Bill Stock wrote:
> I've suffered a few BSODs after the PC has been on for a few hours, so I
> think it must be a heat issue. So I loaded up ASUS probe and it says my
> Opteron 185 is running at 55° at idle. I don't know if ASUS probe is
> accurate for this particular setup or not. MB is an A8N-SLI with stock AMD
> HS and fan. The HS is clean and the fan is spinning at 5K RPM. Room temp. is
> 80° F BTW.
>
> Any idea if PC Probe II is accurate? If so, WTF is the Opteron running so
> hot?
>
1) Run CPUZ and verify your frequencies.
2) Check the core voltage in Asus Probe. Note that if Cool N' Quiet
is enabled and set up properly on the board, that the processor would
operate at a variety of voltage/frequency values, as a function of load.
Thus, you may want to load up the processor with a computing task
(cpuburn, prime95, or the like) while doing measurements. It is harder
to do a reliable idle reading, as the program doing the reading may cause
the FID/VID to change.
For examples of Cool N' Quiet voltage ranges, there are a couple docs here.
Min Pstate is the Vcore at idle, if Cool N' Quiet is running. Max Pstate
is the Vcore as seen when the CPU runs at full load, or if Cool N' Quiet is
disabled. The OS adjusts FID/VID, something like 30 times a second. Which
is why, watching Asus Probe take a reading once a second, is going to miss the
details. Typical values might be 1.3V at Max and 1.1V at Min, and
that helps reduce the idle power output. Similarly, the processor
may have other features (bus disconnect on HLT and the like), which
further improve idle power.
Athlon64
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_doc...0430.pd
Opteron - typical Pdiss 110W full power, 49W in Min Pstate. 185 is not listed ?
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_doc...0417.pd
An Opteron 185 uses 110W at full load. Nominal Vcore at full load is
1.3V or 1.35V. Varies from unit to unit, and may not be stamped on
the box. It is actually pretty hard to draw any conclusions from a
single voltage reading, because in fact there is a load line for the
regulator. But if your voltage was way off, you may be able to detect
that.
http://www.amdcompare.com/us-en/opteron/details.aspx?opn=OSA185DAA6CD
For the CPU cooler to work, the computer case air temperature around
the cooler, must be low enough for good heat transfer. For example, if
room temp is 25C, a computer case or motherboard temperature of 32C or
less, would make it well cooled. If the delta to room temperature is
more than 10C, then you should work on improving your case cooling.
For example, my room temperature is 25C (I have a room temperature sensor
mounted in front of the computer, where the intake vent is). My internal
case temp is 29C. I have a 120mm fan on the back, running at a pretty good
speed. I removed the lower plastic bezel on my case, to expose more vent
space. I removed the dust filter as well. My CPU never goes over about 43C
(but it isn't as powerful as your processor).
You could change CPU coolers, but that only makes sense if the air in the
computer case is relatively cool. There are coolers with theta_R of
less than 0.20C/W . With a 110W processor running flat out, you
would expect a CPU delta_T of 0.20 * 110 = 22C above computer case
air temperature. If the case was 29V, that would leave the processor
at 29C + 22C = 51C. And that is a full load temp - at idle the delta
would drop a bit less than half, or 29C + 11C or 40C. So there is some
room for improvement.
Example of a cooler with theta_R below 0.20C/W:
http://www.zalman.co.kr/eng/product/view.asp?idx=165&code=005
The reason I'm suggesting checking frequencies and voltages, is the power
dissipated is proportional to those. F*C*V**2 at normal operating values
is listed at 110W. If the voltage is higher than the 1.3 or 1.35V nominal
values, the power dissipated goes up as the square of the voltage, so that
can have a bit of an effect.
Hope that gives a few ideas,
Paul
FAQ
Profile
Private Messages
Log in
