eMachine won't keep time

I have an eMachine here with Windows XP SP3 that has a severe time problem.
For instance, it lost 57 minutes in the time it took me to run defrag on it
which was about an hour and a half. I just rebooted it and it lost 10
minutes. There doesn't seem to be any set pattern. I just rebooted it again
and it only lost one minute. It seems to lose or gain randomly whether it
is rebooted or not.
Now I've discovered that in BIOS you can set the time but it does not move.
Presently set at 20:01:01. If you save and reboot to Windows the time shows
8:01PM. If you *don't* save then in Windows the time (like right now)
showed 8:25 PM. Go back into BIOS and it still shows 20:01:01. Actual time
as I post is 8:50 PM.
I put a new battery in it way back in the game. It ain't that. It's got a
PhoenixBIOS 4.0 Rlease 6.0 Rev 1.04. eMachine support says no updates
available. Phoenix led me off to a link at biosagentplus.com which
eventually led me to downloading biosagentplus_752.exe. That eventually
winds you up at http://mewnlite.com/biosagentplus.jpg which leads to
http://mewnlite.com/register.jpg . I did have enough good sense to stop at
that point.
Personally I don't believe there *is* a bios update and even if there was I
doubt seriously it would correct the time thing.
Is there any sense in pursuing it any further?
For what it's worth, it was loaded with MyWebSearch, Hotbar, and a few more
similar adware items, but no other types of malware. I doubt if that has
anything to do with it.

--
-- I'm out of white ink --

Menno Hershberger wrote:
I have an eMachine here with Windows XP SP3 that has a severe time problem.
For instance, it lost 57 minutes in the time it took me to run defrag on it
which was about an hour and a half. I just rebooted it and it lost 10
minutes. There doesn't seem to be any set pattern. I just rebooted it again
and it only lost one minute. It seems to lose or gain randomly whether it
is rebooted or not.
Now I've discovered that in BIOS you can set the time but it does not move.
Presently set at 20:01:01. If you save and reboot to Windows the time shows
8:01PM. If you *don't* save then in Windows the time (like right now)
showed 8:25 PM. Go back into BIOS and it still shows 20:01:01. Actual time
as I post is 8:50 PM.
I put a new battery in it way back in the game. It ain't that. It's got a
PhoenixBIOS 4.0 Rlease 6.0 Rev 1.04. eMachine support says no updates
available. Phoenix led me off to a link at biosagentplus.com which
eventually led me to downloading biosagentplus_752.exe. That eventually
winds you up at http://mewnlite.com/biosagentplus.jpg which leads to
http://mewnlite.com/register.jpg . I did have enough good sense to stop at
that point.
Personally I don't believe there *is* a bios update and even if there was I
doubt seriously it would correct the time thing.
Is there any sense in pursuing it any further?
For what it's worth, it was loaded with MyWebSearch, Hotbar, and a few more
similar adware items, but no other types of malware. I doubt if that has
anything to do with it.


You sure know how to have fun. biosagentplus is part of Phoenix, and esupport
has something to do with it as well.

http://en.wikipedia.org/wiki/Phoenix_Technologies

*******

Time keeping in Windows, is done by counting clock tick interrupts. The OS
makes a copy of the RTC (real time clock) and runs independently of the
crystal in the RTC until shutdown. That is why the clock can advance while
you're in Windows - it doesn't care about the RTC once the time is copied
to memory.

When the system is shut down, the RTC is relied on to keep time. It uses
a 32768 Hz digital watch crystal, strapped to the motherboard. That's what
times the ticks there.

The RTC and CMOS RAM block, are part of the Southbridge chip silicon die, and live
in "the well". The well is a semi-independent section of the chip, separated
from other subsystems by transmission gates. They do it that way, so no
leakage current, or backfeed, can move from one section of the chip to another.
After all, when all power is removed from the PC, "the well" runs off the CR2032
coin cell battery (10uA current draw). The rest of the Southbridge is unpowered,
with the exception of "the well".

So your RTC isn't accepting write cycles, it would appear. Power for the storage
elements, at the time of write attempt, would be 3VSB (which in turn, is regulated
from +5VSB).

Perhaps the PC is set up to do NTP and use an external time server, to set
the correct time, and that's how it is getting moved from 8:01. Since it is
still 8:01 when you return to the BIOS, then it means so far, nothing is able
to change the RTC time setting.

If you didn't have an NTP (network time protocol) process running in Windows, then
every time the PC boots, it should record the time as starting at 8:01 again.

