BIOS setting "Memory over voltage" shows value in red

On Tue, 09 Mar 2010 19:33:09 -0500, Paul wrote:

But since both Prime95 fails and so does memtest86+, then
there is likely a real problem there.

But if Prime95 fails, doesn't that indicate a problem somewhere -
regardless of what memtest says?

In CPUZ memory tab, is your command rate 2T ? Based on your memtest
errors, I might try a 2T setting for the command rate, if the BIOS
hasn't already selected that. Maybe if the memory was running 1T,
that would account for some locations simply not getting written.

What are the implications of changing this value, please?
(I.e. what does 'command rate' govern?!)
Could such a change (to 2T) screw the system?

In the meantime, the supplier has replied as follows:

"Please run the same test using MemTest+ 3.4a as there is an issue
with Version 4 of MemTest throwing false errors."

I could find no "memtest+" 3.4a version, so I shall try "memtest"3.4a.

( I saw a reference to "memtest" not being the same application as
"memtest+". What is the significance of this difference?
I also saw mention of 'spurious errors when using memtest 3.5 on
4GB or more'.

So first I shall try the memtest 3.4a and see how it gets on...

--
/\/\aurice
(Replace "nomail.afraid" by "bcs" to reply by email)

Maurice Batey wrote:
On Tue, 09 Mar 2010 19:33:09 -0500, Paul wrote:

But since both Prime95 fails and so does memtest86+, then
there is likely a real problem there.

But if Prime95 fails, doesn't that indicate a problem somewhere -
regardless of what memtest says?

In CPUZ memory tab, is your command rate 2T ? Based on your memtest
errors, I might try a 2T setting for the command rate, if the BIOS
hasn't already selected that. Maybe if the memory was running 1T,
that would account for some locations simply not getting written.

What are the implications of changing this value, please?
(I.e. what does 'command rate' govern?!)
Could such a change (to 2T) screw the system?

In the meantime, the supplier has replied as follows:

"Please run the same test using MemTest+ 3.4a as there is an issue
with Version 4 of MemTest throwing false errors."

I could find no "memtest+" 3.4a version, so I shall try "memtest"3.4a.

( I saw a reference to "memtest" not being the same application as
"memtest+". What is the significance of this difference?
I also saw mention of 'spurious errors when using memtest 3.5 on
4GB or more'.

So first I shall try the memtest 3.4a and see how it gets on...


The memtest series has two authors working on it. One author stopped
for a period of time, when the other stopped working on it. As far
as I know, both are working on it now.

The memtest86+ site has no "version 3", because they know
the other site is using that number.

http://www.memtest.org/ Versions: 2.00, 2.01, 2.10, 2.11, 4.00

The other site is the Brady one. You can find 3.4a here.

http://www.memtest86.com/download.html

If Prime95 fails, that is confirmation of a problem.

If Prime95 had passed, but memtest had failed, then we might
suspect a problem with the memtest program. In your case,
both failing confirms a problem exists.

*******

Command Rate 1T and 2T, affects setup time for address/command
setup to clock edge. The setting "2T" puts the address/command
value on the bus for two cycles, giving it time to settle. The
value is apparently strobed into the memory chips, on the end
of the second cycle. This gives plenty of setup time to clock
edge, and is preferred for extremely high clock speeds.

Command Rate 1T is a more aggressive setting. It would work
typically with two memory DIMMs, one per channel, and at
more reasonable clock speeds. As the speed of the DIMM goes up,
at some point it makes sense to switch to 2T. The BIOS should be
doing this automatically.

You can use CPUZ to verify the Command Rate is either 1T or 2T.

On some chipsets, only Command Rate 2T is supported (the relaxed one).
And the hardware may not have any mechanism to report this. So we
rely on supposition, that this is the case.

Also note - when you see a setting in the BIOS, that does not
prove the capability exists. A BIOS may show four settings
for a hardware parameter. The actual hardware may only support
two of those values. It is quite easy for the BIOS setup screen
to tell a lie. That is why programs like CPUZ are important,
to verify the truth. I have one motherboard here, where the
timings you set in the BIOS, aren't the ones listed in CPUZ.
It means I have to be careful when making changes.

Paul

On Wed, 10 Mar 2010 11:14:28 +0000, I wrote:

So first I shall try the memtest 3.4a and see how it gets on..

I have now run memtest86 3.4 for just over 3 hours, at which
point it locked solid - the 'wall time' not moving and not
responding to any control keys.

If you point your browser at the URLs below you will see photo's
taken after 4 errors had been found, then 8 (always on test #5),
then after it locked up. At that point, the error count for
test#6 was rapidly cycling up to 32767 - over and over...

http://www.maurice99.ukfsn.org/memtest3.4-1.JPG
http://www.maurice99.ukfsn.org/memtest3.4-2.JPG
http://www.maurice99.ukfsn.org/memtest3.4-3.JPG

I've similarly informed the supplier (also about prime95), and await
their next move...
--
/\/\aurice
(Replace "nomail.afraid" by "bcs" to reply by email)

Maurice Batey wrote:
On Wed, 10 Mar 2010 11:14:28 +0000, I wrote:

So first I shall try the memtest 3.4a and see how it gets on..

