[G.Skill] All kinds of Memtest errors, occasional failure to post on cold boot

GearingMass · Jul 8, 2015

G.SKILL Sniper Series 16GB (2 x 8GB)
DDR3 1866 (PC3 14900)
Model F3-1866C9D-16GSR

Stock voltage and speed/timings via XMP.
Occasional cold boot fails to post and throws out memory-related error on the mobo display.

Memtest immediately found errors; errors in almost all tests.
~80 errors through first 9 tests (disabled bit rot test and didn't finish hammer test).

I should probably test the memory each alone and also try the pair in the other two DIMs, but either way I'm guessing I'm going to have to contact G.Skill.

If I do need to RMA, would it be faster through Newegg or G.Skill? I just got my computer back up and running after my debacle with faulty Swiftech parts and I really would rather my computer not be down for another week. Does either of them do cross-shipping?

sobe · Jul 8, 2015

If you can still do so through NewEgg(they may even expedite shipping/cross-ship) then certainly use the Egg. As reference on G.Skill RMA, my last RMA with them was a set of DDR2-800 and I believe it was a week's turnaround time from my sending in to getting the new set. G.Skill certainly has excellent RMA turnaround.

GearingMass · Jul 21, 2015

Well I RMA'd directly through G.Skill, and they were indeed quick. Their tech supported seemed pretty confident that as long as my bios was at its latest version, my replacement modules shouldn't have any trouble at all.

I updated my bios and the new modules still fail Memtest.

At stock 1.50v and stock speeds via XMP, memtest returns about 85 errors in tests 1-9, single pass.
At 1.55v, still stock speeds via XMP, memtest returns 12 errors under same testing conditions.

So overvolting a tad helps, but the memory still isn't stable.

I know the likelihood of two sets of memory both being defective is slim, so I'm led to believe that this model of memory simply isn't compatible with my MB, or another component of the computer is causing this.

I'll be contacting G.Skill tech support as well to see what they think, but if anyone has any ideas, I'm all ears.

Automata · Jul 21, 2015

Can you verify in the BIOS or Windows (with CPU-Z) what the speed/timings/voltage is actually set at? Are there still errors when you back off on the speed of the RAM?

GearingMass · Jul 21, 2015

thideras said:
Can you verify in the BIOS or Windows (with CPU-Z) what the speed/timings/voltage is actually set at? Are there still errors when you back off on the speed of the RAM?

CPU-z reports that the speed and timings are correct. When the voltage was set to auto, the bios reported that it was at 1.50v, but I'll double-check. When I tried it at 1.52v, HWinfo reported 1.515v, and when I bumped it up to 1.55v in the bios, HWinfo reported exactly that.
I'll try bumping the speed down to 1600MHz @ stock voltage & timings when I get home and see what memtest spits out.

GearingMass · Jul 21, 2015

Stock voltage and timings, with speed downclocked to 1600MHz produces almost as many errors as everything completely stock.

sobe · Jul 21, 2015

Hate to say it but I would get in touch with G.Skill again and tell them your replacement kit is producing errors.

Automata · Jul 21, 2015

Have you tried rotating the sticks through the slots (and only using one stick at a time) to make sure it isn't a motherboard issue? Can you try the memory in another system? I'm not saying getting two bad kits in a row is impossible, but it is far less likely than something else going wrong.

ATMINSIDE · Jul 21, 2015

Tried a bump of VCCSA?

Mandrake4565 · Jul 21, 2015

ATMINSIDE said:
Tried a bump of VCCSA?

Thinking the same Atmin!

GearingMass · Jul 22, 2015

thideras said:
Have you tried rotating the sticks through the slots (and only using one stick at a time) to make sure it isn't a motherboard issue? Can you try the memory in another system? I'm not saying getting two bad kits in a row is impossible, but it is far less likely than something else going wrong.

I haven't yet, but I will. It would look like testing each of the two sticks, individually, through each one of the 4 slots, correct?

ATMINSIDE said:
Tried a bump of VCCSA?

Haven't, no. I'm not familiar with the normal ranges of that voltage, so you'd have to suggest a safe bump I could try.

Edit: Current VCCSA is 1.024v (autoset by bios)

ATMINSIDE · Jul 22, 2015

Try 1.1V

GearingMass · Jul 22, 2015

ATMINSIDE said:
Try 1.1V

Ran a pass of that (everything else stock) and still received 50 errors.

I'm going to start moving single modules through each of the slots, as well as try both modules in different configurations to try and collect as much data as possible.

