Who said chipsets don't wear out?!

[tek_heretik]

*****READ THE WHOLE THREAD FOR FOR THE HAPPY ENDING AND A POSSIBLE SOLUTION IF YOU ARE HAVING THE SAME PROBLEM*****

They do! I was live testing Kubuntu 12.10, for schizz n giggles, I ran Memtest86+, much to my surprise and shock, the dreaded red error lines, AHHHHHHHHH! So this prompted TWENTY FOUR HOURS of process of elimination troubleshooting and testing, tested the sticks one by one, they ran fine 'full blast' individually and even paired in 2 stick/dual channel configuration, but put all 4 in, forget it, had to scale back to the motherboard's nominal RAM speed (sucks but at least I am still using all 4 sticks). Even downgraded the BIOS one version (the latest 4 sticks were installed using an older version), made no difference, so went back to the latest BIOS. A new build is overdue but $ is an ugly reality.

This is just past the 3 years warranty expiration, figures, they sure don't make things to last these days. =(

[woodsmoke]

yeppers,
I set of sticks that I purchases a mere four years ago gave up this summer! lol

woodsmoke

[tek_heretik]

yeppers,
I set of sticks that I purchases a mere four years ago gave up this summer! lol

woodsmoke

The RAM is fine, Kingston Hyper X is good stuff, love it, it's the sucky Intel P45 northbridge that's the problem, but yeah, I've had RAM go bad on me too, built a budget system for some friends years ago, they wanted to go cheap and I warned them, of course WinDOHS! started acting up real bad, no-name RAM from the industrial slums of China was not a wise purchase decision, lmao.

[woodsmoke]

lol
i know it is not funny but I had not heard the term "industrial slums" before, but it is seemingly highly descriptive.

And yeah....love my Kingstons!! lol

woodsmoke

[tek_heretik]

Pah! You won't believe this, the original test that started all this turned out to be corrupt (as in corrupt data in the program itself, or should I say bad burn?) BUT, removed 2 of the 4 sticks, turns out less is more! 2 sticks running in dual channel mode, full out timings and speed pass with flying colours using Memtest86+ v4.20 (SuSE 12.1 DVD), v4.00 on an old floppy and Memtest86 4.0a (not 86+, the original) burned to a DVD from the CD creating iso. I was still getting the odd error after leaving the test run all night (4 sticks, nominal speed), hate those dang red bars blazing across the screen, grrrrrrrrrrr. Would rather run at full speed with 4GB then iffy and slow with 8GB, I don't think I have ever loaded the system so much to use more than 2GB in a session anyway.

Edit: My suspicions were spot on, with all four slots populated and running at full bore AND nominal speeds, it's too much for the chipset and/or CPU (extra work for the memory controller), Memtest86 4.0a kept reporting the same CPU core when the errors occurred, same memory address at the same amount, even when the sticks were switched around, there is no way, with millions of memory addresses, two sticks would have the exact same error, impossible (tested each stick individually at full speed, they are fine), that only leaves my conclusion, the CPU and chipset flake out when under too much load. This also proves, never throw the 'baby out with the bath water' until you have tried everything.

[tek_heretik]

And memtester, from within Kubuntu...

memtester version 4.2.2 (64-bit)
Copyright (C) 2010 Charles Cazabon.
Licensed under the GNU General Public License version 2 (only).

pagesize is 4096
pagesizemask is 0xfffffffffffff000
want 3600MB (3774873600 bytes)
got  3600MB (3774873600 bytes), trying mlock ...locked.
Loop 1/1:
  Stuck Address       : ok         
  Random Value        : ok
  Compare XOR         : ok
  Compare SUB         : ok
  Compare MUL         : ok
  Compare DIV         : ok
  Compare OR          : ok
  Compare AND         : ok
  Sequential Increment: ok
  Solid Bits          : ok         
  Block Sequential    : ok         
  Checkerboard        : ok         
  Bit Spread          : ok         
  Bit Flip            : ok         
  Walking Ones        : ok         
  Walking Zeroes      : ok         
  8-bit Writes        : ok
  16-bit Writes       : ok

Done.

[tek_heretik]

Some photographic evidence...

2 sticks, full speed (SPD says 400Mhz but they are designed to run at 533MHz (1066) 5-5-5-15), take note of the pass count... http://www.flickr.com/photos/7978633...n/photostream/

2 'orphaned' sticks, should put 'em up for sale I guess =(... http://www.flickr.com/photos/7978633...n/photostream/

[SamuraiGhost]

tek_heretik, there is probably nothing wrong with your hardware! There is a bug in memtest86+ on *buntu 12.10 that causes it to incorrectly show errors. Try an older install media, or the ultimate boot cd instead. Here is the bug report:
https://bugs.launchpad.net/ubuntu/+s...+/+bug/1071209

I created this account here just to let you know about this since I just wasted about 3-4 hours trying to track this down. -_-

[tek_heretik]

tek_heretik, there is probably nothing wrong with your hardware! There is a bug in memtest86+ on *buntu 12.10 that causes it to incorrectly show errors. Try an older install media, or the ultimate boot cd instead. Here is the bug report:
https://bugs.launchpad.net/ubuntu/+s...+/+bug/1071209

I created this account here just to let you know about this since I just wasted about 3-4 hours trying to track this down. -_-

Huh! Thank you! That would mean the Memest86+ on the Kubuntu 12.10 disk is buggy too? But I do get the odd different errors in different test #'s when all four sticks are in, not many but they are still there, so it is still my CPU/chipset
(the MB/s RAM speed increases as you put more sticks in, single and 2 sticks are fine->@ FULL speed!, 4 sticks cause occasional other test # failures <other than test#7>, even at SPD settings). As you can see, I stated earlier I used many different tests and versions, so when the original Memtest86 4.0a is showing the same error at the same address, that raises red flags for me, I have to trust my hardware 100%. Memtest86 4.0a is buggy too?! Will put all 4 sticks back in and try memtester again from within Kubuntu, memtester passed 2 sticks at full speed so I know it's working properly for sure. I will post the result.

