Another Attempt, Dell Optiplex SX260 Computer Won't Boot Up

SNIP
With the CPU removed, there is no VID code. The VID code floats.
On a VRD I worked on, the all-1's VID code (float value)
is "zero volts", which causes the chip to switch off.
(That means, with the CPU pulled, VCore switches off.)
On the regulator chip, take logic level readings on the VID inputs.
Compare to the datasheet VID table, and you'll likely find at least
two codes in the table that cause the regulator to be effectively
switched off.

The logic levels on VID codes now, can be quite low levels.
Like perhaps 1.5V logic. Looking at the regulator datasheet,
there may be pullups on the VID lines, and then all you'd need
is to short an input to GND to put a logic zero on it. At one time,
VID lines were open collector, but the processor may use totem pole
drive on them now. Since the processor is not plugged in, you
can do your own driving of the line.

Does VCore need a dummy load on it ? Is it open circuit stable
and properly regulated at no-load ? I don't know the answer to
that. I might use a small array of carbon composition resistors
as a dummy load. Say, twenty 10 ohm 1/4W resistors in parallel,
which would be a 0.5 ohm load across 1 volt output, for 2 amps
flow. And I^2R would be less than the 20/4 = 5 watt rating
of your array. Salt to taste.

*******

As for your ohm meter reading across the processor, I'll not
"go near that with a barge pole". If you need to do such things,
compare the reading to a known-good processor. There are
too many intangibles (sneak paths inside a multi-rail chip),
to go guessing whether "5.5 ohms = busted". My pure guess
would be, it's not dead with a reading like that. If you
read 0.0, I'd be suspicious... The VCore regulator
could still drive such a load - your ohmmeter likely can't
read low values reliably, to determine whether the
processor would actually cause the overcurrent to trip
on the regulator.

I've only seen one good failure on a VCore. Someone posted a
picture once, of a motherboard where the area around the
processor socket was charred (discolored). And that's a
plane to plane short, where the short wasn't close enough
to zero ohms, to trip the overcurrent. And VCore just
made a toaster out of that area of the motherboard.
If the internal plane short had been better quality,
and closer to zero ohms, there would have been no
charring, and the regulator would have switched off.
They do have overcurrent detection, but it's set
very very high. It takes a damn good short to trip it.

Paul

Hi Paul,

Thanks for that information.

I don't know a lot regarding computer hardware. I use to repair TVs,
VCRs, stereos, etc., but not computers.

I failed to find a schematic for this model's mother board. I wonder if
there is a schematic for a "typical" mother board anywhere?

Thanks in advance, John

wrote:
SNIP
With the CPU removed, there is no VID code. The VID code floats.
On a VRD I worked on, the all-1's VID code (float value)
is "zero volts", which causes the chip to switch off.
(That means, with the CPU pulled, VCore switches off.)
On the regulator chip, take logic level readings on the VID inputs.
Compare to the datasheet VID table, and you'll likely find at least
two codes in the table that cause the regulator to be effectively
switched off.

The logic levels on VID codes now, can be quite low levels.
Like perhaps 1.5V logic. Looking at the regulator datasheet,
there may be pullups on the VID lines, and then all you'd need
is to short an input to GND to put a logic zero on it. At one time,
VID lines were open collector, but the processor may use totem pole
drive on them now. Since the processor is not plugged in, you
can do your own driving of the line.

Does VCore need a dummy load on it ? Is it open circuit stable
and properly regulated at no-load ? I don't know the answer to
that. I might use a small array of carbon composition resistors
as a dummy load. Say, twenty 10 ohm 1/4W resistors in parallel,
which would be a 0.5 ohm load across 1 volt output, for 2 amps
flow. And I^2R would be less than the 20/4 = 5 watt rating
of your array. Salt to taste.

