PDA

View Full Version : Need help from a memory expert


Peter Barton
December 7th 03, 12:45 AM
I recently bought a new PC with Windows XP and 1GB RAM. After 5 days XP
crashed with an unrecoverable registry corruption. I reinstalled, then last
week it fell over again. This time it recovered, and XP did some
self-diagnosis. It suggested that based on the nature of the crash it looked
like a memory issue, probably hardware. It directed me to download a memory
test tool from the Microsoft web site:

http://oca.microsoft.com/en/windiag.asp

Sure enough, their test tool regularly comes up with memory errors. The
computer needs to have been on a while mind you, if booted cold it seems to
take longer to detect issues (when warm I get 1 error at least every 30
minutes of testing).

So I unplugged all the hardware in the box, removed the AGP video card and
threw in a basic PCI VGA card - memory error still detected. I'm down to the
CPU (P4 2.67GHz), motherboard (Gigabyte 8SQ800 Ultra), RAM (2x512MB Kingston
DDR 2700) or the power supply (generic).

The motherboard has some newer memory access modes that I'm not totally
familiar with, including Dual Channel Mode. If I disable that, the Microsoft
tool doesn't detect an error (after 1.5 hours, I'm yet to test this way
overnight).

With Dual Channel Mode enabled, I also tried testing using DocMemory
(www.simmtester.com), no errors reported after 12 hours. This tool seems a
little old though, (C)2000? I've also downloaded Memtest86 and am running
that now - no errors so far after 1 full pass. Are these other two tools good
- why do they fail to detect any errors, while the Microsoft tool reports
problems regularly?

Anyone else experience these issues? Is there something else / some other way
I can test?


Peter Barton,
P.O. Box 364, Bulleen, VIC 3105, Australia
http://www.madmaxmovies.com - Mad Max!

D.Currie
December 7th 03, 12:45 AM
This might seem overly obvious, but if you're getting memory errors, the
logical thing to change would be the memory. It might not necessarily be
defective, just incompatible for some reason.

"Peter Barton" > wrote in message
...
> I recently bought a new PC with Windows XP and 1GB RAM. After 5 days XP
> crashed with an unrecoverable registry corruption. I reinstalled, then
last
> week it fell over again. This time it recovered, and XP did some
> self-diagnosis. It suggested that based on the nature of the crash it
looked
> like a memory issue, probably hardware. It directed me to download a
memory
> test tool from the Microsoft web site:
>
> http://oca.microsoft.com/en/windiag.asp
>
> Sure enough, their test tool regularly comes up with memory errors. The
> computer needs to have been on a while mind you, if booted cold it seems
to
> take longer to detect issues (when warm I get 1 error at least every 30
> minutes of testing).
>
> So I unplugged all the hardware in the box, removed the AGP video card and
> threw in a basic PCI VGA card - memory error still detected. I'm down to
the
> CPU (P4 2.67GHz), motherboard (Gigabyte 8SQ800 Ultra), RAM (2x512MB
Kingston
> DDR 2700) or the power supply (generic).
>
> The motherboard has some newer memory access modes that I'm not totally
> familiar with, including Dual Channel Mode. If I disable that, the
Microsoft
> tool doesn't detect an error (after 1.5 hours, I'm yet to test this way
> overnight).
>
> With Dual Channel Mode enabled, I also tried testing using DocMemory
> (www.simmtester.com), no errors reported after 12 hours. This tool seems a
> little old though, (C)2000? I've also downloaded Memtest86 and am running
> that now - no errors so far after 1 full pass. Are these other two tools
good
> - why do they fail to detect any errors, while the Microsoft tool reports
> problems regularly?
>
> Anyone else experience these issues? Is there something else / some other
way
> I can test?
>
>
> Peter Barton,
> P.O. Box 364, Bulleen, VIC 3105, Australia
> http://www.madmaxmovies.com - Mad Max!

Jim
December 7th 03, 12:45 AM
Frankly, I don't know what to make of the Windows diagnostics. I've used
memtest-86 for a long time, highly reliable and has weeded out many a bad
modules several times.

What I suggest is going into your BIOS and "Load BIOS defaults" for the
Advanced Chipset Features page. Overly aggressive memory timings are a
common source of RAM errors and just plain weird behavior in Windows. Then
rerun the Windows tests and see if the errors disappear.

