How To Correct Memory Error Memtest86

[Daharen]

So I've NEARLY tweaked and stabilized my system. Except I went and did a 24 hour Memtest on my RAM and got a single error:
2795.500: Bitpattern Test...
2817.313: Error at 21EE598

Now, it's possible I got one of those 'stray cosmic rays' but I'm gonna just ask people to dismiss that possibility for a moment. Google did tests that show the 'average' system gets 4,000 errors per year over all its DIMMs, a detailed look at the report showed that those errors were all confined to 20% of systems with the other 80% of systems getting absolutely no errors whatsoever, which tells me that it's reasonable and possible to get non-ECC memory that's pretty much error proof barring a cosmic ray switching a bit, and that's my goal here. 

Given that goal I need to figure out if this is a 'timing' issue, voltage issue, frequency issue, CPU issue, Mobo issue, bad RAM, or in a 'best' case scenario if it's just ONE bad sector that I can label as do not use on a single DIMM... 

Now... I'm aware, I can test this by switching DIMM slots, testing alternative RAM, throwing in a different CPU, testing my components in another system, setting the 'default' values, and going through a list of incredibly tedious trouble shooting steps, each requiring their own 24 hour memtest for each configuration to verify the integrity of the RAM given the configuration, and rule out the possibility that it's whatever thing I'm isolating as a variable...

But... Before I go through this tedious, extremely difficult, and time consuming process to chase down a single error, I figured I'd throw a message to the community to see if there were any 'quick' troubleshooting steps I could take to address this error that don't require so much time and effort. Thanks in advance...

[Skiiwee29]

If you only got 1 single error in a 24hr period... Pretty sure you can call it stable and have no issues or worries with it. 

[Daharen]

1 minute ago, Skiiwee29 said:

If you only got 1 single error in a 24hr period... Pretty sure you can call it stable and have no issues or worries with it. 

You're right of course, the 'odds' are in my favor that even if there is a 'single' bit somewhere that is bad, it's not going to hit a critical value and switch it causing a system failure, or major problem somewhere... But... I really am not a 'huge' fan of gambling, even when the odds are in my favor. Given that I know it's possible to get a fully stable system, I'd rather avoid the possibility.

Granted, with 24 hours and 1 error, I think I'm looking at an error every 10 days or so of normal use, approximately 36 a year, and still the odds are pretty good those 36 errors won't be anything major... But... Not a risk I want to take, so any advice would be appreciated if anyone knows any tips or tricks to shortcut troubleshooting an individual error. 

[Skiiwee29]

1 minute ago, Daharen said:

You're right of course, the 'odds' are in my favor that even if there is a 'single' bit somewhere that is bad, it's not going to hit a critical value and switch it causing a system failure, or major problem somewhere... But... I really am not a 'huge' fan of gambling, even when the odds are in my favor. Given that I know it's possible to get a fully stable system, I'd rather avoid the possibility.

Granted, with 24 hours and 1 error, I think I'm looking at an error every 10 days or so of normal use, approximately 36 a year, and still the odds are pretty good those 36 errors won't be anything major... But... Not a risk I want to take, so any advice would be appreciated if anyone knows any tips or tricks to shortcut troubleshooting an individual error. 

Well there are several steps to test and check. 

 

First, 100% reset your BIOS to optimized defaults and run strictly at 100% bare bones stock everything.

 

Second, if everything with the 1st is good, add your overclock back and test each stick individually and see if it produces on a single stick or slot.

[porina]

The immediate question is, is that error repeatable? Do that before you change anything else. I'm not familiar with the exact settings available, but at the least you could run the specific test that failed over and over again. If possible, you might even be able to limit the address range to where the failure happened.

 

If repeatable, you have something to work with. Make a change, see if it goes away. If it is not repeatable, you're never going to prove you fixed it.

[WereCat]

If its only one error in 24h and you don't want to mess with timings and frequency anymore, just bump up the DRAM and IMC voltage by a tiny bit.

Chances are, it will be enough. You can never be sure unless you retest again though. 

[Daharen]

7 minutes ago, porina said:

The immediate question is, is that error repeatable? Do that before you change anything else. I'm not familiar with the exact settings available, but at the least you could run the specific test that failed over and over again. If possible, you might even be able to limit the address range to where the failure happened.

 

If repeatable, you have something to work with. Make a change, see if it goes away. If it is not repeatable, you're never going to prove you fixed it.

I think it is possible to look at the specific address with the professional version of Memtest86 and a specific file made to limit the range, but I am not sure I plan on paying for it... MAYBE as a last resort, and given that, thanks for the suggestion. 

6 minutes ago, WereCat said:

If its only one error in 24h and you don't want to mess with timings and frequency anymore, just bump up the DRAM and IMC voltage by a tiny bit.

Chances are, it will be enough. You can never be sure unless you retest again though. 

This seems to be the most simple and straight forward suggestion. I'll still run another test probably tomorrow evening through when I get off work, and if the same address pops up, I'll just blacklist it and assume its one bad address on the DIMM. If no errors, I'll assume the voltage fixed the stability issue, and if I get an error at another address, I guess I 'might' need to start looking at the CPU or Mobo as the issue, and that sucks... But, this seems pretty simple, and completely eluded me for some reason, so thanks.