Post problem with Z77H2-A3 motherboard

Bob F wrote:

The BIOS seems to default to running at the 1600MHz the RAM is rated
for. If I change memory clk multiplier to manual at 10 instead of 12, it
will now boot every time at 1333 MHz. It does still do the double start,
like it would do if overclocked too much when I was last using it full
time.

I did not previously have to do any of this slowing things down. I am
beginning to think this board is just fading significantly.

I just read your latest comment, and it gives me a lot more to think
about. The DMI rewrite thing could be involved. Could it be that the DMI
is failing and replacing the BIOS chip might clean up my problems?

Another step backwards. Now, I just discovered that the Motherboard is
not shutting off the power supply when it shuts down. This is really
getting frustrating. Resetting the BIOS to default does nothing for this
one.

Well, that one would be a gate level hardware problem.
Like an open collector driver that drives PS_ON# signal
on the ATX main cable.

Since it's open collector, it supports wired-OR operation,
and if a second person was working on the project, they
could have left a shorting jumper (PS_ON# to ground)
on the thing. A pullup resistor, normally pulls
PS_ON# to +5VSB voltage level, when the motherboard
stops trying to drive PS_ON# low. The "#" means
"active low". When the voltage level on PS_ON#
is below 0.8V or so, the ATX supply starts spinning
the fan and delivering 3.3/5/12/-12 rails. If the
PS_ON# voltage rises to 5VSB level, the ATX supply
main section goes off and the fan stops spinning.
The +5VSB should remain running at all times, for
this interface to work of course, as +5VSB powers
the gates on the motherboard looking for that
control signal.

So when it doesn't shut off, it means something
is pulling the signal down, closer to zero volts
than to the logic one value of +5VSB volts.

You can't "drive it high", as this is Open Collector
logic, and only a 2K to 10K ohm resistor might be pulling
it up.

You would need to place a multimeter on PS_ON# to
scope out the problem. You can shove a multimeter
probe into the back of the main cable nylon shroud
and touch the conductive metal on PS_ON#. And take
a reading.

While it could be related to an ACPI/APM kind of
problem, I kind of doubt that would appear out
of the blue like that. I'd start by checking the
PS_ON# to see what kind of swing is present.

One thing I've never understood, is how older
motherboards worked. The pullup resistor scheme,
doesn't require the motherboard end to sink more
than 2mA of current (to make a logic zero and
switch on the PSU). Yet, motherboards have
24mA or 48mA drive capability (knuckle dragger
74F07 driver chips) in those cases. What were they
expecting the PSU to do ??? I couldn't figure
out the level of overengineering on that signal.
It's like the motherboard designers had their
doubts about the PSU designers design credentials :-)

And yet, PS_ON# failures seem quite common. Why ?
I just don't get it. There's no reason for that
interface to be "flake city", and yet that particular
failure stands out. The PSU receiving end, does not use
a logic gate, and uses a transistor-based
substitute. Even though today, in 2020, you can
get 5 pin miniature single logic gates, that
could do a spec compliant TTL level interface
on the thing. The engineers making PSUs
prefer their transistor bodge instead.

If another motherboard had been able to turn the
PSU on and off properly ("soft off"), then this
would implicate the current motherboard as having
a "driving low" problem on PS_ON#. If every motherboard
you connect to the PSU, cannot gain proper control,
the PSU is then implicated (bodge transistor circuit
failure).

Paul

On 4/2/2020 2:41 AM, Paul wrote:
Bob F wrote:

The BIOS seems to default to running at the 1600MHz the RAM is rated
for. If I change memory clk multiplier to manual at 10 instead of 12,
it will now boot every time at 1333 MHz. It does still do the double
start, like it would do if overclocked too much when I was last using
it full time.

I did not previously have to do any of this slowing things down. I am
beginning to think this board is just fading significantly.

I just read your latest comment, and it gives me a lot more to think
about. The DMI rewrite thing could be involved. Could it be that the
DMI is failing and replacing the BIOS chip might clean up my problems?

Another step backwards. Now, I just discovered that the Motherboard is
not shutting off the power supply when it shuts down. This is really
getting frustrating. Resetting the BIOS to default does nothing for
this one.