HTH

Jim

"Peter Barton" > wrote in message
...
> I recently bought a new PC with Windows XP and 1GB RAM. After 5 days XP
> crashed with an unrecoverable registry corruption. I reinstalled, then
last
> week it fell over again. This time it recovered, and XP did some
> self-diagnosis. It suggested that based on the nature of the crash it
looked
> like a memory issue, probably hardware. It directed me to download a
memory
> test tool from the Microsoft web site:
>
> http://oca.microsoft.com/en/windiag.asp
>
> Sure enough, their test tool regularly comes up with memory errors. The
> computer needs to have been on a while mind you, if booted cold it seems
to
> take longer to detect issues (when warm I get 1 error at least every 30
> minutes of testing).
>
> So I unplugged all the hardware in the box, removed the AGP video card and
> threw in a basic PCI VGA card - memory error still detected. I'm down to
the
> CPU (P4 2.67GHz), motherboard (Gigabyte 8SQ800 Ultra), RAM (2x512MB
Kingston
> DDR 2700) or the power supply (generic).
>
> The motherboard has some newer memory access modes that I'm not totally
> familiar with, including Dual Channel Mode. If I disable that, the
Microsoft
> tool doesn't detect an error (after 1.5 hours, I'm yet to test this way
> overnight).
>
> With Dual Channel Mode enabled, I also tried testing using DocMemory
> (www.simmtester.com), no errors reported after 12 hours. This tool seems a
> little old though, (C)2000? I've also downloaded Memtest86 and am running
> that now - no errors so far after 1 full pass. Are these other two tools
good
> - why do they fail to detect any errors, while the Microsoft tool reports
> problems regularly?
>
> Anyone else experience these issues? Is there something else / some other
way
> I can test?
>
>
> Peter Barton,
> P.O. Box 364, Bulleen, VIC 3105, Australia
> http://www.madmaxmovies.com - Mad Max!

Gran Salseron
December 7th 03, 12:45 AM
All three memory testers you mentioned are reliable except that you are
correct in that docmem is old. You should check to see if the RAM you have
on the motherboard supports dual channel mode.

"Peter Barton" > wrote in message
...
> I recently bought a new PC with Windows XP and 1GB RAM. After 5 days XP
> crashed with an unrecoverable registry corruption. I reinstalled, then
last
> week it fell over again. This time it recovered, and XP did some
> self-diagnosis. It suggested that based on the nature of the crash it
looked
> like a memory issue, probably hardware. It directed me to download a
memory
> test tool from the Microsoft web site:
>
> http://oca.microsoft.com/en/windiag.asp
>
> Sure enough, their test tool regularly comes up with memory errors. The
> computer needs to have been on a while mind you, if booted cold it seems
to
> take longer to detect issues (when warm I get 1 error at least every 30
> minutes of testing).
>
> So I unplugged all the hardware in the box, removed the AGP video card and
> threw in a basic PCI VGA card - memory error still detected. I'm down to
the
> CPU (P4 2.67GHz), motherboard (Gigabyte 8SQ800 Ultra), RAM (2x512MB
Kingston
> DDR 2700) or the power supply (generic).
>
> The motherboard has some newer memory access modes that I'm not totally
> familiar with, including Dual Channel Mode. If I disable that, the
Microsoft
> tool doesn't detect an error (after 1.5 hours, I'm yet to test this way
> overnight).
>
> With Dual Channel Mode enabled, I also tried testing using DocMemory
> (www.simmtester.com), no errors reported after 12 hours. This tool seems a
> little old though, (C)2000? I've also downloaded Memtest86 and am running
> that now - no errors so far after 1 full pass. Are these other two tools
good
> - why do they fail to detect any errors, while the Microsoft tool reports
> problems regularly?
>
> Anyone else experience these issues? Is there something else / some other
way
> I can test?
>
>
> Peter Barton,
> P.O. Box 364, Bulleen, VIC 3105, Australia
> http://www.madmaxmovies.com - Mad Max!


---
Outgoing mail is certified Virus Free.
Checked by AVG anti-virus system (http://www.grisoft.com).
Version: 6.0.484 / Virus Database: 282 - Release Date: 05/27/2003

Peter Barton
December 7th 03, 12:46 AM
I should have mentioned in the original post - a friend bought a system from
the same place a week later with the same motherboard, he bought 512MB of DDR
3200 RAM (I forget which brand), I've got 1024MB of DDR 2700 Kingston RAM. He
also bought a different processor. The Microsoft memory tester is also
showing faults in his system.

