc
board:
Core0: 52 C Sound control foam,
inside case wall: 27.8 C
Core1: 52 C Antec NEO 550 PS box
exterior: 30.5 C
Ambient: 0 C ????
System: -65 C ????
Conclusions:
nVidia Monitor and SpeedFan 4.33 report three identical temperatures at
idle and three identical temperatures at 100% TAT load. What nVidia Monitor
calls 'System' and SpeedFan 4.33 calls 'AUX' must be from a thermistor on
the motherboard (the chipset heatsink fins are too hot for 'System' / 'AUX'
to be the chipset temperature, and what is left but the motherboard?
The Core0 and Core1 temperatures at idle reported by SpeedFan 4.33 are
completely bogus as they are more than 4 C below ambient room temperature.
Both nVidia Monitor and SpeedFan 4.33 report identical GPU temperatures. It
seems reasonable to me that the ~ 45 C reported by Centech for the back side
of the 8800 GTS card (opposite the GPU) as it just pokes along is a
reasonable correlate to the ~ 55 C GPU temperature reported by nVidia
Monitor and SpeedFan 4.33.
Air temperature inside the case can easily be measured by aiming the Centech
at thermally nonconductive rough surface (in my case, the black sound
control foam an a piece of black construction paper give the same readings.
What REALLY interests me is the relatively low temperature of the
ThermalTake 7i heatsink fins (temperature measured near fin edges ~ 1/4"
from where the heat pipes attach) compared to the TAT reported temperatures.
It makes me question the thermal resistance between the CPU die and the heat
spreader and/or the thermal resistance between the heatspreader and the base
of the heatsink. When under 100% TAT loads there is a 68 C - 35.2 C = 32.8
C drop from the TAT temperatures to the ThermalTake 7i heatsink fins near
the heat pipes. When at idle, there is a 39 C - 35.2 C = 3.8 C drop from
the TAT temperatures to the ThermalTake 7i heatsink fins near the heat
pipes. I believe the two temperature drops, plus the total power
dissipation plus the ambient air temperature should be enough to figure the
thermal resistance between the CPU die and the heatsink base.
Makes me seriously consider,when I finally have a 'Penryn' Quad upgrade in
hand, removing the E4300 heat spreader and attaching the heatsink directly
to the CPU die (it worked for my bare die Pentium III 1 GHz 100 MHz FSB, so
why not for the E4300? Attaching the heatsink efficiently to a small die (8
X 8 mm?) is going to be very difficult compared to attaching the ~ 4" cube
ThermalTake 7i to the ~ 40 X 40 mm integral heatspreader. I wonder if
variable quality of the CPU die / integral heat spreader contributes a large
part of the difference in overclocking limits seen in overclocking sites?
Hypotheses welcome!
Phil Weldon
FAQ
Profile
Private Messages
Log in