Well, that one would be a gate level hardware problem.
Like an open collector driver that drives PS_ON# signal
on the ATX main cable.

Since it's open collector, it supports wired-OR operation,
and if a second person was working on the project, they
could have left a shorting jumper (PS_ON# to ground)
on the thing. A pullup resistor, normally pulls
PS_ON# to +5VSB voltage level, when the motherboard
stops trying to drive PS_ON# low. The "#" means
"active low". When the voltage level on PS_ON#
is below 0.8V or so, the ATX supply starts spinning
the fan and delivering 3.3/5/12/-12 rails. If the
PS_ON# voltage rises to 5VSB level, the ATX supply
main section goes off and the fan stops spinning.
The +5VSB should remain running at all times, for
this interface to work of course, as +5VSB powers
the gates on the motherboard looking for that
control signal.

So when it doesn't shut off, it means something
is pulling the signal down, closer to zero volts
than to the logic one value of +5VSB volts.

You can't "drive it high", as this is Open Collector
logic, and only a 2K to 10K ohm resistor might be pulling
it up.

You would need to place a multimeter on PS_ON# to
scope out the problem. You can shove a multimeter
probe into the back of the main cable nylon shroud
and touch the conductive metal on PS_ON#. And take
a reading.

While it could be related to an ACPI/APM kind of
problem, I kind of doubt that would appear out
of the blue like that. I'd start by checking the
PS_ON# to see what kind of swing is present.

One thing I've never understood, is how older
motherboards worked. The pullup resistor scheme,
doesn't require the motherboard end to sink more
than 2mA of current (to make a logic zero and
switch on the PSU). Yet, motherboards have
24mA or 48mA drive capability (knuckle dragger
74F07 driver chips) in those cases. What were they
expecting the PSU to do ??? I couldn't figure
out the level of overengineering on that signal.
It's like the motherboard designers had their
doubts about the PSU designers design credentials :-)

And yet, PS_ON# failures seem quite common. Why ?
I just don't get it. There's no reason for that
interface to be "flake city", and yet that particular
failure stands out. The PSU receiving end, does not use
a logic gate, and uses a transistor-based
substitute. Even though today, in 2020, you can
get 5 pin miniature single logic gates, that
could do a spec compliant TTL level interface
on the thing. The engineers making PSUs
prefer their transistor bodge instead.

If another motherboard had been able to turn the
PSU on and off properly ("soft off"), then this
would implicate the current motherboard as having
a "driving low" problem on PS_ON#. If every motherboard
you connect to the PSU, cannot gain proper control,
the PSU is then implicated (bodge transistor circuit
failure).

Â*Â* Paul

This problem showed up while I had a second power supply hooked up to
it, but carried over to the original when I went back to it. I checked
the voltage at PS_ON#, and it is 5V after I power the supply up. It goes
to 0V when I push the power button, and stays there from then on after
shut down. So, it seems like the actual switching works in the power on
direction but it is not reversing on shutdown. I am leaning towards
giving up on this board, and going back to the other board I have with
flakey USB 3.0 ports. I did a little looking, and these boards are a lot
more expensive than I thought old boards should be.

Bob F wrote:


This problem showed up while I had a second power supply hooked up to
it, but carried over to the original when I went back to it. I checked
the voltage at PS_ON#, and it is 5V after I power the supply up. It goes
to 0V when I push the power button, and stays there from then on after
shut down. So, it seems like the actual switching works in the power on
direction but it is not reversing on shutdown. I am leaning towards
giving up on this board, and going back to the other board I have with
flakey USB 3.0 ports. I did a little looking, and these boards are a lot
more expensive than I thought old boards should be.

You could try adding a pullup resistor to the circuit.

+5VSB
|
4.7K ohm resistor
|
Mobo --------+----------------- PS_IN# (PSU)

What that does, is inject a max of 1mA of current
into that signal, and see if the motherboard end
is releasing properly when the intent is to
soft power off.

