Memtest

[IRLeetly]

Hi guys, gals and other beings.

I was wondering how accurate memtest 86+ is? I've used it on ram in the past, but tried it on a new set today and it had found errors before the first pass was even complete.

Ive just upgraded to 8gb of the Kingston HyperX Genesis that was on offer last weekend, I've tried clearing my cmos, ran the test again (with stock settings) and still same amount of errors before the first pass was done. Even tried jiggying about my OC so instead of my usual 191x21(1528Mhz) to 200x20 to give me 1600mhz on the button.

Im using a GA P55A-UD3 which doesnt allow a voltage of 1.65v for ram only 1.64v or 1.66v. Of course I've been using the former of the two, to not risk any damage.

Any other info needed (within reason ) I'll be happy to post.

Is the ram needing to be returned or....?

Any advice & help will be muchly appreciated and possible new names added to my Crimbo card list.

Gracias!

P.S. My Crimbo cards aren't Hallmark, so don't get your hopes up

 

[Plec]

Have you tested each stick individually - and in different DIMM slots to try and isolate which stick/DIMM is faulty etc (may be both sticks not unknown)?

Also make sure you're using the latest release of memtest.

However, if you are using the latest release it would seem to indicate you have memory related problem - be it the sticks themselves, the MB or the controller. It's usually the memory and if a stick/s fails consistently in either slot you can rma away...

EDIT: Obvioulsy check that the timings, frequency, command rate and voltage are set up correctly (CPUz will help verify this - but it's best to set it up manually in the BIOS).

 

[IRLeetly]

Yeah I'm using v4.20, got it yesterday to try out the ram I am going to sell that this set is replacing. That set came up error free after 6 hours of testing

I will try each stick by itself, what would be considered a decent amount of testing? I usually do about 5-6 passes.

EDIT: Using 2T command rate and 9-9-9-24, always check CPU-Z and set timing manually first

 

[Plec]

Yeah I'm using v4.20, got it yesterday to try out the ram I am going to sell that this set is replacing. That set came up error free after 6 hours of testing

The fact that your original set passed is very damning for the new set.

I will try each stick by itself, what would be considered a decent amount of testing? I usually do about 5-6 passes.

6hrs is about right - but i would first test each stick for 1 pass as the faulty stick may give up errors straight away and you could waste 6 hrs testing the healthy stick first (they may both be faulty but even if one stick is faulty you still have to return both sticks - so if one stick fails straight away save yourself some testing time).

If they both pass after one pass then test each stick for 6hrs.

 

[IRLeetly]

So far each stick has passed two passes in slot one by itself, leaving the 2nd stick in for a few hours more.

Hope you don't mean 6 hours per slot, that's like 24 hours of testing each hehe.

Forgot to add in my OP the old set of ram I have was 4gb of corsair xms3, btw thanks for the advice so far. I'll keep things updated here.

 

[Plec]

Forgot to add in my OP the old set of ram I have was 4gb of corsair xms3, btw thanks for the advice so far. I'll keep things updated here.

NP, what was the frequency of the XMS3?

So far each stick has passed two passes in slot one by itself, leaving the 2nd stick in for a few hours more.

Just to clarify - is this at 1600MHz?

Hope you don't mean 6 hours per slot, that's like 24 hours of testing each hehe.

That would be the harsh reality - but if you could test each stick in the 2nd slot now just in case it's a DIMM/memory incompatibilty issue (for whatever reason).

If they both pass in that slot after a couple of passes then you would be in for a long haul... :/

Additional Did you check to see if they passed at 1333MHz in dual channle mode. The reason i mention this is that it may be a compatibilty issue - in that your board is spitting the memory out at 1600MHz in dual channel mode. It shouldn't but if each stick passes memtest individually at 1600MHz it would need to be considered. Do you have another machine you could test the memory in?

 

[IRLeetly]

The Frequency of the XMS3 was the 1528Mhz at 191x21, which is what I'm testing at now.

The current OC im running won't give me 1600 exactly, to do that id have to change my OC which I wouldnt know if that was stable or not or if it affects the test.

Unfortunately I dont have a 2nd DDR3 rig, the one I'm replying on now is DDR2.

Will start the 2nd slot testing now, then try 1333Mhz also after. Oh also I didnt check to see if they passed at 1333 in dual channel mode.

 

[Plec]

>snip<

Sorry, somehow i glossed over the fact your system was clocked.

I should add that you should test the memory at stock CPU settings - just in case the clock is the contributing factor of the instability (unless you've tried this already).

You can save your present clock in a profile within the BIOS.

 

[IRLeetly]

Hows this for as crazy as Katie Price speaking at Oxford.

Cleaned out the DIMM slots, swapped which stick is in which slot, like vice versa'd them if you get me (hope I'm explaining ok, still on first coffee & should really be boosting out the door soon).

9 passes and about 9 hours testing later, no errors!!!

Maybe there was dust or I didn't seat one stick properly into its slot, as usual ram is like "apply pressure BUT not enough to buckle/bend your mobo".

Might IBT or prime Blend whilst I'm out, I finish at lunch, just to make sure.

Helps been...well a great help. Thanks Unlike another certain forum with nearly the same name where you can get ignored

Now I can return to putting my stock corsair fan as the push fan and my GT as the pull as it has slower RPM's, but I think its narrowly pumping out a bigger CFM by like .5 CFM.
Ah well another post for another day.

 

[Plec]

Great stuff - it may even have been a bit of 'post manufacturing residue' on the memory themselves. Either way glad to read that it may be resolved.