Can one believe the measurements shown by 'Speccy'?

[The Other John]

CPU CORE 0.816 V
MEMORY CONTROLLER 1.500 V
+3.3V 2.052 V
+5V 3.105 V
+12V 8.016 V
-12V -8.880 V
-5V -8.880 V
+5V HIGH THRESHOLD 2.782 V
CMOS BATTERY 1.452 V

These look hopelessly misleading because surely the PC wouldn't work with
some of those voltages and/or there would be signs of 'magic smoke'! It's
working fine except a couple of times booting up from cold it fails to
recognise the keyboard but a reboot cures it. This happens with 2
different k/bs, both wired PS2 types. Could it be the CMOS battery? It's
less than 3 years old and they normally last a lot longer than that.

--
TOJ.

[The Other John]

CPU CORE 0.816 V

MEMORY CONTROLLER 1.500 V

+3.3V 2.052 V

+5V 3.105 V

+12V 8.016 V

-12V -8.880 V

-5V -8.880 V

+5V HIGH THRESHOLD 2.782 V

CMOS BATTERY 1.452 V

Sorry, it lost the formatting. Should make more sense above.

--
TOJ.

[Jaimie Vandenbergh]

The CMOS battery won't affect any of this.

Misinterpretation of mobo info is rife in PC land, there's simply too
many ways to get it wrong that it's sometimes only by chance that the
right multipliers/divisions are applied in the software!

Boot into the BIOS, that'll probably have a voltages page showing
accurately interpreted signals.

Cheers - Jaimie
--
The person who says it cannot be done should not interrupt the person doing it.

[Johnny B Good]

It looks like the speccy s/w has 'lost the plot' in interpreting the
voltage readings for this MoBo. Let's assume for now that the 5v line is
actually within tolerance (often shows as 5.05v but we'll assume 5v dead
on the nose). If we divide by the speccy reading of 3.105v to create a
correction factor this gives us 1.6103059 to save in our calculator's
memory to simplify the following calculations for the other 'skewed'
voltage readouts. Let's name this correction factor "Z".

CPU CORE 0.816 V *Z = 1.314v Looks reasonable, if a little high
for a modern CPU

MEMORY CONTROLLER 1.500 V *Z = 2.4154588v Looks plausible.

+3.3V 2.052 V * Z = 3.3043477v Looks spot on to me.

+5V 3.105 V * Z = 4.9999998v (proves validity of "Z" for +ve values)

+12V 8.016 V * Z = 12.908212v is within tolerance, if a little high.

-12V -8.880 V V * Z = -14.299516v This looks bogus

-5V -8.880 V V * Z = -14.299516v This is definitely bogus!

+5V HIGH THRESHOLD 2.782 V V * Z = 4.479871v WTF is this?! [1]

CMOS BATTERY 1.452 V V * Z = 2.3381641v [2]

Our correction factor looks good for the positive readouts but poor old
speccy has well and truly 'lost it' when it comes to the negative ones.
Luckily, a cheap DMM will let you measure the actual voltages for
yourself to confirm the bogusity of speccy's half arsed attempt at
reporting this particular MoBo's voltages. In any case, you can look at
the 'Health' section in the CMOS setup menu to see what this reports (and
even compare those with 'actual voltages' as provided by a cheap DMM...
if you have one, that is! Beg, borrow or steal comes to mind if you
haven't).

[1] Honestly, WTF *is* a "+5V HIGH THRESHOLD" supposed to mean? Anyone?

[2] If this is accurate, then the RTC would run slow or even stop during
periods of absence of mains power if you're in the habit of 'pulling the
plug' in between sessions of use, otherwise the 5VSB line keeps things
'ticking over' removing this load from the Lithium coin cell ("CMOS
Battery"). It's enough to still maintain the 'CMOS settings' though. The
risk of data corruption in the RTC's registers doesn't raise its ugly
head until the voltage drops below the two volt mark.

Usually, it's easy enough, with most desktop PCs, to verify the coin
cell voltage with a DMM. A new coin cell is cheap enough these days
(typically a CR2032 - you can buy a card with 2 or 3 on in most pound
shops). Don't be put off if the BB date if it's this (or even last) years
date, that just means you've only got a minimum of 90% of the new
capacity left. Whoopee, Bloody Do! :-) Even if there are only two useful
coin cells in the pack, they're still good value for money for your one
pound spend.

--
Johnny B Good

[The Other John]

That's an interesting reply JBG, I hadn't thought of your 'correction
factor' approach. You made a slip with the -5V calculation, but in any
case I seem to remember the -ve rails aren't critical to a certain extent.

I have a couple of DMMs and I can check the CR2032 and the rails but I was
looking for pointers before diving inside, mainly because I'm usually the
victim of a 'pingfuckit' whenever I open anything up! :(

--
TOJ.