A sample Intel document I found, seems to have some
idea how the stuff works. The drive capability is 8mA.
A 1K pullup resistor is recommended in the SuperIO
datasheet, whereas the 865G doc uses a 22K (weak) pullup,
which surely is getting close to a minimum for that
point in the circuit. If a 1K ohm resistor (recommended
by SuperIO datasheet) were to be used in place of the 22K,
that would be a sinking current of 5mA. The 4.7K testing
resistor would bring that to 6mA, leaving 2mA to drive
the pullup inside the PSU.
5VSB 5VSB 5VSB
| | |
ICH/PCH SuperIO 22K 4.7K (PU)
SLP_S3# | | |
PWRBTN# ---+ +-----------------+ OD8 +-----+------+------ (PSU) --+
CPU_PRESENT# PS_ON#
---------------+

CPUSLP# ---

The path to switch off the power, is a bugger to trace.
It almost looks like the CPU executes a HALT, that might
assert CPUSLP#, turns into SLP_S3# (because there's some
sequencing going on to make a soft landing), and the
SuperIO then stops driving low on PS_ON#. Even if the
sequenced soft landing fails, you still get to see
"it is safe to turn off this PC" on the screen. And failures
of this sort are blamed on a broken ACPI subsystem in software.
But because I can't find a description of what actually
happens, this is just a guess from looking at signal names.
I see no evidence (yet), of an actual register controlling
the powering off thing.

On the older schematic, instead of a SuperIO performing
the buffer task, the designer used a 74F07 open collector
jelly bean logic chip for the final drive. In other words,
something with more than OD8 (open drain 8 milliamp)
capability, and not going to take "no" for an answer
when it engages. The designer who did that, that's
"virtue signaling" in the engineering profession,
and says "I know something but I'm not telling you" :-)

The CPU_PRESENT type signal, is a static signal asserted
by the CPU (a strap condition), which tells the motherboard
that a CPU is inserted in the socket. It doesn't say the
CPU works, and it's present to prevent the PWRBTN from
working if the motherboard has no CPU installed.

It's hard to believe a proper sized resistor wasn't
placed on PS_ON#. And as far as I know, the power supply
end has its own resistor (because the PSU must behave properly if
sitting on a table, and someone plugs in the power
cord - it can't take off because it lacked a pullup).

*******

You could go to Device Manager while the system is running,
and see if the "Computer" item in device manager, the HAL
value is "ACPI Uniprocessor PC" or "ACPI Multiprocessor PC",
as there is also the possibility of some non-ACPI value.
It's possible in the BIOS, after you've finished fiddling
with the CMOS CR2032 battery, you disabled the ACPI
setting and damaged the HAL value in Device Manager.

On more modern OSes, it might repair this on its own when
the BIOS setting is corrected. With Windows XP, there
were some "horrid procedures" for fixing the damage
caused by starting a system with the BIOS ACPI set
improperly. For example, on some computers here, the
desired setting is "ACPI 2.0" [Enabled] versus
the not-so-nice value of [Disabled]. The older version
of ACPI may not be sufficient to keep any sort of
modern Windows happy. I think I had Win2K refuse to
install, with the [Disabled] value set by accident.

I can't really see a good explanation in terms of the
pullup resistor thing.

Perhaps your evaluation order should be:

1) Check Device Manager "Computer" entry for correct value.
Remedy by checking BIOS setup in Power menu, ACPI.
If the HAL (Hardware Abstraction Layer) label is wrong,
that could account for your "new symptom".

2) If the HAL appears correct, now work on your pullup resistor.
Not really good odds this will be indicative, but it's
what I'd try before "trashing my setup" and moving on.

If it's (1), you don't have to panic :-)

Paul

On Thu, 2 Apr 2020 12:51:56 -0700, Bob F wrote:

I did a little looking, and these boards are a lot
more expensive than I thought old boards should be.

-
They're sometimes actually not, but can be a result of supply and
demand. The last board I dealt in quadrupled its price once Taiwan
facilities began the transition for manufacture cessation. Less
reputability comes to point, as it became destined an OEM item for a
likes of Ebay and secondary-tier sellers. The next manufacture
release is of course bound to the next CPU socket, fewer if any driver
support than for Microsoft dominance, along with shrinking MB real
estate, fewer firmware features, slots and plugs, at smaller form
factors. All of less of nothing, a potential consideration, for more
money at the budget platform entry, effectively in sole dependence on
Windows 10. What that makes old about boards now, to say somewhat
demandingly tedious, is that they're more complicated an affair
required to conclusively discern for widest software compatibility
support (leaving fractional *NIX platforms) at an opportune price.
Handhelds and proprietary services have seen to as much that vision,
one for certain inconveniences if betoken of olden, where personal
computers consequently adjust to fit perhaps a resultant residual of
niche items.

