Can an OS config be OK after RAM failed?

[RickC]

I've had memory that had errors when I received it but this is the first time I've experienced this.

I built a system. It passed MemTest86+ 4.1. I installed the OS and the system passed Prime95 torture test.

The system worked great for a week and then after a week starting blue screening.

The system then failed MemTest86+. I tested each of the two modules individually in the same slot, one passed, one failed. I got new memory with an RMA and it passed MemTest86+.

Everything seems fine now. The system passes sfc /scannow, and passes chkdsk /f on all partitions.

Does anyone have experience with this? Can a configuration be trusted after running with a failed memory module that has been replaced or should I rebuild the OS from scratch?

Thanks,

Rick

[mdettweiler]

Originally Posted by RickC

I've had memory that had errors when I received it but this is the first time I've experienced this.

I built a system. It passed MemTest86+ 4.1. I installed the OS and the system passed Prime95 torture test.

The system worked great for a week and then after a week starting blue screening.

The system then failed MemTest86+. I tested each of the two modules individually in the same slot, one passed, one failed. I got new memory with an RMA and it passed MemTest86+.

Everything seems fine now. The system passes sfc /scannow, and passes chkdsk /f on all partitions.

Does anyone have experience with this? Can a configuration be trusted after running with a failed memory module that has been replaced or should I rebuild the OS from scratch?

Thanks,

Rick

Yes, you should be fine. There's nothing about the OS that was actually built "from scratch" on your computer (well, not unless one is running Gentoo Linux, which you're not); it's all precompiled binaries copied directly from the Windows CD/DVD. There are all sorts of checksum verifications involved in this copying process, so if anything got corrupted it would know and re-do the faulty part. When the Windows installer reports that the OS is installed and ready to use, all should be fine.

[retina]

Hmm, I would say the opposite of mdettweiler. Any time you have a random corruption there is no way of knowing what it affects or what consequences it has/had.

If you want to be 100% certain that there is no lingering nasty little problem awaiting you then you should consider reinstalling again.

If you are content with 99.99% (I just made that figure up, seems about right to a few orders of magnitude) certainty of okayness (at least as far as the software allows) then just carry on with what you have.

[mdettweiler]

Originally Posted by retina

Hmm, I would say the opposite of mdettweiler. Any time you have a random corruption there is no way of knowing what it affects or what consequences it has/had.

If you want to be 100% certain that there is no lingering nasty little problem awaiting you then you should consider reinstalling again.

If you are content with 99.99% (I just made that figure up, seems about right to a few orders of magnitude) certainty of okayness (at least as far as the software allows) then just carry on with what you have.

Hmm...I would have thought the possibility of any critical operating system files being corrupted as a result of bad memory during the OS install next to nil, since everything is verified with a checksum after copying. (At least that's how I'm pretty sure it is in Linux...I wouldn't know about Windows since it isn't as transparent as to what's going on during install. I would hope, though, that it does some kind of checksum verification.)

But, you're right, 99.9999% is not 100%. Still, it's close enough that I would not consider closing the gap worth the extra 1-2 hours of my time it would take to reinstall the OS. That's assuming the computer doesn't have any stuff from personal use on it that would require backing up and restoring; if that was the case then the tradeoff would be even worse for reinstalling.

[retina]

The corruption could be in a configuration file or some other machine/user specific setting. Just because all the downloaded/installed files are checksum verified doesn't mean you have a guaranteed good install.

[mdettweiler]

Originally Posted by retina

The corruption could be in a configuration file or some other machine/user specific setting. Just because all the downloaded/installed files are checksum verified doesn't mean you have a guaranteed good install.

Right...as you said, it's not guaranteed, but it's overwhelmingly likely to be OK, with the likelihood of that being on the order of 99.99% or so (assuming I interpreted that statement correctly?). Personally, I would consider that "good enough" that I wouldn't bother reinstalling.

(Then again, in prime searching we go out of our way to run a separate N-1/N+1 test to prove PRPs, which are perhaps even more likely to be prime than the computer in question is to be OK...so I see your point. I guess it boils down to the question of how much additional human time one can justify spending to eliminate a <.01% possibility. )

[TimSorbet]

As they've said, it's probably ok. My advice would be to just try to use the OS. If you still get blue screens even though the RAM, CPU, and hard drive are functioning correctly, (as tested by torture tools appropriate for each part) reinstall.

[joblack]

Originally Posted by retina

The corruption could be in a configuration file or some other machine/user specific setting. Just because all the downloaded/installed files are checksum verified doesn't mean you have a guaranteed good install.

You will be fine. If not you might reinstall your system. More important is having backups of your data so if one of these are corrupted you can play it back.

[Xyzzy]

http://packages.debian.org/lenny/debsums
http://packages.debian.org/lenny/integrit