Memtest - Proper use of?

[Louise]

Kingston Value RAM is indeed rebandaged. I bought it once- 2 X 256 MB. One
stick was Samsung and one was Nanya. I couldn't even boot up the system. I
was fairly new to building my own computers at the time and ended up calling
Soyo (motherboard). I actually got to talk to a live person who, when told
what memory I had purchased, told me to "get that garbage out of my
motherboad". Replaced with Crucial, and no problems.

Fitz

I just want to thank everyone for responding so thoroughly that I began
to look at things I'd never paid much attention to before.

I found the temp reading in the Bios (rather than using Everest), and
suddenly saw the cpu go up to 75c.

Needless to say I got very nervous. Opened up the machine and
discovered that the cpu fan wasn't turning at all. Raced to turn off
the machine and then started following the wires.

From what I can tell, I accidentally pulled the power plug from the cpu
to the motherboard when I was fighting to get at one of the ram slots.
It seems my machine has been "funny" because the cpu hasn't been cooled
in over 24 hours.

All better now - when I recover from my "trauma" (and get some sleep),
I'll run memtest again and see what happens now that I have a cooling
fan on the cpu

Louise

 

[w_tom]

Method for getting a reliable answer to "is CPU too hot?" is
to read number from Intel(?)/AMD(?) spec sheet. At 70 degrees
F, the CPU must be about 18 degrees C below its maximum
temperature (so that it will not be too hot in a 100 degree F
room). Add a safety margin of 5 degrees. Do those numbers
and you should be able to tell us your CPU is not too hot.

 

[Al Smith]

From what I can tell, I accidentally pulled the power plug from the cpu

to the motherboard when I was fighting to get at one of the ram slots.
It seems my machine has been "funny" because the cpu hasn't been cooled
in over 24 hours.

That'll do it, all right.

 

[Guest]

I found the temp reading in the Bios (rather than using Everest), and

suddenly saw the cpu go up to 75c.

Opened up the machine and discovered that the cpu fan wasn't
turning at all.

All better now - when I recover from my "trauma" (and get some sleep),
I'll run memtest again and see what happens now that I have a cooling

fan on the cpu

It's good you have a Pentium 4 with inherent overheat protection,
rather than an Athlon 32, which lacks this.

The machine is often on for 48 hours at a time - and no, it
never hangs, athough sometimes it gets a bit "sluggish" and
I reboot and it speeds up again.

If Memtest86 finds an error, there definitely is an error that
shouldn't be ignored because will eventually make Windows crash, and
you shouldn't accept the memory unless the problem lies elsewhere, such
as with the power supply or incorrect BIOS settings. Have you checked
those settings? Some can be set too fast even when the BIOS is
configured for automatic memory settings (SPD chip on the memory
module).

 

[spodosaurus]

It's good you have a Pentium 4 with inherent overheat protection,
rather than an Athlon 32, which lacks this.

Wrong. Athlons had this long before motherboard manufacturers bothered
to implement it.




--
spammage trappage: replace fishies_ with yahoo

I'm going to die rather sooner than I'd like. I tried to protect my
neighbours from crime, and became the victim of it. To jump to the end
of the story, as a result of this I need a bone marrow transplant. Many
people around the world are waiting for a marrow transplant, too. Please
volunteer to be a marrow donor:
http://www.abmdr.org.au/
http://www.marrow.org/

 

[Mac Cool]

spodosaurus:

Wrong. Athlons had this long before motherboard manufacturers bothered
to implement it.

I had a 1.4Ghz Athlon and Abit motherboard with built in overheat
protection, that was about 5 years ago.

 

[spodosaurus]

spodosaurus:




I had a 1.4Ghz Athlon and Abit motherboard with built in overheat
protection, that was about 5 years ago.

It probably wasn't using the on-die thermal sensor of the Athlon chip if
it's a Thunderbird core. The diode was put in place in the Palomino core
and later models. So, if the one you're talking about is a Thunderbird
core it's using an external motherboard temperature sensor. Also, if
it's a thunderbird, I wouldn't test that thermal protection by running
it without a heatsink.

Cheers,

Ari

--
spammage trappage: replace fishies_ with yahoo

I'm going to die rather sooner than I'd like. I tried to protect my
neighbours from crime, and became the victim of it. To jump to the end
of the story, as a result of this I need a bone marrow transplant. Many
people around the world are waiting for a marrow transplant, too. Please
volunteer to be a marrow donor:
http://www.abmdr.org.au/
http://www.marrow.org/

 

[Louise]

It's good you have a Pentium 4 with inherent overheat protection,
rather than an Athlon 32, which lacks this.