On 4/2/2020 9:05 PM, Paul wrote:
Bob F wrote:


This problem showed up while I had a second power supply hooked up to
it, but carried over to the original when I went back to it. I checked
the voltage at PS_ON#, and it is 5V after I power the supply up. It
goes to 0V when I push the power button, and stays there from then on
after shut down. So, it seems like the actual switching works in the
power on direction but it is not reversing on shutdown. I am leaning
towards giving up on this board, and going back to the other board I
have with flakey USB 3.0 ports. I did a little looking, and these
boards are a lot more expensive than I thought old boards should be.

You could try adding a pullup resistor to the circuit.

Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* +5VSB
Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* |
Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* 4.7K ohm resistor
Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* |
Â*Â* Mobo --------+----------------- PS_IN# (PSU)

What that does, is inject a max of 1mA of current
into that signal, and see if the motherboard end
is releasing properly when the intent is to
soft power off.

A sample Intel document I found, seems to have some
idea how the stuff works. The drive capability is 8mA.
A 1K pullup resistor is recommended in the SuperIO
datasheet, whereas the 865G doc uses a 22K (weak) pullup,
which surely is getting close to a minimum for that
point in the circuit. If a 1K ohm resistor (recommended
by SuperIO datasheet) were to be used in place of the 22K,
that would be a sinking current of 5mA. The 4.7K testing
resistor would bring that to 6mA, leaving 2mA to drive
the pullup inside the PSU.
Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* 5VSB
5VSBÂ*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* 5VSB
Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* |
|Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* |
Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* ICH/PCHÂ*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* SuperIOÂ*Â*Â*Â* 22K
4.7KÂ*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* (PU)
Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* SLP_S3#Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* |
|Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* |
Â*PWRBTN# ---+Â*Â*Â*Â* +-----------------+Â*Â* OD8 +-----+------+------
(PSU) --+
Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* CPU_PRESENT#Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* Â*Â*Â*Â*Â*Â* PS_ON#
Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â*Â* ---------------+

Â*CPUSLP# ---

The path to switch off the power, is a bugger to trace.
It almost looks like the CPU executes a HALT, that might
assert CPUSLP#, turns into SLP_S3# (because there's some
sequencing going on to make a soft landing), and the
SuperIO then stops driving low on PS_ON#. Even if the
sequenced soft landing fails, you still get to see
"it is safe to turn off this PC" on the screen. And failures
of this sort are blamed on a broken ACPI subsystem in software.
But because I can't find a description of what actually
happens, this is just a guess from looking at signal names.
I see no evidence (yet), of an actual register controlling
the powering off thing.

On the older schematic, instead of a SuperIO performing
the buffer task, the designer used a 74F07 open collector
jelly bean logic chip for the final drive. In other words,
something with more than OD8 (open drain 8 milliamp)
capability, and not going to take "no" for an answer
when it engages. The designer who did that, that's
"virtue signaling" in the engineering profession,
and says "I know something but I'm not telling you" :-)

The CPU_PRESENT type signal, is a static signal asserted
by the CPU (a strap condition), which tells the motherboard
that a CPU is inserted in the socket. It doesn't say the
CPU works, and it's present to prevent the PWRBTN from
working if the motherboard has no CPU installed.

It's hard to believe a proper sized resistor wasn't
placed on PS_ON#. And as far as I know, the power supply
end has its own resistor (because the PSU must behave properly if
sitting on a table, and someone plugs in the power
cord - it can't take off because it lacked a pullup).

*******

You could go to Device Manager while the system is running,
and see if the "Computer" item in device manager, the HAL
value is "ACPI Uniprocessor PC" or "ACPI Multiprocessor PC",
as there is also the possibility of some non-ACPI value.
It's possible in the BIOS, after you've finished fiddling
with the CMOS CR2032 battery, you disabled the ACPI
setting and damaged the HAL value in Device Manager.