The Southbridge is hard to replace, if that was indeed what was broken.
Could it be the power source for "the well" at fault ? Possibly. Just
for fun, you can stick a multimeter set to volts, across the CR2032 top
surface, and connect the other lead to grounded metal on the chassis. And
see if it is still 3.0V. You say it's new, but batteries can drain pretty
quickly if there is a fault. Very occasionally, someone will report that
their battery drains in a matter of days, which means there is a leakage
path there somewhere. The battery is not charged by the ATX supply, so cannot
be "pumped back up".

*******

I see no reason to change the BIOS at this point, because you don't have
any other anomalies except the clock. You would have noticed an inability
to set the clock, if that had happened when the machine was new. And since
the main body of the BIOS is protected by a checksum, there would likely be
an error message of some sort, if there was "bit rot" in there. The boot block
code, should be computing the checksum of the main BIOS code, before running it.

*******

You left out the most important part - Emachine model number info. Did you
Google the model number, and see if there were any other reports like yours ?

Paul

Paul wrote in :

Menno Hershberger wrote:
I have an eMachine here with Windows XP SP3 that has a severe time
problem. For instance, it lost 57 minutes in the time it took me to
run defrag on it which was about an hour and a half. I just rebooted
it and it lost 10 minutes. There doesn't seem to be any set pattern.
I just rebooted it again and it only lost one minute. It seems to
lose or gain randomly whether it is rebooted or not.
Now I've discovered that in BIOS you can set the time but it does not
move. Presently set at 20:01:01. If you save and reboot to Windows
the time shows 8:01PM. If you *don't* save then in Windows the time
(like right now) showed 8:25 PM. Go back into BIOS and it still shows
20:01:01. Actual time as I post is 8:50 PM.
I put a new battery in it way back in the game. It ain't that. It's
got a PhoenixBIOS 4.0 Rlease 6.0 Rev 1.04. eMachine support says no
updates available. Phoenix led me off to a link at biosagentplus.com
which eventually led me to downloading biosagentplus_752.exe. That
eventually winds you up at http://mewnlite.com/biosagentplus.jpg
which leads to http://mewnlite.com/register.jpg . I did have enough
good sense to stop at that point.
Personally I don't believe there *is* a bios update and even if there
was I doubt seriously it would correct the time thing.
Is there any sense in pursuing it any further?
For what it's worth, it was loaded with MyWebSearch, Hotbar, and a
few more similar adware items, but no other types of malware. I doubt
if that has anything to do with it.


You sure know how to have fun. biosagentplus is part of Phoenix, and
esupport has something to do with it as well.

http://en.wikipedia.org/wiki/Phoenix_Technologies

*******

Time keeping in Windows, is done by counting clock tick interrupts.
The OS makes a copy of the RTC (real time clock) and runs
independently of the crystal in the RTC until shutdown. That is why
the clock can advance while you're in Windows - it doesn't care about
the RTC once the time is copied to memory.

When the system is shut down, the RTC is relied on to keep time. It
uses a 32768 Hz digital watch crystal, strapped to the motherboard.
That's what times the ticks there.

The RTC and CMOS RAM block, are part of the Southbridge chip silicon
die, and live in "the well". The well is a semi-independent section of
the chip, separated from other subsystems by transmission gates. They
do it that way, so no leakage current, or backfeed, can move from one
section of the chip to another. After all, when all power is removed
from the PC, "the well" runs off the CR2032 coin cell battery (10uA
current draw). The rest of the Southbridge is unpowered, with the
exception of "the well".

So your RTC isn't accepting write cycles, it would appear. Power for
the storage elements, at the time of write attempt, would be 3VSB
(which in turn, is regulated from +5VSB).

Perhaps the PC is set up to do NTP and use an external time server, to
set the correct time, and that's how it is getting moved from 8:01.
Since it is still 8:01 when you return to the BIOS, then it means so
far, nothing is able to change the RTC time setting.

If you didn't have an NTP (network time protocol) process running in
Windows, then every time the PC boots, it should record the time as
starting at 8:01 again.

Yep... that's what it's doing. Ooops wait. I just checked the BIOS again
and it had advanced to 20:28.xx. But this was after 10 o'clock which should
have been 22:xx:xx. I went ahead and set it right again. Again just now I
looked at the windows time at it said 10:36 and it was actually 10:44. I
went into the time settings and synchronized it with time.windows.com and
it did jump up to 10:44. But if I look back in a few minutes it will be
behind again. There's some place in the registry where you can change the
frequency of the time checks. Maybe I should set it to every 10 minutes!