[tek_heretik]

4 sticks @ full speed memtester results...

memtester version 4.2.2 (64-bit)
Copyright (C) 2010 Charles Cazabon.
Licensed under the GNU General Public License version 2 (only).

pagesize is 4096
pagesizemask is 0xfffffffffffff000
want 7550MB (7916748800 bytes)
got  7550MB (7916748800 bytes), trying mlock ...locked.
Loop 1/1:
  Stuck Address       : ok         
  Random Value        : ok
  Compare XOR         : ok
  Compare SUB         : ok
  Compare MUL         : ok
  Compare DIV         : ok
  Compare OR          : ok
  Compare AND         : ok
  Sequential Increment: ok
  Solid Bits          : ok         
  Block Sequential    : ok         
  Checkerboard        : ok         
  Bit Spread          : ok         
  Bit Flip            : ok         
  Walking Ones        : ok         
  Walking Zeroes      : testing 112FAILURE: 0x00008000 != 0x01008000 at offset 0x26aaf280.
  8-bit Writes        : ok
  16-bit Writes       : ok

Done.

Testing 4 sticks @ SPD values except voltage, 1.9v instead of 1.8 as suggested by manufacturer (set 'statically', aka manual, seems to work better than 'auto')...

Huh! I am shocked, the results, 4 sticks, nominal speed using memtester from within Kubuntu...

memtester version 4.2.2 (64-bit)
Copyright (C) 2010 Charles Cazabon.
Licensed under the GNU General Public License version 2 (only).

pagesize is 4096
pagesizemask is 0xfffffffffffff000
want 7550MB (7916748800 bytes)
got  7550MB (7916748800 bytes), trying mlock ...locked.
Loop 1/1:
  Stuck Address       : ok         
  Random Value        : ok
  Compare XOR         : ok
  Compare SUB         : ok
  Compare MUL         : ok
  Compare DIV         : ok
  Compare OR          : ok
  Compare AND         : ok
  Sequential Increment: ok
  Solid Bits          : ok         
  Block Sequential    : ok         
  Checkerboard        : ok         
  Bit Spread          : ok         
  Bit Flip            : ok         
  Walking Ones        : ok         
  Walking Zeroes      : ok         
  8-bit Writes        : ok
  16-bit Writes       : ok

Done.

Now testing with Memtest86 (not 86+) 4.0a, will post photographic results...see next post for results and conclusion vvvvvvvv

[tek_heretik]

4 sticks @ SPD values except voltage, 1.9v instead of 1.8v as suggested by manufacturer, RAM speed, CAS set manually, CPU, FSB, MCH Latch, all set manually to default nominal values (ultra fine-tuning sub-menus were not touched, all left on auto, you can lost in those for days, lol).

Mestest86 4.0a results, at which I am really surprised it passed this time, will explain at the end of this post why I think it did... http://www.flickr.com/photos/7978633...n/photostream/

Memtest86+ 4.20 results (NOT from any 'buntu disk, I downloaded the stand alone iso CD creator from the Memtest86+ website)... http://www.flickr.com/photos/7978633...n/photostream/

Conclusion:

-Never trust any software, always get a second opinion, or even third.

-This is proof motherboards/chipsets age and weaken (the RAM, or should I say MB, the MB is the problem here, was fine at RAM full speed when the RAM was first installed, it was tested extensively), I still get errors now (though not many) running the RAM at it's full rated speed. I should note here, even though the board is designed to run RAM up to "1333+" (so Gigabyte says), it's 'native' speed is '800' (aka 400MHz).

-MANUAL settings are better than everything 'auto' (at least in my case, this is what finally solved my problem), letting the MB set everything on every boot causes problems.

-Try EVERYTHING before running out and blowing a gazillion dollars on new hardware.

I am finally running with all 4 slots populated, error free but at the MB's 'native' RAM speed, which is fine, only 339MB/s slower, not a big deal, barely noticeable, just glad I didn't have to let the 2nd kit sit and do nothing or go through the hassle of selling it.

Edit: I should also mention I cleaned the computer case intake air filter, upped the case fan to 1500 RPM and the CPU cooler fan to 1600 RPM.

[SamuraiGhost]

The memtest bug in *buntu 12.10 (yes, Kubuntu is affected) always causes failure in the same test (#7), starting at the same MB (129M or so). It has been speculated that the problem is due to it being built with GCC 4.7. If you get errors in different locations or tests, it's probably correct in that your memory is not correct. I first suspected the program was bad when both memory sticks failed in the exact same location at the exact same test, even though it passed all the others. And then it showed my laptop had the exact same problem too. lol I tried again with UBCD 5.1.1 and everything passed.

[SamuraiGhost]

I have read about some motherboards failing early if the motherboard manufacturer's advertised "overclock" memory speed is used instead of what is recommended by the chipset/cpu manufacturer.

[tek_heretik]

I have read about some motherboards failing early if the motherboard manufacturer's advertised "overclock" memory speed is used instead of what is recommended by the chipset/cpu manufacturer.

I believe you, it's happening to mine, but now have it stable at nominal values set manually, thank you ever so much for your replies.

Edit: I will be considering the memory overclock feature thing for my next build, I will just get a fast board to begin with, with matching RAM.