On more modern OSes, it might repair this on its own when
the BIOS setting is corrected. With Windows XP, there
were some "horrid procedures" for fixing the damage
caused by starting a system with the BIOS ACPI set
improperly. For example, on some computers here, the
desired setting is "ACPI 2.0" [Enabled] versus
the not-so-nice value of [Disabled]. The older version
of ACPI may not be sufficient to keep any sort of
modern Windows happy. I think I had Win2K refuse to
install, with the [Disabled] value set by accident.

I can't really see a good explanation in terms of the
pullup resistor thing.

Perhaps your evaluation order should be:

1) Check Device Manager "Computer" entry for correct value.
Â*Â* Remedy by checking BIOS setup in Power menu, ACPI.
Â*Â* If the HAL (Hardware Abstraction Layer) label is wrong,
Â*Â* that could account for your "new symptom".

2) If the HAL appears correct, now work on your pullup resistor.
Â*Â* Not really good odds this will be indicative, but it's
Â*Â* what I'd try before "trashing my setup" and moving on.

If it's (1), you don't have to panic :-)


Paul and Flasherly

Thank you both for your thoughtful consideration of my problems here. I
have, for the moment set the Z77 board aside, and gone back to the less
problematic Z68 board which does work other than the USB 3 problem.
Since I only have the one 1155 processor, I will not be able to continue
with debugging the Z77 board at this time. You both gave me plenty to
think about on this project, and I appreciate the education.

Bob

On Mon, 13 Apr 2020 20:23:49 -0700, Bob F wrote:

Since I only have the one 1155 processor, I will not be able to continue
with debugging the Z77 board at this time. You both gave me plenty to
think about on this project, and I appreciate the education.

Last Intel I ran was a nice experience and the first time, apart from
richer processor costs, I went with over a spread of Intel processors
supported on that brand motherboard. A motherboard purchased to build
with a new Intel dual core, and the first time subsequently I'd
considered a supply of used processors. Three or four times I went
back to used processors, from that support list, eventually to max it
out with the fastest Intel listed. Each time as used processors
dropped to a fraction of their original values. The first new Intel
was probably $50 and a budget model. The rest, afterwards, I didn't
have to pay usually more than $20 a processor.

EBAY business pulls. Look them through and some are respectable
enough. Look a little more distantly, generally in such a market,
which I did on an AMD platform since for a budget AMD processor I
bought twice to get the best value for performance from the first
purchased misapplied, but later corrected from South Korea, my first
time that distant, among sellers of used processors. Overall, I've
been surprised how very nice very little cost-outlay can be for
options indeed available in a secondary CPU market.

On Tue, 14 Apr 2020 05:49:45 -0400, Flasherly
wrote:

Intel CPU updates over a course of 2011-2015 --

http://stores.ebay.com/greenITexchange/Processors
Coreâ„¢ 2 Duo E6850 $12 free ship

ebay
Intel Pentium D 805 2.66 GHz LGA 775 CPU SL8ZH 2M/533 dual core 64-bit
10.95 shipped

ebay Item number361026204698
INTEL CORE 2 DUO PROCESSOR E6850 3GHz/4M/1333MHz SLA9U CPU
Qty: 1 $11.95 Standard Shipping FREE
tax $0.84 tot 12.79

ebay INTEL CORE 2 QUAD Q8200 SLG9S 2.33GHZ 4MB CACHE 1333MHZ CPU
Qty: 1 $24.99 USPS First Class Package FREE
Sold by ci-elec

--
Intel board replaced since and most recent, including corrected AMD
already mentioned --

ebay s korea
Sold by gtron7 Jan 2019
AMD FX-Series FX-4100 Zambezi FD4100WMW4KGU Socket AM3+ 95W CPU
Quantity 1 Item 192079749758
Order total Shipping Free
$24.50

(
I didn't like this
Amd Phenom II X4 810 - 2.6Ghz (Used/Pull)
$29.95
and replaced it with the above FX-4100
[arguably, combined, cost of an AMD-3+ hexcore]
)