The Southbridge is hard to replace, if that was indeed what was
broken. Could it be the power source for "the well" at fault ?
Possibly. Just for fun, you can stick a multimeter set to volts,
across the CR2032 top surface, and connect the other lead to grounded
metal on the chassis. And see if it is still 3.0V. You say it's new,
but batteries can drain pretty quickly if there is a fault. Very
occasionally, someone will report that their battery drains in a
matter of days, which means there is a leakage path there somewhere.
The battery is not charged by the ATX supply, so cannot be "pumped
back up".

OK. Just did that. The new one pegged my meter on the 2.5VDC setting. Next
increment is 50... too hard to read. I checked the old battery and it has
less than half a volt left in it.

*******

I see no reason to change the BIOS at this point, because you don't
have any other anomalies except the clock. You would have noticed an
inability to set the clock, if that had happened when the machine was
new. And since the main body of the BIOS is protected by a checksum,
there would likely be an error message of some sort, if there was "bit
rot" in there. The boot block code, should be computing the checksum
of the main BIOS code, before running it.

*******

You left out the most important part - Emachine model number info. Did
you Google the model number, and see if there were any other reports
like yours ?

Yep. It's a T2042. It's got a 2.0 GB Celeron processor and over a gig of
memory. There's drivers on the eMachine support site (Gateway actually) for
most everything but under "BIOS" it says nothing available (or something
like that). I also checked on driversguide.com and didn't come up with
anything. The only place that hinted that it might be out of date was that
biosagentplus thing. Then, if you looked at my screenshots, when I clicked
on the bios update thing it wanted me to "register"
-----------------------
HOME OF THE LARGEST BIOS DIRECTORY ON THE INTERNET
* Get instant access to over a million driver files.
* Better PC Performance.
* Improved PC Stability.
* From a trusted source in the computer industry.
------------------------
Sounds kind of spammy to me and I'll bet they want $$$$ to become
"registered". If it's Phoenix and there are upgrades available, then they
should offer downloads for free.

As usual thanks for sharing your knowledge

--
-- I'm out of white ink --

Menno Hershberger wrote in
news:Xns9F67EBB622DAEbutter@wefb973cbe498:

Paul wrote in :

Menno Hershberger wrote:
I have an eMachine here with Windows XP SP3 that has a severe time
problem. For instance, it lost 57 minutes in the time it took me to
run defrag on it which was about an hour and a half. I just rebooted
it and it lost 10 minutes. There doesn't seem to be any set pattern.
I just rebooted it again and it only lost one minute. It seems to
lose or gain randomly whether it is rebooted or not.
Now I've discovered that in BIOS you can set the time but it does
not move. Presently set at 20:01:01. If you save and reboot to
Windows the time shows 8:01PM. If you *don't* save then in Windows
the time (like right now) showed 8:25 PM. Go back into BIOS and it
still shows 20:01:01. Actual time as I post is 8:50 PM.
I put a new battery in it way back in the game. It ain't that. It's
got a PhoenixBIOS 4.0 Rlease 6.0 Rev 1.04. eMachine support says no
updates available. Phoenix led me off to a link at biosagentplus.com
which eventually led me to downloading biosagentplus_752.exe. That
eventually winds you up at http://mewnlite.com/biosagentplus.jpg
which leads to http://mewnlite.com/register.jpg . I did have enough
good sense to stop at that point.
Personally I don't believe there *is* a bios update and even if
there was I doubt seriously it would correct the time thing.
Is there any sense in pursuing it any further?
For what it's worth, it was loaded with MyWebSearch, Hotbar, and a
few more similar adware items, but no other types of malware. I
doubt if that has anything to do with it.


You sure know how to have fun. biosagentplus is part of Phoenix, and
esupport has something to do with it as well.

http://en.wikipedia.org/wiki/Phoenix_Technologies

*******

Time keeping in Windows, is done by counting clock tick interrupts.
The OS makes a copy of the RTC (real time clock) and runs
independently of the crystal in the RTC until shutdown. That is why
the clock can advance while you're in Windows - it doesn't care about
the RTC once the time is copied to memory.

When the system is shut down, the RTC is relied on to keep time. It
uses a 32768 Hz digital watch crystal, strapped to the motherboard.
That's what times the ticks there.

The RTC and CMOS RAM block, are part of the Southbridge chip silicon
die, and live in "the well". The well is a semi-independent section
of the chip, separated from other subsystems by transmission gates.
They do it that way, so no leakage current, or backfeed, can move
from one section of the chip to another. After all, when all power is
removed from the PC, "the well" runs off the CR2032 coin cell battery
(10uA current draw). The rest of the Southbridge is unpowered, with
the exception of "the well".