If Memtest86 finds an error, there definitely is an error that
shouldn't be ignored because will eventually make Windows crash, and
you shouldn't accept the memory unless the problem lies elsewhere, such
as with the power supply or incorrect BIOS settings. Have you checked
those settings? Some can be set too fast even when the BIOS is
configured for automatic memory settings (SPD chip on the memory
module).

My bios is configured, as far as I could make out, for automatic
settings. But....how do I know what to change it to?

Thanks.

Louise

 

[Guest]

My bios is configured, as far as I could make out, for automatic
settings. But....how do I know what to change it to?

Try the most conservative settings -- the slowest memory bus speed,
highest number of clock cycles (i.e., choose 4T over 2T), and "normal"
over "fast" or "turbo." You may have to switch the memory timing to
manual mode to make some of these changes.

 

[David Maynard]

Memtst (like all memory testers) has a problem in that some
memory locations are not in memory.

In a word: no.

For example, the NIC can
reserve a block of memory.

No, how much it needs is specified through it's PnP/ACPI configuration but...

If it's an ancient legacy device it's memory is in the 640k to 1 meg area,
which we now call a 'hole', that was reserved for I/O devices (and BIOS) in
the original PC. If it's a modern device it gets a high memory assignment
above RAM up to the 32 bit 4 gig boundary. If you have more than 4 gig of
RAM then there's another 'hole'.

Then when Memtst thinks it is
testing motherboard memory, instead, it is screwing with the
NIC. Video controller creates same complication.

Memory tester must understand and avoid these problems.
Therefore Memtst is constantly being upgraded. Memtst must be
constantly upgraded as these new 'landmines' are discovered.

And you really think the O.S. and apps would enjoy having memory addresses
mysteriously become 'land mines' because an I/O card decided, all on its
own, to insert itself in the middle of system RAM? Or that anyone would be
foolish enough to design a system which allowed it?

It don't work that way. I/O devices are not in system RAM.

Too many reasons why this test may do as you report. Too
many others have simply posted 'try this and try that'
speculations. Do you even know if other computer hardware is
good? Does you computer manufacturer provide comprehensive
diagnostics for free? It not, then you must download Memtst
and other diagnostics, one at a time, to verify your other
hardware.

However start by determining YOUR Memtst is the latest
version.

BTW heat is not a reason for failure. Heat is the
diagnostic tool we use to find defective hardware.

Depends on how much, or are you willing to let me test your memory with an
acetylene torch?

 

[Frank]

Thanks -

I guess I'll just test it again and see if it hangs at the same place.
Perhaps there was some kind of blip in the power - that does happen in
this building.

You are just asking for numerous subtle problems if you do not have your
system on a UPS.

 

[w_tom]

David is posting carefully scripted contrarian responses.
If memory blocks are dedicated to other functions, then Memtst
must be modified so as to identify and not interfere with
these other functions. These hardware 'landmines' can cause
Memtest execution problems. Some of these are legency
functions. Some of these 'landmines' are new functions.
Older versions of Memtest can have problems with newer
hardware.

David even creates strawmen only to argue. The earlier post
never mentioned IO addresses. David had to invent the problem
so at to attack it. He has done this in other discussions.
It is his history.

As hardware changes - as new devices sometimes use dedicated
memory - then Memtst must be upgraded to avoid these
complications. Otherwise anyone could write comprehensive
memory diagnostics - as I did even 30 years ago.

 

[w_tom]

This need for a UPS is only true if using an obsolete
Operating System or obsolete filesystem. An unexpected
blackout should never cause problematic system problems. But
that assumes the system is properly constructed. A plug-in
UPS is for the protection of unsaved data. A UPS to solve
Louise's problems is better described as curing a symptom.

 

[David Maynard]

David is posting carefully scripted contrarian responses.
If memory blocks are dedicated to other functions, then Memtst
must be modified so as to identify and not interfere with
these other functions. These hardware 'landmines' can cause
Memtest execution problems. Some of these are legency
functions. Some of these 'landmines' are new functions.
Older versions of Memtest can have problems with newer
hardware.

Wrong. System RAM does not go in the reserved I/O region and I/O doesn't go
into the system RAM.

David even creates strawmen only to argue. The earlier post
never mentioned IO addresses. David had to invent the problem
so at to attack it. He has done this in other discussions.
It is his history.

I suppose you think the device I/O region, that you claim arbitrarily pops
itself into system RAM, has no address.

As hardware changes - as new devices sometimes use dedicated
memory - then Memtst must be upgraded to avoid these
complications.

Wrong. The regions reserved for potential I/O device use are specific and
unambiguous.

Otherwise anyone could write comprehensive
memory diagnostics - as I did even 30 years ago.

Things have changed in 30 years.