Re: ATTN: Paul... ASUS P4C800-E

On Tue, 12 Jul 2011 00:07:30 -0400, Paul <nospam@xxxxxxxxxx> wrote:

Steve Sr. wrote:
Paul,

I don' know if you saw my earlier post or not. The message ID is
below. I would appreciate your input or ideas as a current owner of
one of these boards. Sorry it is a long post.

<ibrk07h3gkpvj0sdjs45grspnm4irmj3hi@xxxxxxx>


Thanks,

Steve

******* copy of <ibrk07h3gkpvj0sdjs45grspnm4irmj3hi@xxxxxxx> posting *******

Paul & Other Asus Gurus,

I have a system built around a P4C800-E motherboard. It is not
overclocked and has been reliable for a number of years. Recently it
developed a video noise problem which appear to be the video card
which I replaced with an EVGA 512MB 8X AGP unit.

during the video card replacement I noticed some bad caps on the
motherboard so replaced all of the main electrolytics with Nichicon HZ
series replacements. On re-powering the system I noticed an
intermittent burning smell emminating from the power supply, an Antec
True Power 430. I opened the supply but didn't see anything obviously
burnt. I did see some more bad caps in the output section of the
supply but since the PC was working I put it all back together and
began researching a new power supply.

4-5 weeks elapsed before receiving the new supply. During this time
the system continued to work flawlessly. The chosen replacement supply
is a Seasonic by Corsair TX-650 V2. Immediately upon installing this
supply I began to experience all kinds of weird POST failures.
However, before blaming the power supply I actually put an
oscilloscope on it to look at the ramp up of voltages and the PWR_OK
signal. All of this checked out to the ATX specification and the
in-circuit voltages were almost perfect when measured with a good
digital meter.

The initial symptoms were black screen, no beeps. The same may happen
on a subsequent power up or eventually it will POST but give a bogus
"Overclocking Failed!" message. From this point either restoring BIOS
defaults or just entering and exiting the setup without changing
anything will result in booting to Windows. Once it gets to Windows it
is rock-solid reliable until the next cold boot (usually the next
day). Another interesting thing is that once everything is "warmed up"
it will POST and boot properly until it is allowed to cool off usually
overnight. As far as component temperatures go the CPU and GPU are
both at about 50C. Hard drives are about 27-29C. This is with the
tower case sitting non-optimally horizontal on my desk.

This system has 2-1GB sticks of Corsair ValuSelect RAM (Upgrade 2008)
and 2-512MB sticks of Crucial/Micron RAM (Original 2004). Thinking
that this might somehow be memory related I began swapping memory
sticks around. Things got really strange with ONLY the Micron RAM
installed, first both sticks and then each individually. The main
error now was "BIOS flash checksum bad" and prompts to insert a
diskette or CD with the BIOS code on it! NOTE: This is not the same
error as "BIOS Checksum Bad" which usually applies to the setup data
being corrupted. Installing the corsair RAM first one stick and then 2
allowed proper POSTing and booting. Thinking that this Crucial RAM was
somehow bad I got an RMA and sent it back to Crucial. BTW, once POSTed
this system ran MemTest86 overnight with no failures with BOTH memory
sets in it.

The joy didn't last long! The next day I was back to intermittent POST
issues. However, it seemed that it didn't take as much "warm up" time
as before for the system would boot normally.

Using SpeedFan or CPUZ I noticed that the CPU core voltage was all of
the way at the bottom at 1.56 volts. I read a post that suggested it
be raised a little so I bumped it up to 1.66 volts. At first this
seemed to have a beneficial effect at least getting through the POST
more often. However now there are some new symptoms. Now if the board
makes it through the POST it may fail to identify one or BOTH hard
drives and prompt for bootable media. At this point the reset switch
will cause a re-POST but the drives may or may not be picked up on the
re-POST. Turning the power supply completely off (rear panel switch)
thereby removing standby power for 30 seconds or so will likely allow
both drives to be picked up on the next POST.

So do any of you have any more thoughts on how to proceed to
troubleshoot and fix this? Clearly all of the hardware including the
new power supply can't ALL be bad. Troubleshooting to this point
appears to point to a memory sensitivity but I don't see the
connection with the new power supply. I have seen a few web reports of
similar symptoms with this same MB but no conclusive cause and/or fix.

If I had to guess it would appear that the BIOS is too agressive at
setting memory perhaps before even configuring it to the setup data.
When this issue first appeared I tried turning off "Performance Mode"
in the BIOS. As I recall the system then POSTed but Win 2K threw an
"IRQ not less than or equal" Stop error on 2 consecutive boots so I
turned it back off. Apparently this was some sort of memory driver
error that Windows couldn't handle.