In some additional testing today, I moved the RAM from the first bank in the
motherboard to the second bank - and the Microsoft memory tester began
failing around 10 times more frequently. Memtest 86 now also shows failures,
it didn't pick up anything after a few passes in the first bank. DocMemory
didn't show any failures after 4 hours.

Looks like the motherboard is the issue, and I have now lodged a support
question with Gigabyte, I'll post if I get a response.

Thanks to everyone for the help.


"D.Currie" > wrote:

>This might seem overly obvious, but if you're getting memory errors, the
>logical thing to change would be the memory. It might not necessarily be
>defective, just incompatible for some reason.

Peter Barton,
P.O. Box 364, Bulleen, VIC 3105, Australia
http://www.madmaxmovies.com - Mad Max!

Peter Barton
December 7th 03, 12:46 AM
Thanks for the tip, I hadn't thought to check that. The Kingston web site has
the Gigabyte motherboard that I have listed, and showed a selection of RAM
which I presume are all considered compatible - the 512MB DIMMs that I have
were on that list. So I guess it is expected to work.

"Gran Salseron" > wrote:

>All three memory testers you mentioned are reliable except that you are
>correct in that docmem is old. You should check to see if the RAM you have
>on the motherboard supports dual channel mode.

Peter Barton,
P.O. Box 364, Bulleen, VIC 3105, Australia
http://www.madmaxmovies.com - Mad Max!

Peter Barton
December 7th 03, 12:46 AM
I tried this early on, and it didn't seem to alter the behaviour. There are
various RAM settings in there (on a 'hidden' page, CTRL-F1 in the BIOS), at
the moment I've just been leaving the settings to AUTO - the BIOS is
supposedly picking what is best for the RAM I have installed. The only
feature I've found that makes a difference is that if I set Dual Channel Mode
to DISABLED, the errors go away.

"Jim" > wrote:

>Frankly, I don't know what to make of the Windows diagnostics. I've used
>memtest-86 for a long time, highly reliable and has weeded out many a bad
>modules several times.
>
>What I suggest is going into your BIOS and "Load BIOS defaults" for the
>Advanced Chipset Features page. Overly aggressive memory timings are a
>common source of RAM errors and just plain weird behavior in Windows. Then
>rerun the Windows tests and see if the errors disappear.
>
>HTH
>
>Jim
>

Peter Barton,
P.O. Box 364, Bulleen, VIC 3105, Australia
http://www.madmaxmovies.com - Mad Max!

Peter Barton
December 7th 03, 01:35 AM
In case anyone is still following this, disabled Dual Channel Mode did make
the situation better - but I still got errors when running the test
overnight.

I've now had other users with 8SQ800 boards run the same test (some who had
also been having problems with XP crashing), and all boards the test has been
run with so far have reported an error - despite different memory
configurations/brands, different speed processors, etc. I'm now concluding
that it's a fault in the motherboard itself, and will be swapping it over for
another brand to see if I can remove the problem that way.


Peter Barton > wrote:

>I tried this early on, and it didn't seem to alter the behaviour. There are
>various RAM settings in there (on a 'hidden' page, CTRL-F1 in the BIOS), at
>the moment I've just been leaving the settings to AUTO - the BIOS is
>supposedly picking what is best for the RAM I have installed. The only
>feature I've found that makes a difference is that if I set Dual Channel Mode
>to DISABLED, the errors go away.
>
>"Jim" > wrote:
>
>>Frankly, I don't know what to make of the Windows diagnostics. I've used
>>memtest-86 for a long time, highly reliable and has weeded out many a bad
>>modules several times.
>>
>>What I suggest is going into your BIOS and "Load BIOS defaults" for the
>>Advanced Chipset Features page. Overly aggressive memory timings are a
>>common source of RAM errors and just plain weird behavior in Windows. Then
>>rerun the Windows tests and see if the errors disappear.
>>
>>HTH
>>
>>Jim
>>
>
>Peter Barton,
>P.O. Box 364, Bulleen, VIC 3105, Australia
>http://www.madmaxmovies.com - Mad Max!

Peter Barton,
P.O. Box 364, Bulleen, VIC 3105, Australia
http://www.madmaxmovies.com - Mad Max!

Google