Do you have a question? Post it now! No Registration Necessary. Now with pictures!
- Posted on
- Measuring CPU vcore voltage?
September 23, 2006, 9:11 pm
rate this thread
Re: Measuring CPU vcore voltage?
Yes, easiest would be to reverse engineer the board VRM
subcircuit to figure out where to put your multimeter probe.
We might be able to help determine a suitable spot if we
knew what board you have.
In general there will be multiple stages of mosfets, and
last of which is followed by an inductor and a bank of
capacitors. You can measure at the output leg of the
inductor. Depending on board design, other boards may have
large copper planes behind the CPU socket, I mean on the
back of the board, but they're often in the socket-well too
on sockets that have an open center. You will typically see
several paralleled ceramic capacitors between the vcore and
ground plane there and can put a probe across one of those
ceramic caps, on each end of one.
If you have access to the rear of the board to do that while
it's running, you also have access to the capacitor leads of
those caps following the inductor mentioned above. Which
spot is easiest depends on your particular board. If you
happen to have needle tips for your multimeter probes, you
can even pick your own spot to measure as gently pushing in
can get to the top copper layer almost anywhere (but given
your question, probably better to fixate on a physical
component lead instead).
Re: Measuring CPU vcore voltage?
This is actually a tricky question.
The reason it is tricky, is the voltage is small to start with,
and the currents that flow are large (like 119 amps on a high end
Intel processor). The current is distributed by means of a copper plane,
extending from the output phases of the Vcore regulator, until it gets
to the contacts on the bottom of the processor.
The Vcore regulator circuit gets feedback from some point on the
plane, and the regulator will try to maintain the voltage as seen at
that feedback point. But the output voltage is not a constant, and
is expected to follow a "load line". The Vcore circuit might be
0.150V lower at full load, than when there is a minimal load. And
that "load line" is documented in the processor specification -
the processor maker fully expects and specifies the amount of
drop with load. That complicates exactly what a Vcore voltage
reading would mean. If you saw 1.44V when idle in the Windows
desktop, and 1.25V running Prime95, then exactly what is the
voltage ? (Inside the BIOS, the setting could well be 1.40V
as seen in the BIOS screen Vcore setting.)
The important thing here, is not to take your eye off the ball.
The voltage as seen inside the processor is the quantity you want
to observe. Pins on the bottom of the processor socket, like
VCC_SENSE and VSS_SENSE, allow a person in the lab to observe
the voltage being received by the processor core. And that tells
you exactly what the processor is seeing. PDF page 45, upper
right hand corner, shows the VCC_SENSE and VSS_SENSE pins on
a Prescott processor. You would need to have access to the bottom
of the motherboard, to get at these.
Another way to look at the problem might be, how does my
voltmeter compare to the voltage as measured by the hardware
monitor chip on the motherboard ? I don't know what motherboard
you've got, but a typical Super I/O chip with built-in hardware
monitor block, has eight voltage channels for measuring voltages.
Vcore should be directly connected to one channel of the hardware
monitor. Typically the Super I/O chip might have 100 pins, and one
of those pins has Vcore on it. The connection is direct, since the
voltage level is low enough to stay within the measurement range
of the monitor chip. When the monitor chip measures a voltage like 12V,
a resistive divider consisting of two resistors is used, and that
resistor network attenuates the voltage so it stays with the 4.096V
or less measurement range of the monitor chip. If you can find the
Super I/O chip on the motherboard, and can read off the part number,
you can go to winbond.com or iteusa.com and download a datasheet.
One problem with probing the Super I/O chip, is the pins are tiny,
so the probing would need well lighted conditions. I would
recommend disassembling the computer, and placing the motherboard
on your work table. Find a nice halogen work light, so you
can see what you are doing. (I use 50W held 12" from the work.)
Make your measurement either when the computer is idle, or when
Prime95 is running, so you can get some idea of the range of voltage
values for the load line. If you look at PDF page 28 of the above
sample processor datasheet, Figure 2-2 shows three line - the upper
and lower lines define the acceptable voltage range, while the center
line in the plot shows the "nominal" load line.
Are you still keen to measure it ? :-)