So your RTC isn't accepting write cycles, it would appear. Power for
the storage elements, at the time of write attempt, would be 3VSB
(which in turn, is regulated from +5VSB).

Perhaps the PC is set up to do NTP and use an external time server,
to set the correct time, and that's how it is getting moved from
8:01. Since it is still 8:01 when you return to the BIOS, then it
means so far, nothing is able to change the RTC time setting.

If you didn't have an NTP (network time protocol) process running in
Windows, then every time the PC boots, it should record the time as
starting at 8:01 again.

Yep... that's what it's doing. Ooops wait. I just checked the BIOS
again and it had advanced to 20:28.xx. But this was after 10 o'clock
which should have been 22:xx:xx. I went ahead and set it right again.
Again just now I looked at the windows time at it said 10:36 and it
was actually 10:44. I went into the time settings and synchronized it
with time.windows.com and it did jump up to 10:44. But if I look back
in a few minutes it will be behind again. There's some place in the
registry where you can change the frequency of the time checks. Maybe
I should set it to every 10 minutes!


The Southbridge is hard to replace, if that was indeed what was
broken. Could it be the power source for "the well" at fault ?
Possibly. Just for fun, you can stick a multimeter set to volts,
across the CR2032 top surface, and connect the other lead to grounded
metal on the chassis. And see if it is still 3.0V. You say it's new,
but batteries can drain pretty quickly if there is a fault. Very
occasionally, someone will report that their battery drains in a
matter of days, which means there is a leakage path there somewhere.
The battery is not charged by the ATX supply, so cannot be "pumped
back up".

OK. Just did that. The new one pegged my meter on the 2.5VDC setting.
Next increment is 50... too hard to read. I checked the old battery
and it has less than half a volt left in it.

*******

I see no reason to change the BIOS at this point, because you don't
have any other anomalies except the clock. You would have noticed an
inability to set the clock, if that had happened when the machine was
new. And since the main body of the BIOS is protected by a checksum,
there would likely be an error message of some sort, if there was
"bit rot" in there. The boot block code, should be computing the
checksum of the main BIOS code, before running it.

*******

You left out the most important part - Emachine model number info.
Did you Google the model number, and see if there were any other
reports like yours ?

Yep. It's a T2042. It's got a 2.0 GB Celeron processor and over a gig
of memory. There's drivers on the eMachine support site (Gateway
actually) for most everything but under "BIOS" it says nothing
available (or something like that). I also checked on driversguide.com
and didn't come up with anything. The only place that hinted that it
might be out of date was that biosagentplus thing. Then, if you looked
at my screenshots, when I clicked on the bios update thing it wanted
me to "register"
-----------------------
HOME OF THE LARGEST BIOS DIRECTORY ON THE INTERNET
* Get instant access to over a million driver files.
* Better PC Performance.
* Improved PC Stability.
* From a trusted source in the computer industry.
------------------------
Sounds kind of spammy to me and I'll bet they want $$$$ to become
"registered". If it's Phoenix and there are upgrades available, then
they should offer downloads for free.

As usual thanks for sharing your knowledge


I'm a sucker for punishment so I went ahead and filled out that
registration screen just to see what would happen.
Sure enough, http://mewnlite.com/shurnuf.jpg
$29.95. No thanks.



--
-- I'm out of white ink --

Menno Hershberger wrote:
Paul wrote in :

Menno Hershberger wrote:
I have an eMachine here with Windows XP SP3 that has a severe time
problem. For instance, it lost 57 minutes in the time it took me to
run defrag on it which was about an hour and a half. I just rebooted
it and it lost 10 minutes. There doesn't seem to be any set pattern.
I just rebooted it again and it only lost one minute. It seems to
lose or gain randomly whether it is rebooted or not.
Now I've discovered that in BIOS you can set the time but it does not
move. Presently set at 20:01:01. If you save and reboot to Windows
the time shows 8:01PM. If you *don't* save then in Windows the time
(like right now) showed 8:25 PM. Go back into BIOS and it still shows
20:01:01. Actual time as I post is 8:50 PM.
I put a new battery in it way back in the game. It ain't that. It's
got a PhoenixBIOS 4.0 Rlease 6.0 Rev 1.04. eMachine support says no
updates available. Phoenix led me off to a link at biosagentplus.com
which eventually led me to downloading biosagentplus_752.exe. That
eventually winds you up at http://mewnlite.com/biosagentplus.jpg
which leads to http://mewnlite.com/register.jpg . I did have enough
good sense to stop at that point.
Personally I don't believe there *is* a bios update and even if there
was I doubt seriously it would correct the time thing.
Is there any sense in pursuing it any further?
For what it's worth, it was loaded with MyWebSearch, Hotbar, and a
few more similar adware items, but no other types of malware. I doubt
if that has anything to do with it.