Your thoughts?


Thanks,

Steve
***********************8**** End copy *****************************

You've changed a few things in your system, and when I saw your
posting, I decided to wait and see.

First of all, there is this

"main electrolytics with Nichicon HZ"

If you used those for VCore, you have to be careful to replace
"like with like". For example, say there were seven middle-of-the-road
ESR caps originally in the design. If you look at the design
equations (using the datasheet for the regulator chip, which
in this case is an Analog Devices part), the characteristics
of the caps are all part of the design. If you were to slap
a couple OSCONs in place of the original seven "average" caps,
the circuit would no longer be centered. So when replacing
caps, you at least should be consulting the VCore regulator
datasheet, for any words of wisdom. Perhaps it makes
no difference to the circuit, but the datasheet will provide
guidance. This is one reason, you don't want the ESR to be
"zero", simply because the design takes ESR into account
in its own way. The staff at Analog Design, want their
regulator to be able to use cheap caps, and so living
with ESR is all part of that. Cranking the caps, after
the fact, would require re-working the equations, and see
if any resistors in the feedback chain etc., need to be changed.

That datasheet has an awe inspiring collection of equations :-)

Nichicon HZ capacitors are designed for PC motherboard Vcore usage.
You are correct that ESR can affect regulator stability but there are
two things that appear to rule out MB capacitor issues.

1. The MB worked flawlessly with the new caps for 5 weeks before
installing the new power supply.

2. The system is solid as a rock in Win2k or Memtest 86... If it gets
past the POST. I would Vcore expect stability issues to cause other
problems.

Seeing "1.56V" is virtually perfect. This assumes a 0.13u
processor with 1.500V nominal VID setting, plus the "usual
Asus overvolt of 60 millivolts". I have several Asus
motherboards that do something similar - when the processor
is idle, the VCore voltage rests about 60mV above nominal.
So I wouldn't immediately assume it needed help.

Droop in a circuit like that is also normal, and accounted
for in the load line. If you saw the measured value dip by
0.15V under full load, there wouldn't be a need to be alarmed.

I can't remember clearly now, but this might be the board
that has an anomaly in the VCore boost. If you use the
highest settings, the voltage boost is added with GPIO
bits somewhere (since the VID table in the datasheet,
doesn't go high enough to account for the BIOS setting).
In my case, I think the highest boost actually goes backwards
(so the VCore settings aren't monotonically increasing, at
the highest settings). This is only a problem, if the
BIOS VCore setting is higher than what you find
in the Analog Devices regulator datasheet (>1.600V ???).
I'm just going from memory here.

Now, to the power supply. I measured my P4C800-E Deluxe
years ago, with a clamp on ammeter. This is what I got
under load.

3.3V @ 14.4 amps (4 DIMMs in dual channel mode)
5.0V @ 0.56 amps (not much mobo 5V load evident)
12.0V @ 0.43 amps (fan headers + GD75232 chip)

12.0V @ 5.62 amps (Northwood VCore on ATX12V 2x2)

Virtually any modern supply should be able to handle that.
An exception, is early Seasonic supplies, which had a
weakness in one of the low voltage rails. The Seasonic in
question was only good for about half the rated current
on the label (one rail tended to droop at high load).
Later Seasonic designs, seemed to fix this. At least,
I haven't seen complaints about it since. You'd have to
be pretty unlucky, to combine that Seasonic with your
motherboard. (No, I didn't keep records of which one.
But it's one of their early high efficiency ones.)

To check for that, measure your 3.3V at the motherboard.
Most ATX supplies have the "remote sense" wire on one
of the 3.3V pins on the main connector, which allows very
good tracking of 3.3V levels.

Part of that 3.3V current, would be for my video card, and
your video card could be different. The AGP spec allows
something like 6 amps on that rail ? That would presumably
be a component part of the 14.4 amps.

The 3.3 volt rail on the new supply measures a near perfect 3.362V in
with a good digital meter. It also looks perfect (low noise) on an
oscilloscope.


Now, Asus boards of that era, they tended to do linear
regulators, in a chain. 3.3V would be used to make
2.5V for the DIMMs. The 2.5V would then be run
through another linear stage, to make 1.5V or whatever
is needed for Vnb or Vsb and so on. I didn't see switchers
on there, neither did I see three terminal or five terminal
linears. It looks like they used home-brew regulators
based on op-amps.

If your low voltage was slightly weak, perhaps the
linear chain isn't working quite right, and that might
help account for the disks not being detected. (If they
were SATA disks, and some low voltage rail was used
for the SATA I/O pads. It wouldn't be as good
a theory for the IDE ribbon cable interface, as that
could be powered from something else.)

