Memtest, Asrock and CMZ8GX3M2A1600C8

Hello!

In standard, non XMP profile setup, Memtest on my Asrock board, I tested each stick in the same slot and Memtest came back clean.

I tested the matched pair in the same slots, Memtest came back clean.

I tested all four sticks in and Memtest reported one error in Test 6.

On a lark, I set the XMP profile setup, ran Memtest and one reported error came back in Test 8.

Shouldn't the error have been duplicated in Test 6, and not Test 8?

(or vice versa)

Shouldn't the error have been consistent, given Memtest, and the setup having stayed the same?

And does this mean that the Mobo is the culprit or the memory? I tested the individual sticks and they all came back clean in the same memory slot, ergo one of the other memory slots is the culprit, and not the actual stick?

Am I testing correctly?

On (another) lark I ran the system all night, nothing special running, just virus scans and Ad-Aware, and nothing untoward happened.

Any diagnosis or help?

The background to this story is that I *just* RMA'd one of the two pairs of Corsair memory. I am hoping that I don't have to do it again, but I am also hoping that I don't have to RMA an A$rock motherboard either?

But I think one of the two component types must be playing foul?

Any feedback much appreciated.

Am I testing correctly?

Nothing wrong with the way your testing ...However your issue is more than likely going to be because you mixed two 8 gig kits and not 1 single 16gig kit. Your 4 modules have not been tested with each other like a single 16 gig kit would have been. All it takes is one little thing to be different in the IC's to make them not work with each other.

So it is conclusive that any potential errors are caused by the mismatching of sticks as opposed to natively faulty RAM (esp. since the test came back clean on individual RAM)?

Ahhh..DOH! :(

Either I take my chances and live with the downside of the 4 sticks, or I try to get a refund, re-sell them on EBay and get a set of 4 sticks which are matched?

.f

So it is conclusive that any potential errors are caused by the mismatching of sticks as opposed to natively faulty RAM (esp. since the test came back clean on individual RAM)?

Yeah, this is correct. Have you tried to use just one set or the other to confirm.? See if they run fine with just the new set and then just the old set.

Either I take my chances and live with the downside of the 4 sticks, or I try to get a refund, re-sell them on EBay and get a set of 4 sticks which are matched?

Basically. But if you can get all four to run good , but at a reduce frequency , it's still a huge performance boost. Greater capacity at lower frequency can actually be better for performance than the reverse.

So it's up to you. But in the end a complete matched set would be the best bet. And it would be guaranteed to run at rated specs.

But before you decide either way, can you check to see what your voltages are for both DIMM voltage and memory controller voltage is being set to? There is always a chance that increasing the memory controller voltage will help.

I've tried a few iterations and tests since the last set of (great) recommendations and advice.

Variations and combinations of Lower Frequency, higher voltage, higher frequency, looser timings, as well as XMP with, what the heck I figured, LOWER voltage (got frustrated at one point, yes I was..).. LOL.

The winner so far is looser timings of 10-10-10-24 and higher speed frequency, the XMP profile 1600MHz..

One error in 6 passes, 10 microbits in size. Test #4.

Ironically, after a nice day away from the RAM, but letting Memtest run for many many hours, I got back home and quickly checked, and there wasn't a single Error.

However, after supper and such, a few hours later, the (only?) error showed up on the radar. My outlook on life was so positive only a few hours earlier. ;)

I see the frequency of error(s) in total, per hour, as a probability of a memory 'blurp' (a bad hiccup) and so far this is the lowest probability that I've seen thus far.

The big decision is keep it and take my (relatively okay) chances, or do the deed, Ebay one pair, use the system at a stable, to spec, 8gigs of RAM, and wait for the full slot amounts (ie 32gigs) to one day approach my level of budget.

*sighs* argh. :( So close. :(

Thank you for all the advice and recommendations Peanut!

Very helpful and much appreciated!

AT the very least a good learning process. :)

No worries Froozeball. Glad to help!

The winner so far is looser timings of 10-10-10-24 and higher speed frequency, the XMP profile 1600MHz..

One error in 6 passes, 10 microbits in size. Test #4.

Is this with all four sticks or just 2?

The rig now has 16gbs of 4x4sticks ...

Ran the usual virus/adaware/MS defender scans overnight and everything was still up and running this morning. So at least, I'm relatively certain that I can protect my system without giving up the ghost! :)

Doing anything else is still up in the air. :o