You sure know how to have fun. biosagentplus is part of Phoenix, and
esupport has something to do with it as well.

http://en.wikipedia.org/wiki/Phoenix_Technologies

*******

Time keeping in Windows, is done by counting clock tick interrupts.
The OS makes a copy of the RTC (real time clock) and runs
independently of the crystal in the RTC until shutdown. That is why
the clock can advance while you're in Windows - it doesn't care about
the RTC once the time is copied to memory.

When the system is shut down, the RTC is relied on to keep time. It
uses a 32768 Hz digital watch crystal, strapped to the motherboard.
That's what times the ticks there.

The RTC and CMOS RAM block, are part of the Southbridge chip silicon
die, and live in "the well". The well is a semi-independent section of
the chip, separated from other subsystems by transmission gates. They
do it that way, so no leakage current, or backfeed, can move from one
section of the chip to another. After all, when all power is removed
from the PC, "the well" runs off the CR2032 coin cell battery (10uA
current draw). The rest of the Southbridge is unpowered, with the
exception of "the well".

So your RTC isn't accepting write cycles, it would appear. Power for
the storage elements, at the time of write attempt, would be 3VSB
(which in turn, is regulated from +5VSB).

Perhaps the PC is set up to do NTP and use an external time server, to
set the correct time, and that's how it is getting moved from 8:01.
Since it is still 8:01 when you return to the BIOS, then it means so
far, nothing is able to change the RTC time setting.

If you didn't have an NTP (network time protocol) process running in
Windows, then every time the PC boots, it should record the time as
starting at 8:01 again.

Yep... that's what it's doing. Ooops wait. I just checked the BIOS again
and it had advanced to 20:28.xx. But this was after 10 o'clock which should
have been 22:xx:xx. I went ahead and set it right again. Again just now I
looked at the windows time at it said 10:36 and it was actually 10:44. I
went into the time settings and synchronized it with time.windows.com and
it did jump up to 10:44. But if I look back in a few minutes it will be
behind again. There's some place in the registry where you can change the
frequency of the time checks. Maybe I should set it to every 10 minutes!

The Southbridge is hard to replace, if that was indeed what was
broken. Could it be the power source for "the well" at fault ?
Possibly. Just for fun, you can stick a multimeter set to volts,
across the CR2032 top surface, and connect the other lead to grounded
metal on the chassis. And see if it is still 3.0V. You say it's new,
but batteries can drain pretty quickly if there is a fault. Very
occasionally, someone will report that their battery drains in a
matter of days, which means there is a leakage path there somewhere.
The battery is not charged by the ATX supply, so cannot be "pumped
back up".

OK. Just did that. The new one pegged my meter on the 2.5VDC setting. Next
increment is 50... too hard to read. I checked the old battery and it has
less than half a volt left in it.
*******

I see no reason to change the BIOS at this point, because you don't
have any other anomalies except the clock. You would have noticed an
inability to set the clock, if that had happened when the machine was
new. And since the main body of the BIOS is protected by a checksum,
there would likely be an error message of some sort, if there was "bit
rot" in there. The boot block code, should be computing the checksum
of the main BIOS code, before running it.

*******

You left out the most important part - Emachine model number info. Did
you Google the model number, and see if there were any other reports
like yours ?

Yep. It's a T2042. It's got a 2.0 GB Celeron processor and over a gig of
memory. There's drivers on the eMachine support site (Gateway actually) for
most everything but under "BIOS" it says nothing available (or something
like that). I also checked on driversguide.com and didn't come up with
anything. The only place that hinted that it might be out of date was that
biosagentplus thing. Then, if you looked at my screenshots, when I clicked
on the bios update thing it wanted me to "register"
-----------------------
HOME OF THE LARGEST BIOS DIRECTORY ON THE INTERNET
* Get instant access to over a million driver files.
* Better PC Performance.
* Improved PC Stability.
* From a trusted source in the computer industry.
------------------------
Sounds kind of spammy to me and I'll bet they want $$$$ to become
"registered". If it's Phoenix and there are upgrades available, then they
should offer downloads for free.

As usual thanks for sharing your knowledge


If it's like eSupport, they'll want money for your free upgrade. If
the BIOS file comes from the manufacturer, that would normally be free.
eSupport used to sell generic BIOS of some sort (the people who bought
from them, never made clear to me, whether the BIOS interface stayed the same
or not). And the sales people would actually negotiate price (kinda like
high pressure salesmen). And that to me was the tipoff. Via "chat" somebody
had the salesman convince them to buy at half price.