*******

As for your ohm meter reading across the processor, I'll not
"go near that with a barge pole". If you need to do such things,
compare the reading to a known-good processor. There are
too many intangibles (sneak paths inside a multi-rail chip),
to go guessing whether "5.5 ohms = busted". My pure guess
would be, it's not dead with a reading like that. If you
read 0.0, I'd be suspicious... The VCore regulator
could still drive such a load - your ohmmeter likely can't
read low values reliably, to determine whether the
processor would actually cause the overcurrent to trip
on the regulator.

I've only seen one good failure on a VCore. Someone posted a
picture once, of a motherboard where the area around the
processor socket was charred (discolored). And that's a
plane to plane short, where the short wasn't close enough
to zero ohms, to trip the overcurrent. And VCore just
made a toaster out of that area of the motherboard.
If the internal plane short had been better quality,
and closer to zero ohms, there would have been no
charring, and the regulator would have switched off.
They do have overcurrent detection, but it's set
very very high. It takes a damn good short to trip it.

Paul

Hi Paul,

Thanks for that information.

I don't know a lot regarding computer hardware. I use to repair TVs,
VCRs, stereos, etc., but not computers.

I failed to find a schematic for this model's mother board. I wonder if
there is a schematic for a "typical" mother board anywhere?

Thanks in advance, John


Yes, there is one. There are two reference schematics I located, one
being for the Slot 1 era, the other is for S478 with the 875P dual channel
Northbridge. The latter is the one I'd suggest as a reference. One
problem with this schematic, is it does not use a "real" SuperI/O,
so lots of valuable info is unavailable from it. The other problem, is
the schematic is drawn in Mentor Graphics "stick font", which makes
the Acrobat search function useless. If you're designing at a company
with Mentor as your CAD system, you can change that to get "real"
text fonts, and I'm surprised the Intel staff haven't fixed theirs.
At my company, our "IT" type hardware manager, figured out how
to fix that, all on his own (the manuals for that crap, are a couple
feet thick and just the thought of wading through that gives me nausea).

(875P chipset reference schematic, File is 25281202.pdf.
Chip notable because of the extra (CSA) interface for a LAN chip.
In effect, it's almost like the Northbridge has enough interfaces
to support two Southbridges.)

http://developer.intel.com/design/ch...ics/252812.htm

You already have the datasheet for the 4 phase regulator on my
S478 / 875P motherboard. That's this one. Table 1 on page 7,
shows when the CPU is pulled from the socket, the all 1's
code on VID is expected to happen. It's a 6 bit code, and
you'd need anywhere from one to six jumpers stuffed into
the S478 socket to set a code other than the "float" value.
(I.e. Jumper a VID to logic 0 GND, if you need a zero on that signal.)

http://web.archive.org/web/200403310...5ADP3180_0.pdf

The logic levels on the VID, reading between the lines,
seem to be 0 to 1.25V. The VCore chip will tolerate up to
a 5.5V level on the pins (almost like it's TTL or CMOS
compatible), but the internal pullup on the pin, is to
a 1.0V internal source. So both the thresholds and
pullup, are centered around much lower voltages.
With the CPU pulled, all six VID lines will float to logic
1, and the VCore regulator will put out zero volts. If you
use grounding jumpers on some of the VID signals, you could
attempt to define a different value.

On PDF page 89 of 25281202.pdf, in the lower right hand corner,
you can see the VID bus is tied to VCC3 (3 volts) with 1K ohm
resistors. Which means, for whatever reason, Intel thinks they're
3V levels. If that were the answer, then using your multimeter,
the six VID signals, with the CPU pulled, would sit at around
3V to 3.3V or so. If the motherboard had no provision like that,
the regulator chip itself has internal pullups, and on the
regulator chip my board uses, they bias at 1.0V when the CPU is
missing. The 1K ohm pullups would "override" the weak pullups
inside the 4 phase regulator chip.

When a board won't run, the "RESET" is a key thing to check.
As a lot of initialization logic funnels into it. A motherboard
isn't allowed to "start", until a number of subsystems report
in that they're "ship-shape". If you have fans spinning, and
a black screen, anything that will jam RESET is enough to
prevent POST.

Good luck reading the schematic :-)

Paul