Line Rate

Blogging because someone out there is probably banging their heads against the wall just like I did today.  I've had some "interesting" stability problems on my main desktop, an Asus P9X79 Deluxe motherboard, Intel i7-4930k CPU, and lots of g.skill 4GB sticks (32gb total).

I've had memory go bad on me before, so I figured that would be a good place to start.  I checked the basics (timing, voltage, etc.) and they were all configured properly in the BIOS. 

I had a few copies (and various versions) of Memtest86+ and MemTest86 burned to USB sticks.  I can't remember which one I started with, but both of them started spraying errors after a few tests.  I went down to a single memory stick and still had the errors.  I swapped the stick around with other sticks.  Still no luck.  All 8 sticks were erroring out.  At this point I was pretty concerned because that could mean I had motherboard or CPU problems. 

I quickly grabbed new/current versions of both utilities and tried them out.   One of them ran overnight without any issues, and the other failed as soon as I got to test 3... very perplexing.  I fiddled with timing, with ram slots, and still couldn't figure it out - although I did start seeing a pattern.  The test(s) that passed were all single-CPU.  The tests that failed were all SMP tests.  

After a little research, I ran across this gem of information.  Essentially, there are some motherboard+cpu combinations out there that exhibit CPU register corruption when used with certain versions of Memtest.  This is the reason why the single CPU tests ran overnight just fine.

With the stability problems, I wasn't happy with only testing a single CPU (in case my problem was  CPU-related).  The recommendation is to use SMP with Round Robin.

 "the work-around implemented was to change the default CPU selection mode to round robin. In this mode only one CPU is used at a time, but after each test the CPU in use is rotated. So all CPUs will still get used, but only after a longer period of time."

I'm feeling a little better about my system now, at least as far as RAM is concerned.  I have a round robin test running now and will let that go for another 24 hours... but so far so good.