*******

I see some pictures of the motherboards used in T2042 here.

http://www.e4allupgraders.info/dir1/...imperial.shtml

I'd probably be booting something like Ubuntu, and testing time keeping there,
to rule out an OS software issue.

On one of my Linux VMs, I discovered something called sntp. A command like

sntp -s time.nrc.ca

would set the clock in Linux immediately (as long as Linux uses DHCP to set
up networking for you). In one of my Ubuntu virtual machines,
there is something called "rdate", which doesn't work for me at all. The
date/time application (looks like a calendar) also has access to NTP protocol,
but only seems to update the time from the Internet, once in a blue moon.
Which is why I was investigating other options. For a hardware guy like me,
the sntp one was great, because it worked immediately, and would also print out
some delta information. But it isn't listed in all the package managers, so I
can't guarantee you can easily find it. And compiling it from source, would require
you to find the source for it.

*******

Timekeeping while Windows is running, is done by counting clock tick interrupts.
The clock tick interrupt has a relatively high priority, which is to prevent
other interrupts from affecting time keeping. Servicing of that interrupt
would have to be stopped for more than one tick, in order to lose one, and
make the (software) clock slow down.

There have been chipsets with a hardware problem, where the time wandered all
over the place. And it had to do with a bug in the interrupt logic. I'm not
aware of anything like that for an Intel chipset.

So anyway, give a Linux LiveCD a try, and see if the clock screws up there
or not. If it's a hardware issue, it should be every bit as bad in Linux.

*******

And 2.5V for the battery, isn't that good. If that was the real value,
there isn't a lot of headroom between 2.5V and 2.3V (as low as I'd want to go),
and with the computer unplugged, it might only last a matter of weeks. The
Southbridge is rated for 2.0V, but doesn't get the full battery voltage,
due to 0.3V drop across a Schottky diode. The diode is there, to prevent
the motherboard from pumping charge into the battery (which would cause it
to burst). So the battery has to supply 2.3V, so the Southbridge can see 2.0V
(its minimum allowed for accurate functioning).

(ATX plugged in)
3VSB ----------------- diode -- ------+
|
+----- Southbridge (well)
| RTC and 256 byte CMOS RAM block
CR2032 --- 1000 ohm ---diode -- ------+
resistor

Measuring the voltage where it feeds the SB would be nice... But
who can find that junction ? I've only managed it once or twice.
I doubt I could find it on my current motherboard. To find the
BAT45 dual diode (looks like a flat transistor with three legs), you
need a magnifying glass to verify the part number. And some
motherboards have a ton of things that look like that.

HTH,
Paul

Menno Hershberger wrote:

I'm a sucker for punishment so I went ahead and filled out that
registration screen just to see what would happen.
Sure enough, http://mewnlite.com/shurnuf.jpg
$29.95. No thanks.

I'd have dug out my credit card, if only Oprah had approved of it :-)

Paul

Paul wrote in :

Menno Hershberger wrote:
Paul wrote in :

Menno Hershberger wrote:
I have an eMachine here with Windows XP SP3 that has a severe time
problem. For instance, it lost 57 minutes in the time it took me to
run defrag on it which was about an hour and a half. I just
rebooted it and it lost 10 minutes. There doesn't seem to be any
set pattern. I just rebooted it again and it only lost one minute.
It seems to lose or gain randomly whether it is rebooted or not.
Now I've discovered that in BIOS you can set the time but it does
not move. Presently set at 20:01:01. If you save and reboot to
Windows the time shows 8:01PM. If you *don't* save then in Windows
the time (like right now) showed 8:25 PM. Go back into BIOS and it
still shows 20:01:01. Actual time as I post is 8:50 PM.
I put a new battery in it way back in the game. It ain't that. It's
got a PhoenixBIOS 4.0 Rlease 6.0 Rev 1.04. eMachine support says no
updates available. Phoenix led me off to a link at
biosagentplus.com which eventually led me to downloading
biosagentplus_752.exe. That eventually winds you up at
http://mewnlite.com/biosagentplus.jpg which leads to
http://mewnlite.com/register.jpg . I did have enough good sense to
stop at that point. Personally I don't believe there *is* a bios
update and even if there was I doubt seriously it would correct the
time thing. Is there any sense in pursuing it any further?
For what it's worth, it was loaded with MyWebSearch, Hotbar, and a
few more similar adware items, but no other types of malware. I
doubt if that has anything to do with it.

