Re: sensor identification
- From: floyd@xxxxxxxxxx (Floyd L. Davidson)
- Date: Fri, 11 May 2007 17:42:30 -0800
ebenZEROONE@xxxxxxxxxxx (Hactar) wrote:
Floyd L. Davidson <floyd@xxxxxxxxxx> wrote:
ebenZEROONE@xxxxxxxxxxx (Hactar) wrote:
VCore: +1.25 V (min = +0.00 V, max = +4.08 V)
+3.3V: +3.36 V (min = +0.00 V, max = +4.08 V)
+5V: +5.01 V (min = +0.00 V, max = +7.02 V)
+12V: +12.03 V (min = +0.00 V, max = +16.32 V)
-12V: -12.01 V (min = -0.00 V, max = -16.20 V)
V5SB: +5.01 V (min = +0.00 V, max = +7.10 V)
VBat: +3.06 V
The voltages appear to be correct, but max/min values aren't
very useful.
Oh, I don't rely on those anyhow.
You *can't* rely on them; but if set properly you can and
probably will...
I wrote a script that makes graphs of
what they were over the last two weeks, and look for deviations from
their nominal values.
That doesn't do much of anything useful though (even if it is
very interesting!).
For examples of some extensive graphing, look up Phil
Edelbrock's "tellerstats" and also find the extension to it that
I wrote. Last time I knew tellerstats was included in the
lm_sensors distribution, and my extensions were listed in the
"useful links" on their web page. The link is,
http://web.newsguy.com/floyd_davidson/code/sensors/
However I haven't looked at it for a couple of years now, so it
may or may not work out of the box with the current
distribution.
You might want to pull those down to 1%, 5% or 10%,
of the target voltage, depending on which voltage it is. For
example the VCore voltage has very tight regulation, but the -12VDC
voltage does not.
Survey says the -5 and -12 are varying degrees of "inaccurate" too. The
-5V sensor read "0V", so I disabled it. The -12V sensor read -1.48V
before I rescaled it. I assumeic that it was linear, and that when the
sensor reports "-1.48V" it actually reads "-12V" (this PS is only a few
weeks old, and it's a decent one I think -- Raidmax RX-530SS). OTOH, it
could report "-1.48V" for any voltage below -1.48. I don't know, until
I have two (different) known points.
There may or may not actually be a -5 Volt rail. The -12 Volts
calculation is indeed open to debate as far as accuracy goes.
Or, a better way to describe it is that none of the voltage
measurements are necessarily accurate, and the negative voltage
readouts are perhaps less so that the positive voltage readouts.
The formula to calculate the displayed voltage can be changed to
make any actual variation display as either a large or a small
variation.
Another problem is noise on the voltage line, which can cause
what appears to be out of spec voltages to be indicated.
Apparently some motherboards are much worse than other in this
regard (due to not having bypass capacitors right at the sensor
chip).
In the end, the most useful configuration is one designed to do
what you have described: indicate abnormal variations. The
actual voltage or the exact variation isn't really very
important, compared to the fact that something has changed.
For that reason I would recommend using a configuration that
just sets the voltages to the voltage it is supposed to be,
graph it over an extended period of time, and then set alarm
limits at just beyond the largest variations seen in the actual
data. Then any significant change a year or more later will be
quite obvious.
CPU Fan: 1344 RPM (min = 0 RPM)
Cs front: 2472 RPM (min = 0 RPM)
Cs rear: 2755 RPM (min = 0 RPM)
Those may or may not be accurate. If you have slow speed fans they
might be. It could just as easily be that they are going twice that
fast. In particular it would be amazing if the CPU fan is only going
1344 RPM.
Well, it is at the slowest setting. At top speed it's turning about
2600 RPM. But I notice no additional CPU cooling that way, so I figure
why have the noise?
Okay, assuming those are all ~2600-2700 RPM fans. Most CPU fans
are faster, but lots of case fans are 2400-3000 RPM.
And as long the fan is automatically temperature controlled, it
might well be that slow. But if it is manually
controlled... that is not a particularly good method because
when it does need more cooling it will not be available. If it
is manually controlled it should be set to the fastest speed
that would ever be required.
Another problem with setting fans for very low speeds is that
when power cycled they may not start. A brand new fan will have
no problem, but a couple years later the wear on the bearings
might mean it requires more torque to start than the low speed
setting provides. It will run fine at the low speed once
started, but can't start there. Turning the computer off and
restarting it might result in damage if the fan doesn't start.
Also, you might want to set the minimum RPM (where an alarm will be
indicated) at 1/2 or 2/3 the normal speed. A 0 RPM alarm is worthless.
Gkrellm is set up to trigger an warning for the CPU fan at 1200 RPM, and a
more severe alert at 100 RPM. It actually speaks these alerts through
Festival (a text-to-speech system).
That should do the job.
CPU Temp: +29 C (low = -1 C, high = +127 C) sensor = diode
temp2: +36 C (low = -1 C, high = +127 C) sensor = thermistor
temp3: +25 C (low = -1 C, high = +127 C) sensor = thermistor
Those do not look particulary accurate. Unless you just turned on
the machine, the CPU is almost certain to be higher than 30C (86F).
You might try changing the sensor and see if there is one that gives
a reasonable value. Something over 40C is probably normal.
I do have a big honkin' heatsink:
http://www.newegg.com/Product/Product.aspx?Item=N82E16835118019
Could that make a difference?
Well, lots of things could. Big heat sinks, high volume fans,
and low power CPU's would all bring it down. But lots of CPU's
run as high as 55C, and commonly CPU's run 40-50C. And at 30C
(86F), that CPU is all but cool to the touch.
"CPU temp" and the two numbers returned by k8temp all look CPU-like
(judging by their performance over time), but they're all in the same
ballpark. temp1 is actually a few degrees lower than the other two.
Hmmm... does your BIOS setup have a monitor program? Lots of
motherboards can give readouts for the voltages and the
tempurature sensors when looking at the BIOS setup. That would
give you an idea what the absolute coolest temperature in normal
use would be.
--
Floyd L. Davidson <http://www.apaflo.com/floyd_davidson>
Ukpeagvik (Barrow, Alaska) floyd@xxxxxxxxxx
.
- References:
- sensor identification
- From: Hactar
- Re: sensor identification
- From: Floyd L. Davidson
- sensor identification
- Prev by Date: Re: Using XT keyboard with Linux
- Next by Date: Connecting USB devices to VMware Server?
- Previous by thread: Re: sensor identification
- Next by thread: Re: sensor identification
- Index(es):
Relevant Pages
|
