Take the 2-minute tour ×
Super User is a question and answer site for computer enthusiasts and power users. It's 100% free, no registration required.

I'm running Ubuntu 12.04 on a P8P67 Deluxe Asus motherboard, with 3 x 1TB+ WD Caviar Green SATA disks.

When I run:

cp /media/X/ubuntu.iso /media/Y/ubuntu.iso
cmp -l /media/X/ubuntu.iso /media/Y/ubuntu.iso

where X and Y are each one of the three disks (including X=Y), I get corruptions like (multiple runs of cmp):

279222903 212  12
 80534135 270  70
689056375 356 156
704375415 221  21
 77208183 302 102
 18418295 305 105
524483191 315 115
120892023 322 122
638548599  25 225
638548599 225  25

What is the underlying cause of this pattern?

Other information: I RMA'd the board back in 2011 because of "SATA degradation issues" and got the current board as a replacement. The current corruption rate is around 1 byte corrupted every 1.5 GB, though it is sometimes as high as 1 byte in 50 MB.

share|improve this question

migrated from serverfault.com Mar 7 at 22:06

This question came from our site for professional system and network administrators.

2  
Have you performed extensive memory tests to make sure it's not your ram? –  Mxx Mar 7 at 22:07
    
ran memory tests from an ubuntu live CD - no issues there –  Karl C Mar 7 at 22:12
    
I'm not sure if you ran that test within 10 minutes between our replies..If so, I'd re-run a much more extensive test, if possible even overnight. –  Mxx Mar 8 at 3:30
    
I had ran a couple passes of memtest86+ a couple days ago - I am re-running it for the next 12+ hours tonight... –  Karl C Mar 8 at 4:23
1  
All the corruption is at addresses of the form ...101001000100, i.e. at the same location modulo 4096. This kind of pattern is a telltale sign of bad memory. The RAM on the motherboard is the most likely culprit. It could be RAM elsewhere, for example in the SATA controller, but such RAM tends to fail less because it stretches the boundaries of reliable engineering less. RAM tends to fail only for certain usage patterns which memtest86+ tries to emulate but occasionally doesn't. I strongly recommend trying with different RAM sticks. –  Gilles Mar 8 at 13:58

1 Answer 1

up vote 0 down vote accepted

If you print out the corrupted addresses in a binary base, a pattern emerges:

10a49a77
04ccda77
29122a77
29fbea77
049a1a77
01190a77
1f42fa77
0734aa77
260f7a77
260f7a77

Not only is all the corruption on the same bit in a byte, it's always at addresses of the form ...101001000100, i.e. at the same location modulo 4096. This kind of pattern is a telltale sign of bad memory (buses and cables don't stick to corrupting one bit in 32768, and software bugs hardly ever do either).

The RAM on the motherboard is the most likely culprit. It could be RAM elsewhere, for example in the SATA controller. However the motherboard RAM is the one that most stretches the boundaries of reliable engineering and is thus the most failure-prone.

RAM tends to fail only for certain usage patterns which memtest86+ tries to emulate but occasionally doesn't. Use a recent version by preference, but even then it sometimes gives a green light to memory that fails in real use. I strongly recommend trying with different RAM sticks.

share|improve this answer

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.