You are not logged in.

#1 2009-05-10 15:48:53

skottish
Forum Fellow
From: Here
Registered: 2006-06-16
Posts: 7,942

Question about memtest86+ errors

I'm nine hours into running memtest86+ on 8GB of DDR3-1333, PC3-10600 RAM and it's come up with two errors:

Pass        Where        Good            Bad           Err-Bits
--------------------------------------------------------------------
  3          8840MB   e36e5578   636e5578    80000000
  5          8800MB   8a029d49   0a029d49    80000000

I'm not sure how concerned to be. Is there a margin of error with this program? Is the fact that memtest86+ is mis-reporting the type and clock speed of the RAM, as well the processor type, a factor in the results?

--EDIT--

I'm stopping the test at 9.5 hours. I'll swap the sticks and see if the same one produces any errors. I'd still like the questions above to be answered though. Getting good information is proving to be a little tricky.

Offline

#2 2009-05-12 04:27:12

skottish
Forum Fellow
From: Here
Registered: 2006-06-16
Posts: 7,942

Re: Question about memtest86+ errors

No one? This is a big concern for me and the time frame on returning the memory is short. I guess that the most important thing I mentioned above is the memtest86+ is seeing the RAM differently from what it is. I understand that overclocked RAM can generate lots of errors with the tests. I'm still wondering if underclocked or misrepresented memory could do the same.

Offline

#3 2009-05-12 06:04:06

cerbie
Member
Registered: 2008-03-16
Posts: 124

Re: Question about memtest86+ errors

skottish wrote:

No one? This is a big concern for me and the time frame on returning the memory is short. I guess that the most important thing I mentioned above is the memtest86+ is seeing the RAM differently from what it is. I understand that overclocked RAM can generate lots of errors with the tests. I'm still wondering if underclocked or misrepresented memory could do the same.

What seems odd to me is that it's showing the spot being 8800MB. 8GB should stop below 8200MB.

However, try testing each stick individually, in slot 1, and see if you get errors with any one of them..

Last edited by cerbie (2009-05-12 06:04:56)


"If the data structure can't be explained on a beer coaster, it's too complex." - Felix von Leitner

Offline

#4 2009-05-12 11:09:26

R00KIE
Forum Fellow
From: Between a computer and a chair
Registered: 2008-09-14
Posts: 4,734

Re: Question about memtest86+ errors

I have no idea of why it is still testing so far up .... maybe due to memory reallocation to allow access to system devices (maybe around 1GB).
I would do as cerbie says, test stick by stick and let the motherboard set the timings and don't reduce the voltage on the ram.

As far as I know there is no margin for error with memtest, underclocked or misrepresented ram should not change that. memtest just tries to figure out how much memory you have (it tries to ask the bios for that) and then writes a series of patterns to the ram and tries to read them back and compares them.

Overclocking is the best way to start getting errors and possibly damage the ram, tighter timings should be harmless I believe but more voltage can kill the ram.

Try to test the ram in another system if you can, the ram may be ok but the motherboard could be starting to fail, it should be less likely but can happen.


R00KIE
Tm90aGluZyB0byBzZWUgaGVyZSwgbW92ZSBhbG9uZy4K

Offline

#5 2009-05-12 19:19:26

skottish
Forum Fellow
From: Here
Registered: 2006-06-16
Posts: 7,942

Re: Question about memtest86+ errors

I'm not underclocking the RAM, it's just how it reads. I have to manually set the correct timings in BIOS to get the sticks to register properly.

The reason why I'm asking about margin of error because of what I've read at the memtest86+ forums. The general idea is that the values may not be trustworthy if the program doesn't know the correct system settings, which it doesn't. But, their solution is produce a machine dump with CPU-Z, a Windows only program that doesn't run properly under Wine. So I guess that the only solution is to swap out sticks and hope that memtest86+ is actually producing useful data.

Offline

#6 2009-05-12 19:38:14

MoonSwan
Member
From: Great White North
Registered: 2008-01-23
Posts: 881

Re: Question about memtest86+ errors

Is there no other equivalent program you could try under linux?  I don't know of one but that doesn't mean a lot.

Offline

#7 2009-05-13 09:04:36

toad
Member
From: if only I knew
Registered: 2008-12-22
Posts: 1,775
Website

Re: Question about memtest86+ errors

You prolly done that already, but have you checked your message log? Do you get any mem errors in there?


never trust a toad...
::Grateful ArchDonor::
::Grateful Wikipedia Donor::

Offline

#8 2009-05-13 14:06:37

R00KIE
Forum Fellow
From: Between a computer and a chair
Registered: 2008-09-14
Posts: 4,734

Re: Question about memtest86+ errors

It should work even if it doesn't know much about your system, I believe that showing the timings, speed and type of the memory are more of a cosmetic thing than anything else because that as far as I know memtest will not mess with any memory settings unless you go into the advanced options and change them yourself, I may be wrong though.

You mention you have to set the timings manually, these days you can set a ton of things no one knows for sure what they do except for the people that make the hardware .... maybe one of those cryptic timings is set too tight, also check the recommended voltage for the modules and the voltage you have set, some modules need slightly more voltage to work properly (how much should be stated by the manufacturer).

Memory related problems are always a pain, most of the time intermittent and depend on temperature hmm


R00KIE
Tm90aGluZyB0byBzZWUgaGVyZSwgbW92ZSBhbG9uZy4K

Offline

#9 2009-05-16 06:51:25

skottish
Forum Fellow
From: Here
Registered: 2006-06-16
Posts: 7,942

Re: Question about memtest86+ errors

MoonSwan wrote:

Is there no other equivalent program you could try under linux?

That's a great question that deserves a response. I searched around for a while and found two programs already in the repos: stress and memtester. I don't know how comprehensive that they are compared to memtest86+, but they both can be run from within a regular install.

ROOKIE, it looks like the voltage may have been the problem. I had the timings set fine and I was getting segfaults very frequently. I brought the voltage up a tiny bit and the system seems to have stabilized. I'm grateful for that. I screw my system up perfectly fine without the help of any hardware.

Offline

Board footer

Powered by FluxBB