[Johnny B Good]

On Fri, 16 Feb 2018 20:49:31 +0000, The Other John wrote:

> On Fri, 16 Feb 2018 17:41:08 +0000, Johnny B Good wrote:

====snip====

>> reporting this particular MoBo's voltages. In any case, you can look at
>> the 'Health' section in the CMOS setup menu to see what this reports

====snip====

>
> That's an interesting reply JBG, I hadn't thought of your 'correction
> factor' approach. You made a slip with the -5V calculation, but in any
> case I seem to remember the -ve rails aren't critical to a certain
> extent.

I can't see how I slipped up on the -5v calculation. Speccy had
apparently reported the same erroneous voltage for the -12v line
according to your post.

It's true that the tolerances on the negative rails aren't quite as
critical as the tolerances required for the 3.3 and 5 volt rails (+/-5%
), probably on a par with the 12v rail which is specced to a +/-10%
tolerance. Also, it's been a while since ATX motherboards have actually
needed a -5v rail and the latest ATX PSUs lack such a rail since modern
motherboards and PCI/PCIe adapters just feed the -12v to a voltage
converter if anything requires a -5v supply (ditto for any on-board
components that require a -5v or other negative voltage level).

The -5v supply used to be used for negative biassing supplies which only
required the voltage with virtually no current draw such as in the case
of the early 1Kbit and 4Kbit DRAM chips which had a -5v bias pin (the
later 16Kbit chips did away for the need of such a pin by generating the
required negative bias voltage in the chip itself at the same time as
doing away with the need for a 12v pin as a result of reduced transistor
size which allowed the 5v pin to then supply all the positive voltages,
neatly freeing up two pins for extra address lines allowing a 16 pin
package to still be used on DRAM chips that could now address 16 times
more memory than what the the 4Kbit chips had been limited to.

Since "Pin Count" was a major driver of 'cost', the price of these
larger capacity chips swiftly dropped down to that of the older 4Kbit
chips and carried on dropping as demand drove mass production.

Think negative grid bias battery as used by the early thermionic valves
(vacuum tubes) which biassed the grid circuits which drew no current at
all but throttled the anode current down to the desired value in
amplification applications such as battery powered portable radios.

The grid bias battery only had to supply voltage to what was effectively
an open circuit. The voltage merely had to be present to achieve its
effect without the need to supply any current whatsoever and the same
condition applied to those early DRAM chips.

About the only things that required current were line driver chips used
by serial ports and even then, the demand was only a matter of milliamps
per driver chip.

With modern compact HF switching voltage converter technology, there's
no real need to supply any negative voltages from the main PSU so even
the -12v half amp rail could disappear from the ATX spec. Its presence
today is merely to satisfy legacy requirements which will eventually
disappear when motherboards no longer support PCI expansion slots (I'm
not sure but I suspect that PCIe has already removed the -12v requirement
from its specification - I'd be very surprised if it's still in the spec
- compact HF switching converters have been around longer than PCIe).

>
> I have a couple of DMMs and I can check the CR2032 and the rails but I
> was looking for pointers before diving inside, mainly because I'm
> usually the victim of a 'pingfuckit' whenever I open anything up! :(

I've never seen any PC cursed by the presence of pingfuckits. Removing
the cover(s) off any desktop PC is generally a fairly trivial task,
typically involving nothing more than a cross head screwdriver if the
case isn't of a 'toolless' design. In any case, as I mentioned in my
follow up post, you can enter the cmos setup menu and select the 'system
health' menu which, as well as showing fan speeds and temperatures, will
also show all the various voltages that Speccy so spectacularly failed to
correctly display, including voltages that can't be readily checked out
with a DMM without some risk of harm to the motherboard.

--
Johnny B Good

[Robin]

On 16/02/2018 20:49, The Other John wrote:
> I'm usually the
> victim of a 'pingfuckit' whenever I open anything up! :(
>

You need to eat more bran.






I'll get my coat.

--
Robin
reply-to address is (intended to be) valid

[The Other John]

On Sat, 17 Feb 2018 05:53:33 +0000, Johnny B Good wrote:

> I can't see how I slipped up on the -5v calculation. Speccy had
> apparently reported the same erroneous voltage for the -12v line
> according to your post.

Sorry, you're right. I was suffering brain-fart!

> I've never seen any PC cursed by the presence of pingfuckits. Removing
>the cover(s) off any desktop PC is generally a fairly trivial task,
>typically involving nothing more than a cross head screwdriver if the case
>isn't of a 'toolless' design. In any case, as I mentioned in my follow up
>post, you can enter the cmos setup menu and select the 'system health'
>menu which, as well as showing fan speeds and temperatures, will also show
>all the various voltages that Speccy so spectacularly failed to correctly
>display, including voltages that can't be readily checked out with a DMM
>without some risk of harm to the motherboard.

Opening the case is no problem, easing the clip off the CMOS battery could
well end up with damage knowing my luck. The cmos setup system health
section only shows core voltage and memory supply.

However, it's gone a couple of days now without losing the keyboard so I
think I'll just let sleeping dogs lie. Thanks for your input.


--
TOJ.