You sure know how to have fun. biosagentplus is part of Phoenix, and
esupport has something to do with it as well.

http://en.wikipedia.org/wiki/Phoenix_Technologies

*******

Time keeping in Windows, is done by counting clock tick interrupts.
The OS makes a copy of the RTC (real time clock) and runs
independently of the crystal in the RTC until shutdown. That is why
the clock can advance while you're in Windows - it doesn't care
about the RTC once the time is copied to memory.

When the system is shut down, the RTC is relied on to keep time. It
uses a 32768 Hz digital watch crystal, strapped to the motherboard.
That's what times the ticks there.

The RTC and CMOS RAM block, are part of the Southbridge chip silicon
die, and live in "the well". The well is a semi-independent section
of the chip, separated from other subsystems by transmission gates.
They do it that way, so no leakage current, or backfeed, can move
from one section of the chip to another. After all, when all power
is removed from the PC, "the well" runs off the CR2032 coin cell
battery (10uA current draw). The rest of the Southbridge is
unpowered, with the exception of "the well".

So your RTC isn't accepting write cycles, it would appear. Power for
the storage elements, at the time of write attempt, would be 3VSB
(which in turn, is regulated from +5VSB).

Perhaps the PC is set up to do NTP and use an external time server,
to set the correct time, and that's how it is getting moved from
8:01. Since it is still 8:01 when you return to the BIOS, then it
means so far, nothing is able to change the RTC time setting.

If you didn't have an NTP (network time protocol) process running in
Windows, then every time the PC boots, it should record the time as
starting at 8:01 again.

Yep... that's what it's doing. Ooops wait. I just checked the BIOS
again and it had advanced to 20:28.xx. But this was after 10 o'clock
which should have been 22:xx:xx. I went ahead and set it right again.
Again just now I looked at the windows time at it said 10:36 and it
was actually 10:44. I went into the time settings and synchronized it
with time.windows.com and it did jump up to 10:44. But if I look back
in a few minutes it will be behind again. There's some place in the
registry where you can change the frequency of the time checks. Maybe
I should set it to every 10 minutes!

The Southbridge is hard to replace, if that was indeed what was
broken. Could it be the power source for "the well" at fault ?
Possibly. Just for fun, you can stick a multimeter set to volts,
across the CR2032 top surface, and connect the other lead to
grounded metal on the chassis. And see if it is still 3.0V. You say
it's new, but batteries can drain pretty quickly if there is a
fault. Very occasionally, someone will report that their battery
drains in a matter of days, which means there is a leakage path
there somewhere. The battery is not charged by the ATX supply, so
cannot be "pumped back up".

OK. Just did that. The new one pegged my meter on the 2.5VDC setting.
Next increment is 50... too hard to read. I checked the old battery
and it has less than half a volt left in it.
*******

I see no reason to change the BIOS at this point, because you don't
have any other anomalies except the clock. You would have noticed an
inability to set the clock, if that had happened when the machine
was new. And since the main body of the BIOS is protected by a
checksum, there would likely be an error message of some sort, if
there was "bit rot" in there. The boot block code, should be
computing the checksum of the main BIOS code, before running it.

*******

You left out the most important part - Emachine model number info.
Did you Google the model number, and see if there were any other
reports like yours ?

Yep. It's a T2042. It's got a 2.0 GB Celeron processor and over a gig
of memory. There's drivers on the eMachine support site (Gateway
actually) for most everything but under "BIOS" it says nothing
available (or something like that). I also checked on
driversguide.com and didn't come up with anything. The only place
that hinted that it might be out of date was that biosagentplus
thing. Then, if you looked at my screenshots, when I clicked on the
bios update thing it wanted me to "register"
-----------------------
HOME OF THE LARGEST BIOS DIRECTORY ON THE INTERNET
* Get instant access to over a million driver files.
* Better PC Performance.
* Improved PC Stability.
* From a trusted source in the computer industry.
------------------------
Sounds kind of spammy to me and I'll bet they want $$$$ to become
"registered". If it's Phoenix and there are upgrades available, then
they should offer downloads for free.

As usual thanks for sharing your knowledge


If it's like eSupport, they'll want money for your free upgrade. If
the BIOS file comes from the manufacturer, that would normally be
free. eSupport used to sell generic BIOS of some sort (the people who
bought from them, never made clear to me, whether the BIOS interface
stayed the same or not). And the sales people would actually negotiate
price (kinda like high pressure salesmen). And that to me was the
tipoff. Via "chat" somebody had the salesman convince them to buy at
half price.