Now, when I tried to trace the regulation chain on
my motherboard, I "got lost". I was not able to make
sense out of the voltages I was seeing on the legs of
the MOSFETs. They didn't make sense. And yet, there weren't
any switchers in the chain. I could see op-amps strategically
placed, and MOSFETs, and my assumption was, they were
being used to build linear regulators.

As you can see, I'm hard pressed to come up with a
credible theory that explains all the symptoms. Which
is the reason I didn't write this post in the first place.
I wanted to wait and see if some other fact would come
along, to help things fall into place.

Oh, another thing. The

"BIOS flash checksum bad" and prompts to insert a
diskette or CD...

That hardly ever is caused by the data in the flash
chip being wrong :-) But I suspect you know that.
That's usually a sign there is something wrong with
the BIOS chip reading process. Just another flaky
symptom, to add to your pool of symptoms. On
motherboards without clock locking, just running
the clock on the EEPROM too fast, is enough to do it.

You've got so many symptoms, it's hard to cast a
net over them and tie them to one specific thing...

Paul

I think that the common thread in all of these symptoms is that the
BIOS appears to be getting lost during POST and not executing properly
hence causing all of these weird and seemingly unrelated errors.

Here is another tidbit. This issue appears to be temperature related
and only happens when the system is cold and has been off for several
hours, usually overnight. After the system has been turned on for
more than a couple of minutes it will pass POST repeatedly. This is
making it extremely hard to troubleshoot as I only get one opportunity
for a few minutes per day.

The symptom is also VERY sensitive to the type of memory modules
installed. The original Crucial/Micron RAM bought in 2003 performs
MUCH worse in POST than the Corsair ValuSelect upgrade bought in 2008.
The Micron RAM takes considerably longer to "warm up" and generates a
greater variety of errors than the Corsair. The Corsair is in it now
and in the last 3 cold boots it POSTed once without incident and the
other two times it only took a single push of the front panel reset
switch (after it failed the first time) to bring it to life.

I think I may have found a smoking gun... While looking at CPUZ to
answer another question I noticed a memory timing discrepancy. BTW, I
have overclocking set to STD and Performance Mode (which supposedly
optimizes memory timing) also set to STD.

CPUZ is showing memory timing set to 200MHz, 2.5-3-3-6. However, if
one looks at the SPD timing for the same memory at 200MHz it should be
3-3-3-8 at 2.5V. Any idea why in the H... the BIOS is setting these
aggressive memory timings when Performance Mode is turned OFF? Is CPUZ
accurate?

Now why all of this happened when I replaced a failing power supply is
beyond me. My only speculation is that the voltage levels on the old
supply may have been a little different due to regulation differences.
I can just about guarantee that the ripple was much higher on the old
supply. This may have gotten through the regulator chain and may have
gotten rectified and boosted the effective operating voltage of the
RAM and hidden this issue until now.

Thanks for your input. The light bulb is starting to illuminate!

Steve
.

Relevant Pages

  • Re: ATTN: Paul... ASUS P4C800-E
    ... I did see some more bad caps in the output section of the ... From this point either restoring BIOS ... you at least should be consulting the VCore regulator ... the VCore voltage rests about 60mV above nominal. ...
    (alt.comp.periphs.mainboard.asus)
  • Re: ATTN: Paul... ASUS P4C800-E
    ... From this point either restoring BIOS ... that this might somehow be memory related I began swapping memory ... the VCore voltage rests about 60mV above nominal. ... device supplying Vtt needs to be able to source or sink up ...
    (alt.comp.periphs.mainboard.asus)
  • Re: Puzzling Problems -- Pulling Hair
    ... These fancy memory sticks have blinky LEDs on top. ... Is it possible I've been killing my memory with bad BIOS settings? ... The one setting I changed from default was the DRAM Frequency item in BIOS. ... BIOS seems to indicate that with everything set to AUTO and N.O.S. not specifically turned on, the system will sense the increased voltage demand from the memory set to 1066, and will adjust accordingly. ...
    (alt.comp.periphs.mainboard.asus)
  • Re: Memory voltage on P8H67-M EVO
    ... Should I manually set the voltage in the BIOS to 1.5v or leave on ... I test my RAM, before I use it for real work, and using the various ... from me (and the memory wasn't exotic, ...
    (alt.comp.periphs.mainboard.asus)
  • Re: Memory setting of AsusŪ A8N SLI Premium ??
    ... >>> What is your DDR Voltage set to in the BIOS setup? ... It should be 2.7 Volts or 2.8 Volts depending on ... >>> When I originally installed the memory I had forgotten to change the DDR ...
    (alt.comp.periphs.mainboard.asus)