Two questions before I start:

1) I'll run all these different module configurations at stock speed/timings/voltage, but I was curious of one thing -- when the bios sets the dram voltage to 1.500v, it reports as much in the monitoring section, but when I'm in windows, HWiNFO reports in fluctuating between 1.49 & 1.50, averaging 1.495 -- should I bump it up a little so that it's at least is 1.50v minimum in windows, or should I just trust the bios on that one?

2) I recall reading somewhere that letting memtest use all your cores/threads instead of just one can cause it to produce incorrect errors -- is that true?
If that's not true, I'd like to let it use all cores so that it's not 30 minutes per test (would that make the test run faster?)

Thanks!

Mandrake4565 · Jul 22, 2015

Give the Dram a bump it voltage and see if it helps, 1.55 V wont hurt.

GearingMass · Jul 22, 2015

Mandrake4565 said:
Give the Dram a bump it voltage and see if it helps, 1.55 V wont hurt.

I did run it in it's current layout with 1.55v and still had errors.

I'm trying to run them stock so that the only variable is their position, I was just wondering if I needed to give it a tiny bump to make absolutely sure it's running at it's stock-rated voltage.

Mandrake4565 · Jul 22, 2015

Think your best bet at this point is, trying one stick at a time in each memory slot and see if there is still an issue. I feel the issue may lie within the system, as I am doubtful that you would have received 2 sets of faulty sticks. Though, stranger things have happened.

GearingMass · Jul 22, 2015

Mandrake4565 said:
Think your best bet at this point is, trying one stick at a time in each memory slot and see if there is still an issue. I feel the issue may lie within the system, as I am doubtful that you would have received 2 sets of faulty sticks. Though, stranger things have happened.

I'm doubtful too, but I'm hoping so, as replacing the MB or CPU would be excruciating.

Couldn't it also be that neither the CPU, MB or Memory is faulty, just that this model memory happens to be incompatible with this model MB?

Edit: Just got a system_service_exception bsod while testing Rage config edits in-game. No idea if that's memory related or not.

Edit2: Just got another - DRIVER_IRQL_NOT_LESS_OR_EQUAL
Could definitely just be me running a frankenstein config, but when i looked it up, it said memory could be a cause, so I thought I'd note it.

Mandrake4565 · Jul 23, 2015

GearingMass said:
I'm doubtful too, but I'm hoping so, as replacing the MB or CPU would be excruciating.

Couldn't it also be that neither the CPU, MB or Memory is faulty, just that this model memory happens to be incompatible with this model MB?

Sure it's likely that they are just not compatible but highly unlikely. If you were using a set of stick from say 4-5 years ago or some no name brand ram, then I'd would say yeah it's the sticks. Having issues such as you described, nowing that you're getting Memtest errors certainly point to them being memory related. Give G.Skill a call and let them in on the situation. I have had noting but good experiences with their support. One other possibility, however unlikely, is the cpu isn't making good contact with the pins on the board. Other then testing the slots individually to see if one of them are bad I am at a loss.

GearingMass · Jan 24, 2016

[I apologize for necroing this thread, but I didn't want to throw away the prior info]
----------
I went ahead and tested the sticks in every possible configuration and documented the results on the attached sheet below.

Unfortunately, all four of the proper configuration positions (the first four tested) produce errors.
----------
The first column is the memory layout with '1' and '9' being the two unique sticks, and 'x' denoting an empty slot.
The second column is the total number of errors after a single pass of memtest86 without the bitrot test; I don't remember if I enabled the hammer test or not.
The third column includes anything special that happened when the memory was in this configuration.
Everything other than the position was run at default/stock configuration.

Mandrake4565 · Jan 24, 2016

Your thread, you're allowed to Neco

If I'm reading everything correctly it sure looks like either you have 2 bad memory slots on the board. Looks like any time you had a stick or both in the last two memory slots you were fine but any other combination you had errors.

[G.Skill] All kinds of Memtest errors, occasional failure to post on cold boot

Member

Unscathed Member

Member

Destroyer of Empires and Use

Member

Member

Unscathed Member

Destroyer of Empires and Use

Sim Racing Aficionado Co-Owner

Mr. Clean Senior Member

Member

Sim Racing Aficionado Co-Owner

Member

Mr. Clean Senior Member

Member

Mr. Clean Senior Member

Member

Mr. Clean Senior Member

Member

Mr. Clean Senior Member

Similar threads