*******

I see some pictures of the motherboards used in T2042 here.

http://www.e4allupgraders.info/dir1/...8/imperial.sht
ml

I'd probably be booting something like Ubuntu, and testing time
keeping there, to rule out an OS software issue.

On one of my Linux VMs, I discovered something called sntp. A command
like

sntp -s time.nrc.ca

would set the clock in Linux immediately (as long as Linux uses DHCP
to set up networking for you). In one of my Ubuntu virtual machines,
there is something called "rdate", which doesn't work for me at all.
The date/time application (looks like a calendar) also has access to
NTP protocol, but only seems to update the time from the Internet,
once in a blue moon. Which is why I was investigating other options.
For a hardware guy like me, the sntp one was great, because it worked
immediately, and would also print out some delta information. But it
isn't listed in all the package managers, so I can't guarantee you can
easily find it. And compiling it from source, would require you to
find the source for it.

*******

Timekeeping while Windows is running, is done by counting clock tick
interrupts. The clock tick interrupt has a relatively high priority,
which is to prevent other interrupts from affecting time keeping.
Servicing of that interrupt would have to be stopped for more than one
tick, in order to lose one, and make the (software) clock slow down.

There have been chipsets with a hardware problem, where the time
wandered all over the place. And it had to do with a bug in the
interrupt logic. I'm not aware of anything like that for an Intel
chipset.

So anyway, give a Linux LiveCD a try, and see if the clock screws up
there or not. If it's a hardware issue, it should be every bit as bad
in Linux.

*******

And 2.5V for the battery, isn't that good.

But but but.... I said it pegged the needle when set at the 2.5 range. It
clicked at the top actually a little past 2.5. I'm sure it would have hit 3
if there was any more room... :-) It was going where the yellow line is at
http://mewnlite.com/plink.jpg - you can see how hard it would be to read
the 50 scale. I can't afford one of those fancy digital meters for no more
use than I would get out of it.

If that was the real value,
there isn't a lot of headroom between 2.5V and 2.3V (as low as I'd
want to go), and with the computer unplugged, it might only last a
matter of weeks. The Southbridge is rated for 2.0V, but doesn't get
the full battery voltage, due to 0.3V drop across a Schottky diode.
The diode is there, to prevent the motherboard from pumping charge
into the battery (which would cause it to burst). So the battery has
to supply 2.3V, so the Southbridge can see 2.0V (its minimum allowed
for accurate functioning).



(ATX plugged in)
3VSB ----------------- diode -- ------+
|
+----- Southbridge (well)
| RTC and 256 byte
CMOS RAM block
CR2032 --- 1000 ohm ---diode -- ------+
resistor

Measuring the voltage where it feeds the SB would be nice... But
who can find that junction ? I've only managed it once or twice.
I doubt I could find it on my current motherboard. To find the
BAT45 dual diode (looks like a flat transistor with three legs), you
need a magnifying glass to verify the part number. And some
motherboards have a ton of things that look like that.

HTH,
Paul






--
-- I'm out of white ink --

Menno Hershberger wrote:


But but but.... I said it pegged the needle when set at the 2.5 range. It
clicked at the top actually a little past 2.5. I'm sure it would have hit 3
if there was any more room... :-) It was going where the yellow line is at
http://mewnlite.com/plink.jpg - you can see how hard it would be to read
the 50 scale. I can't afford one of those fancy digital meters for no more
use than I would get out of it.

They make a model, with slightly better spacing. $22

http://images10.newegg.com/NeweggIma...001-522-01.jpg

Since your meter is pegged, that's good enough for now. It's over 2.5V
but we don't know by how much. It's probably not the problem right now.

On with the testing...

Paul

Paul wrote in :

Menno Hershberger wrote:


But but but.... I said it pegged the needle when set at the 2.5
range. It clicked at the top actually a little past 2.5. I'm sure it
would have hit 3 if there was any more room... :-) It was going
where the yellow line is at http://mewnlite.com/plink.jpg - you can
see how hard it would be to read the 50 scale. I can't afford one of
those fancy digital meters for no more use than I would get out of
it.

They make a model, with slightly better spacing. $22

http://images10.newegg.com/NeweggIma...001-522-01.jpg

Since your meter is pegged, that's good enough for now. It's over 2.5V
but we don't know by how much. It's probably not the problem right
now.

On with the testing...

Damn, when do you sleep? :-)

--
-- 2:15 AM in Iowa --

Menno Hershberger wrote:


Damn, when do you sleep? :-)

I'm asleep right now.

Paul