Do you have a question? Post it now! No Registration Necessary. Now with pictures!
- Posted on
- Pentium 3 Current consumption
- Ricky Romaya
April 27, 2006, 12:51 pm
rate this thread
Re: Pentium 3 Current consumption
Start at http://processorfinder.intel.com
Select the family you want.
When the processor list comes back, click the one you want.
You'll end up on a page like this:
The processor TDP in that case is 29.9W. TDP is "thermal design
power", and is some kind of estimate of processor power. In
some ways, it can be a "boiler plate" or imaginary number. If I
saw a 2.8GHz, a 3.0GHz, and 3.2GHz processor, all rated at
89W, I know that the lower two are not burning 89W. If the
3.4GHz was rated at 115W, I know the 3.4GHz was a bit above
89W, while the 3.2GHz must be pretty close. TDP is good for
budgeting purposes, and may give you a safe number, but technically
that number is a bit too "foggy" (unreal) for my liking.
Not all the P3 listings have a TDP, because someone at Intel
is a slacker :-)
The motherboard designer is free to use whatever rail they
want as a power source. If the designer chooses the 12V
rail, and the Vcore conversion circuit is 90% efficient,
the amps would be (29.9W/12V)*(1/0.9)=2.77A from +12V.
If the 5V rail was used, and the Vcore conversion was 80%
efficient, the current is (29.9W/5V)*(1/0.8)=7.48A from +5V.
In my USENET postings, I've traditionally used 90% efficiency,
but motherboards could be anywhere from 80% upwards. A cross
check for the efficiency, is to feel the Vcore components with
your fingers, and estimate how many watts are being dissipated.
Based on checking a few motherboards here, the losses are
pretty low and the circuits run cool.
As to how you figure out whether +12V or +5V is being used,
that is a tough one. On a P4 motherboard, the presence of the
ATX12V 2x2 power connector, tells you immediately the source of
power is 12V. On the older computers, you'll need a multimeter,
and a bit of cleverness, to figure out the power source.
Where the power enters the circuit, there is usually an LC.
The L is an inductor. On a multiphase Vcore circuit, the "odd
man out" inductor, like a cylinder with wire on it, could be
the input inductor. One side of that inductor goes to the
processor power source, and touching your voltmeter there will
tell you whether +5V or +12V is used.
Perhaps someone else knows the tradition of those designs
better than I do, and can tell you which one is more likely
to be used.
To see a sample schematic of a P3 type powering circuit, look
at page 12 of this document. L1 would be the "cylinder"
inductor, while L2 would be the "donut" output inductor.
Around L1 is where you want to measure, and preferably in
this case, on the left hand side of L1. The side not
touching the input caps labelled "Cin 5x1000uF" in the
diagram. What that means, is they want the designer to use
five separate cylindrical capacitors for input filtering.
You want to touch the side opposite the caps, to detect the
DC voltage on the inductor. The L1*Cin node will be "pulsing"
a bit, at several hundred kilohertz. The other side of L1
should be a bit more stable.
Re: Pentium 3 Current consumption
Your 90% figure is likely fairly close, closer than 80%.
One other thing to point out is that since these are TDP, if
an OS is using ACPI/HALT the CPU will be well below that
current in typical uses.
Another things some frown upon but is often possible with
the lower speed ~700MHz) chips is to reduce the vcore down
to about 1.5V, some might not remain stable but others might
even go lower but I doubt below 1.4V is attainable.
Except some potentially odd workstation or servers, any P3
board that lacks the 12V (or another aux. connector) should
not be using 12V for CPU.
5V was by far the most common, but since it was also used
for so many other things, it may be splitting hairs to nail
down a number of amps the CPU is using vs the cumulative
total for rest of the system. If power usage is really THAT
important, OP could just downclock the FSB, memory, and
expect a corresponding power usage decrease from these
subsystems as well, in addition to it being ever more likely
a CPU vcore reduction would be tolerated. If the vcore can
be lowered to about 1.5V or less it's fairly easy to run one
passively, given an appropriate PSU or duct to draw enough
air past the CPU... but such things (resultant temps) must
be measured in full load situations, something you probably
know but some would overlook it.
- » Exasperating graphics problem, motherboard related?
- — Next thread in » Home-built Computers