View Single Post
  #1  
Old August 24th 04, 08:42 AM
Paul
external usenet poster
 
Posts: n/a
Default P4G8X core voltage

In article , moe @mci.com wrote:

I'm running a P4 2.66Ghz (533fbs/512cache/1.53 core v) on a P4G8X- Dlx
motherboard.
The bios core voltage is set at 1.525v but the voltage readouts
in Asus Probe, MBM5, and the Bios hardware monitor all show the core voltage
running at 1.6v which is a substantially higher than the set point. Is this
normal or does it indicate a problem with the boards voltage regulator?

Are there any other factors which could result such a large offset between the
set and actual core voltage. ( I'm assuming the voltage read by the monitor
programs is correct)

thanks
moe


It is really hard to say what the Asus circuit is measuring, and
how the software is converting the reading to the value you see on
your screen. MBM5 should be the most "honest" software, as it
reads the hardware directly and then computes the value. Probe or
the BIOS, could be shielding the user from the truth.

First of all, the hardware monitor will have some tolerances.
A typical monitor chip has an 8 bit ADC to make the reading.
Full scale might be 4.096 volts, meaning the step size is 0.016V.
But note though, that if you watch the readings, the steps don't
seem to be a multiple of 0.016, so I don't know if a running
average is being used, or how to explain the pattern of
readings you get. The ADC will have a voltage reference in it,
and typically references are pretty sloppy, like maybe 1% or
2% of full scale etc.

Also, Vcore is distributed on a solid copper plane on the
motherboard, and the Vcore measurement is supposed to be made
at a particular point, to be valid. Perhaps Asus isn't connecting
the monitor to the correct physical point in the circuit, or
another possibility is that there is a ground potential difference
between where the monitor chip sits, and the processor Vcore plane.
Measurements like this should be made differentially (a + lead and
a - lead), instead of single ended (a + lead and a shared ground
level). Measuring low voltages like this, requires a good deal of
care, to be done properly.

On some Asus motherboard models, the high Vcore value is very
consistent from motherboard to motherboard, which considering
the tolerances in the measurement, makes it hard to believe
the reading is an honest one.

If you are at all curious, download a datasheet for an Intel P4
processor. This is one from my collection:

ftp://download.intel.com/design/Pent...s/29864312.pdf

The VID specification is not a "single point" spec. In fact,
the expected voltage follows a "load line". When the processor
is running a program at 100% load, the voltage will be low by
roughly 0.14 volts, for your nominal 1.525V processor. (That is
assuming the processor is drawing 54 amps at the time.)
Similarly, the voltage will be higher at idle, but should still
be less than VID by roughly 0.04 volts. (That is assuming
the processor is drawing 7.5 amps at the time, based on a guess.)

The Intel spec would make it seem that voltages higher than the
stated VID value, are out of spec. But the thing is, Asus
is going to design the thing, such that they don't get a lot of
calls from users complaining that "my Vcore is low, I want to
RMA".

So, try measuring the voltage when Windows is idle, and also
when your computer is running Prime95 in torture test mode
(mersenne.org). See how the results line up with the P4
datasheet.

The absolute max for the P4 is 1.75 volts, and if Vcore is
headed there, then you are right to worry. While by definition,
a failure of the Vcore circuit to keep the processor voltage
between the "high" and "low" load lines, is a failure to
meet spec, I would only expect real trouble if the processor
spends a lot of its time at 1.75 or higher volts. For
example, a Tualatin processor has a rating of 1.75V max,
and some die after 4 weeks at 1.8V.

HTH,
Paul