memory checker

Ferris

hi, i seem to have taken over this thread

bbb - when i say "completely different location", i mean the memory addresses M$ gives as failing are completely different to those produced by memtest.

so, for example, memtest finds errors in addresses that look like:
05faa74c
17daa74c
memtest also indicates that these are in the region 386MB.

while M$ finds errors that look like:
0192b53c
054ab53c

i notice that both sets of errors have consistent last 4 characters. does this indicate any kind of proximity... i.e. are they "in the same location" as you put it?

Joe_bloggs:
all my temperatures seem normal under windows (e.g. never really going higher than 47deg under very heavy load (Prime95 CPU torture test). my voltages are fairly stable. my 12V sits at around 11.25V which i was a little concerned about but have been assured is normal.

i can run half life 2 full screen with all graphics settings at maximum without problems, but Prime95 (on the maximum ram test) usually crashes after less than one minute with a rounding error.

and i still get completely random BSOD.

i think i will be trying the swapping memory modules around option to see if it helps. one question: what do you do with a memory module when it is not in the PC? where do you store it that won't damage it, if any static at all could wreck it?

Joe_Bloggs

@Ferris the 11.25V is not good from my viewpoint. The 12V line is the most stressed in the system It may be wobbling around and you just measure the average value. You might consider reducing the load by removing a CD /DVD for example. Both power and IDE cable have to be removed.
So many problems are overcome by getting a quality power supply.
J_B.

Ferris

Joe_Bloggs wrote:

@Ferris the 11.25V is not good from my viewpoint. The 12V line is the most stressed in the system It may be wobbling around and you just measure the average value. You might consider reducing the load by removing a CD /DVD for example. Both power and IDE cable have to be removed.
So many problems are overcome by getting a quality power supply.
J_B.

interesting... i was told that +/- 10% is ok for the 12V line (+/-5% for the others). in bios it comes up as 11.85V, and in windows i use speedfan which measures the voltage over time, and it it never goes lower than 11.13V and often sits at 11.31V, both of which are within the 10% allowance, but i am really very green when it comes to motherboard issues so you may well be right; yet another thing to look into...

thanks!

bbb_uk

Joe_Bloggs wrote:

.....I don't think it's worth backing things up in a system that fails a memory test. Better to disconnect the hard drive and leave things the way they are . Installing an operating system in such circumstances is as prone to memory errors as much as any other process. Only when you pass a memory test should you consider adding a memory hogging operating system....

I only suggested backing up the system and a fresh install of windows because there is no guarantee that the problem is memory (although I have to admit is is pointing in that direction). I only suggested backing-up and a fresh install of windows if obtaining replacement memory, etc wasn't possible due to financial constraints. A fresh install of WinXP would prove the BSoD's are not driver issues (providing all non-essential hardware was removed from motherboard - you should only need a graphics card). I even admitted that a backup and fresh install was a time-consuming process.

Ferris, what are the BSoD's error messages and you mention about swapping memory modules around if you have for example two memory modules then I would just take one out and test that and then swap over. Also, I have read somewhere that it is possible that memory errors can occur (very rare i would have thought) on a particular memory bank so say you used memory bank 0 for the tests, then do the same again with them in memory bank 1.

I realise memory tests can be time-consuming and I agree with Joe_Bloggs that cooling is important so if its not possible to do the tests from cold then at least have a fan on them or something.

Ferris, does M$ memory tester not report in MB where the problems are as those values you mentioned are hex values and using the windows 'calc' in scientific mode reports that the value from M$ (0192b53c) is 26,391,868 in decimal (26mB) region. Are the values the your providing the actual location of a memory address or the value that is written to memory at the time of the error? Using Memtest, the actual errors show up half way down the page when they occur but the value that is written to the memory at the time is at the top. Does this make sense?

blinky

Memtest is very sensitive to memory errors. The fact that you cannot run prime95 suggests the system isn't stable.

