KTwo said:
AGP 4X runs at 1.5volts. 8X takes it even lower, down to .8V. But
since there is neither a 1.5V or a .8V lead from an ATX power supply,
the mobo must be doing the conversion on the board. Question is just
which of the three power sources does it us?
1.5V can be made with a linear regulator from +3.3V. The I/O pads
run from 1.5V, whether the AGP I/O signals are at 1.5V or 0.8V.
The 0.8V signal voltage happens when a voltage divider terminator
scheme is used - the driver pin swings 1.5V p-p, but the terminator
at the receiving end reduces the swing to 0.8V. So, there is no 0.8V
regulator on the board, only a 1.5V regulator. On a DDR memory based
board, the 1.5V for AGP can come from linear regulation using the 2.5V
for the DIMMs, and the DIMMs in turn can be linear regulated from 3.3V.
It must be expensive to make switching converters, because Asus loves
to use linear regulation schemes on new boards, using MOSFETs for
series pass elements.
My list of periphals is this . . .
120 GB HD
40 GB HD
48X CD-RW
IDE Zip Drive
GeForce3 Ti200 Video Card
Soundblaster Audigy
10/100 NIC
PCI LPT2 Card
512 MB RAM
ASUS TUSL2-C Mobo
While this totals to around 270W of power, if you look at the all these
pieces in context of which of the three power rails they draw from,
these all hit the 5V rail pretty hard. An Antec True330 (which I've
gone through two of) has 30A output on the 5V rail. The above
components, when you add in the mother board itself, total about 27A
current on the 5V rail. That's something like 92% of max capacity. No
power supply I know likes to get hit that hard. So, it worked for
awhile and slowly died.
The lesson I learned out of this is to look at all three voltages when
sizing a power supply, not just the total watts.
-K2
I've actually measured the input power to two of my computers.
This is current flowing into the PSU, multiplied by the measured
voltage on the line. (Phase angle unknown, and for simplicity,
assumed V and I are in phase.) As these are _input_ power to
the PSU, they must be reduced to about 70% of these numbers,
to get an idea of how much power is consumed inside the computer.
70% is a rough number for the efficiency of the average PSU.
On my TUA266 with 1.1GHz Tualatin: Idle=120W CPUBURN=132W
On my P4B with 1.8GHz P4: Idle=106W CPUBURN=160W
These computers had Radeon 8500 video card, 3x512MB SDRAM,
single disk drive, Ethernet card, Promise IDE controller,
CDROM reader.
I have trouble believing you are making it all the way to 270W.
An extra disk drive is 12W. The Tualatins were in the 35W ballpark,
if memory serves.
There are some other limits in your computer. For example,
the ATX 20 pin power connector has a 6A limit per pin. Since there
are four +5V pins, that is a 24A limit. There are three +3.3V pins,
for an 18A limit. The +12V has only one pin, giving a 6A limit.
The max power you could draw through the ATX connector on +3.3V
and +5V combined, is 180W.
When designers put together the "power train" for a board, they
are careful not to run all the switching converters off the same
supply. For example, if the Vcore has a switching converter, and
there is a second switcher for VIO (SDRAM VDIMM, AGP, PCI, chipset
voltage), the two of them will use different sources of power. One
will use +5V and the other +12V. For the reasons you state, that
they don't want to hammer a single rail.
I would venture to guess that something else is killing the
PSU. Otherwise, your ATX motherboard connector would be melted
or burned by now.
Another sign that the PSU is providing a lot of power, is
the PSU will be kicking out a lot of heat. Based on the 70%
efficiency, a fair amount of heat is dissipated by the PSU,
to make its output power. If the load is really drawing
lots of current, the PSU fan speed should pick up, and the
exhaust air should be hot.
One poster here holds the record for power dissipation. His
PSU got so hot, that he couldn't touch the metal on the PSU
casing. Now, you know a lot of current was going _somewhere_
on that computer.
If you can find one, I recommend a clamp-on DC ammeter, as
you can quickly grab bundles of wires and measure the total
current through them. For example, you can grab the four
+5V wires on the ATX 20 pin connector, slip the clamp around
the four wires, and the meter will read out the total current
in the four wires. I bought one for home use, and it was $400
CDN, the most expensive test equipment I own. All my other
meters are sub $100 units.
http://www.repaircalibration.com/380947.html
(I made my own line separator for measuring AC current. You
have to select one of the three wires in an AC cord, and put
it by itself in the jaws of the meter, as a clamp-on meter
works by measuring the magnetic field, and slipping the meter
around the whole AC cord will read exactly zero.)
HTH,
Paul
