Need help understanding MemTest x86+ results

[Skypix7]

Hi folks, added this post to my trouble-with-p8z68-v Pro/Gen3 build thread but nobody's chimed in all day so starting a new thread here.

I'm up and running but weird and random BSODs, sometimes mid-boot, sometimes in Win update, sometimes in reboot, sometimes hangs shutting down, other times works like a charm...but only for awhile.

Tried...tried to run Prime 95 stress test, crashed in 2.3 nanoseconds.

So... I ran a boot disk version of memtest 86+ v 4.2, all night last night... 12 hours. Seems to me memory is the culprit.

Would sure appreciate interpretation and suggestions re these results, I'm a noob in this area, I've never done much in the way of testing:

Test 0 thru 4: no errors
Test 5: 244 errors
Test 6: 9
Test 7: 1270
Test 8: 16
Test 9: 0

Errors per memory slot:
0-3: 1 for each slot
All the rest, 4-16, no errors
Bits in error -total 1 min 1 max 1 avg 0
Max contiguous errors 1

Total errors 3152 (yeah, 3152)

Just a wild guess here: memory and mobo don't like each other, or the mem is messed up.

Is it obviously RMA time? Don't want to spend another two days chasing down slots, which stick etc. unless there's a good reason, I'll swap it out for something on the Asus compatibility list. This exact memory, although it met all specs, wasn't on that list.

This memory is G Skill 4x4GB Ripjaws X DDR 1333 (1.5v 7-7-7-21 timing, bios saw it correctly, I also pushed Mem Ok button initially which seemed to help, but only temporarily. I figured since smaller sizes of G Skill were on the list, maybe Asus hadn't gotten around to putting these 4x4s on yet. Their notoriously slow in keeping up to date, the DRAM compatibility list was last updated in May!!...makes it kind of tough to know what's good to buy.

Thanks for any insights fellow Anandians. Live long and prosper!

 

[BonzaiDuck]

I use the RipJaws "GBRL" kit -- 2x4GB = 8GB. It looks as though you have all four slots filled. The GBXL RAMs were supposedly made specifically for the Z68 motherboards, but the GBRLs work just fine -- stellar, if you were to ask me . . .

Start by finding any pair of sticks that don't generate errors such as those you described. Put those aside. Test each module of the remaining pair in "single channel" mode.

If it's only a single module, you can continue to run the two good sticks (8GB total), set the third good module aside, and send them the defective unit.

This way, you'll also confirm for yourself that it isn't a motherboard problem or some other such cause. G.SKILL allows you to replace single modules of a kit, so this approach may offer the least inconvenience.

 

[krose]

Yes you could have a bad stick of RAM, but it is also likely that the memory controller on the CPU is a having a hard time with all four slots filled. You must increase the QPI and/or system agent voltage to stabilize the memory. I had the same problem on my Gigabyte Z68 with 4x4Gb GSkills, tested all the sticks separately and they were fine, but I couldn't get all four to work together so I went with 2x4GB. Test one pair and if it passes test the other. If both pairs are fine it is the voltages.

 

[Skypix7]

:whiste:Thank you both, excellent detailed suggestions but here's what solved the problem (at least I think it's solved based on another overnight memtest (8 hours) without a single error.

Solution: late last night, roaming around the net like Sherlock Holmes hoping to uncover information it seems is buried to keep people busy looking for it (as in hide in plain sight), I found a reference in a forum to switch on XMP in Bios.

Of course there are references to XMP along with many other things in the manual but where, I wonder, does it say "May solve memory problems?"

Anyway, the first boot after I enabled XMP, my system storm became calm seas.

I'll do more research on what it is, but after three hours, not a hiccup, not a blue screen, not so much as a display stutter (running two 23" screens).

Wonderful!

It's fun playing Sherlock. But wouldn'lt it be great if motherboard manufacturers hired somebody to write a manual on all the things to try that are built right into their own products so you wouldn't have to go through the "try slot 1 with stick 3, then stick 2 in slot 3" etc. troubleshooting to solve what turn out to be simple problems?

Also, all four slots populated seem to be working just fine.

So thanks for the help, but my hand is up voting for some unified field theory of computing that makes builds less time-consuming, like mine became when my memory acted so schizoid.

BTW, my memory readout according to memtest show 666 mHz at DDR3 (1333) and the correct timing according to G Skill: 7-7-7-21. Can anyone tell me if that 666 mHz is a correct value for the memory? I'm having a devil of a time figuring that out. (groan).

thanks again for the help

 

[krose]

Your memory at 666 is correct, DDR is 2x that or 1333 MHz. Glad you got it working.

 

[bryanW1995]

I use the RipJaws "GBRL" kit -- 2x4GB = 8GB. It looks as though you have all four slots filled. The GBXL RAMs were supposedly made specifically for the Z68 motherboards, but the GBRLs work just fine -- stellar, if you were to ask me . . .

Start by finding any pair of sticks that don't generate errors such as those you described. Put those aside. Test each module of the remaining pair in "single channel" mode.

If it's only a single module, you can continue to run the two good sticks (8GB total), set the third good module aside, and send them the defective unit.

This way, you'll also confirm for yourself that it isn't a motherboard problem or some other such cause. G.SKILL allows you to replace single modules of a kit, so this approach may offer the least inconvenience.

I had to rma some memory from a 3x2gb set in 2009, they made me send all 3 sticks back. Of course, that was crucial. The new set didn't work any better, unfortunately...

 

[Skypix7]

Your memory at 666 is correct, DDR is 2x that or 1333 MHz. Glad you got it working.

Right...DDR...double data rate...thanks for clarifying that! Working like a champ now...all I have to do is figure out how to force Windows Update to update all those failed patches (18 or so didn't install), which of course being windows isn't a simple thing to do.

 

[Puppies04]

Windows should pick up on the fact that they didn't install give it a day or so

 

[Skypix7]

Thanks Adam, I have noticed it's picking up one or two at a time...long as I'm not messing things up by moving on, you never know with all this stuff talking to each other and having secret conversations we're not privvy to.