What's the full spec of the machine?

bbb_uk

blinky wrote:

Memtest is very sensitive to memory errors. The fact that you cannot run prime95 suggests the system isn't stable.

What's the full spec of the machine?

It is also not always reliable as stated in their literature hence the reason why I suggested doing several tests (ideally whilst motherboard is cold or at least cooled down via a house fan or something).

Ferris

bbb_uk wrote:

Ferris, what are the BSoD's error messages and you mention about swapping memory modules around if you have for example two memory modules then I would just take one out and test that and then swap over. Also, I have read somewhere that it is possible that memory errors can occur (very rare i would have thought) on a particular memory bank so say you used memory bank 0 for the tests, then do the same again with them in memory bank 1.

I realise memory tests can be time-consuming and I agree with Joe_Bloggs that cooling is important so if its not possible to do the tests from cold then at least have a fan on them or something.

Ferris, does M$ memory tester not report in MB where the problems are as those values you mentioned are hex values and using the windows 'calc' in scientific mode reports that the value from M$ (0192b53c) is 26,391,868 in decimal (26mB) region. Are the values the your providing the actual location of a memory address or the value that is written to memory at the time of the error? Using Memtest, the actual errors show up half way down the page when they occur but the value that is written to the memory at the time is at the top. Does this make sense?

the only memory swap thing i have done to date is to try swapping the DIMMS around in their banks. when i get the time, i will try to run an overnight test with one DIMM at a time. the problem is, the PC is in my bedroom, and the five fans inside the case make a heck of a racket!

the M$ and the memtest results are the "failing addresses" (in hex) rather than the failing patterns. what i am wondering is whether the last four characters denote a particular area of memory. i have taken screenshots (digicam) of the output of both programs if you would like to see them for more clarification.

full spec:
athlon 64 3200+
asus k8v MB
1GB 400mhz ddr ram PC3200 (2x512)
160GB maxtor s-ata hard drive
asus nvidia geforce 5900 128mb graphics card
windows xp pro (SP1a)
jlms dvd-rom
liteon dvd-rw
d-link wireless card

bbb_uk

the only memory swap thing i have done to date is to try swapping the DIMMS around in their banks

If the problem does seem to be around 381MB region then swapping the memory around would see the errors occur higher up in the memory (ie after swapping memory modules around the so-called defective memory would only be accessed after the other is used (ie higher than 512MB).

when i get the time, i will try to run an overnight test with one DIMM at a time.

I would definitely recommend this as what are the chances of both memory modules failing at the exact same memory location (380MB).

Remember, that it may not be the memory but the motherboard, or (if you're really unluck) both and the reason for testing one memory module at a time and to then test this one memory module in both the memory banks one at a time, is to determine for sure its the memory and not motherboard.

what i am wondering is whether the last four characters denote a particular area of memory.

I'm not sure about that to be honest. Maybe someone else could answer this question?

Ferris

bbb_uk wrote:

If the problem does seem to be around 381MB region then swapping the memory around would see the errors occur higher up in the memory (ie after swapping memory modules around the so-called defective memory would only be accessed after the other is used (ie higher than 512MB).
I would definitely recommend this as what are the chances of both memory modules failing at the exact same memory location (380MB).

that's what i would have thought, but it didn't seem to happen. even after i swapped them around the failing addresses were in the same place.

sigh. i guess that suggests a problem with the motherboard (memory controller or something), right?

bbb_uk

Ferris wrote:

that's what i would have thought, but it didn't seem to happen. even after i swapped them around the failing addresses were in the same place.

sigh. i guess that suggests a problem with the motherboard (memory controller or something), right?

I'd test one module on its own in memory bank 0 and then again in memory bank 1, and then the other memory module in bank 0 and then again in bank 1. Remember try and do two tests per module - to be sure.

I realise that would take you about 2 years ( :rotfl: ) but you need to know beyond a doubt as another motherboard would cost more than memory.

If the motherboard is still under warranty then you would still need to do these tests beforehand anyhow as they would try and say its not the m'board but it could be driver/hardware issue, or even memory.