I have now run memtest86 3.4 for just over 3 hours, at which
point it locked solid - the 'wall time' not moving and not
responding to any control keys.

If you point your browser at the URLs below you will see photo's
taken after 4 errors had been found, then 8 (always on test #5),
then after it locked up. At that point, the error count for
test#6 was rapidly cycling up to 32767 - over and over...

http://www.maurice99.ukfsn.org/memtest3.4-1.JPG
http://www.maurice99.ukfsn.org/memtest3.4-2.JPG
http://www.maurice99.ukfsn.org/memtest3.4-3.JPG

I've similarly informed the supplier (also about prime95), and await
their next move...

I like the "errors per memory slot" feature of that version.
I haven't used the Brady version in years.

Did you notice that only one "slot" has errors and the
other one doesn't ? Interesting.

The program actually moves its code segment out of the way,
in order to carry out a test on the memory area the program
is resting in. That is why there is a short test period and
a long test period, per each test. The "freeze" may have happened
when the code got moved, and maybe the code was corrupted.

Paul

On Wed, 10 Mar 2010 11:49:03 -0500, Paul wrote:

Did you notice that only one "slot" has errors and the
other one doesn't ? Interesting.

Yes - but as the test hadn't finished (how long might it take?!)
who knows what it might have found on the 2nd slot...
--
/\/\aurice
(Replace "nomail.afraid" by "bcs" to reply by email)

On Wed, 10 Mar 2010 06:31:28 -0500, Paul wrote:

You can use CPUZ to verify the Command Rate is either 1T or 2T.

It says "2T", but I can't find anything in the BIOS manual about
it, or in the BIOS Setup menus.

It's also showing 5-5-5-15 for DRAM Timing Control.

--
/\/\aurice
(Replace "nomail.afraid" by "bcs" to reply by email)

Maurice Batey wrote:
On Wed, 10 Mar 2010 11:49:03 -0500, Paul wrote:

Did you notice that only one "slot" has errors and the
other one doesn't ? Interesting.

Yes - but as the test hadn't finished (how long might it take?!)
who knows what it might have found on the 2nd slot...

Memtest runs forever. It'll run until you stop it.

I was noting that only one of your DIMMs is reporting errors.
At least, that is what the Brady program is claiming in its
test output. Figuring out which DIMM is which, isn't trivial
in software, so there is still some room for doubt, when the
program shows you such a result. It requires the software
author to know exactly how addresses map to memory sticks.

It is one of the reasons I recommend to people, that they
install one DIMM at a time, if they are interested in
identifying for sure that a DIMM is bad. You don't need
to worry about that, as you will likely be using your
warranty services. And they'll have to figure it out.

One other thing. Any time you have a computer, where you
send it out for service, *always* back up any personal
data on a disk you can keep at home. There have been
numerous horror stories of sending a computer to "the geeks",
and having the hard drive erased. Even though the hard
drive had nothing to do with the problem, some places
will just "level and reload" no matter what. If you have
files on there, or even wish to image the entire disk, do it
before sending the machine back. If there is nothing of
value on the new disk, then don't worry about it.

Paul

On Wed, 10 Mar 2010 22:41:12 -0500, Paul wrote:

Any time you have a computer, where you
send it out for service, *always* back up any personal
data on a disk you can keep at home.

Yes, indeed! I have an easy way of keeping my desktop in sync with
laptop, for use when away from base.
Also, I keep a regular external-drive backup of all files on the
desktop, so I would be able to re-create the system.

So while the PC is 'away', I shall just soldier on with the laptop,
and re-sync when it returns.

(They offered to send replacement RAM, but I accepted the alternative
of (free) collection and return instead.)
--
/\/\aurice
(Replace "nomail.afraid" by "bcs" to reply by email)

Maurice Batey wrote:
On Wed, 10 Mar 2010 22:41:12 -0500, Paul wrote:

Any time you have a computer, where you
send it out for service, *always* back up any personal
data on a disk you can keep at home.

Yes, indeed! I have an easy way of keeping my desktop in sync with
laptop, for use when away from base.
Also, I keep a regular external-drive backup of all files on the
desktop, so I would be able to re-create the system.

So while the PC is 'away', I shall just soldier on with the laptop,
and re-sync when it returns.

(They offered to send replacement RAM, but I accepted the alternative
of (free) collection and return instead.)

At least you have a second machine. That takes some of the pain out
of waiting for it to be returned.

Paul

On Thu, 11 Mar 2010 17:07:06 +0000, Maurice Batey wrote:

(They offered to send replacement RAM, but I accepted the alternative
of (free) collection and return instead.)

They now tell me it was a faulty CPU (Intel Q8300) - not memory - so awaiting
new stock of that so can fix and return the PC...

(Under warranty)

--
/\/\aurice
(Replace "nomail.afraid" by "bcs" to reply by email)