haswell platform stability problem

[Manoa]

I got a 4790K on a Gigabyte Z87X OC-Force and on it I have 4x8GB Corsair Dominator Platinum CMD16GX3M2A2400C10
using the XMP profile

have BSOD every few days, once I saw IRQL_NOT_LESS_OR_EQUAL and other times SYSTEM_SERVICE_EXCEPTION

tried to reduce from 4600 mhz to 4400 mhz, no help
tried to increase voltage from 1.24 to 1.265, no help
(1.24 froze the system during prime95 large within less than 5 minutes all the way while increasing to 1.265, at all times program chose FMA3 instructions and that increased my cores temperature to almost 100C, damn intel and their shitty grease :x - but it was stable for 8 minutes at the end at 1.265V - the point at which the thermal conductivity of the intel shitty grease allowed the cores to reach 100C)

tried to reduce RING to 4500 and 4400, no help
tried to increase RING voltage, no help
tried to increase RAM voltage from 1.65 to 1.70, no help
tried to increase SA/DIO/AIO by +0.05V and then by +0.1V, no help
right now disabled the XMP profile and running 1333 mhz 1.5V HCI memtest

but throughout the all history ran consistently memtest86 and it never found any errors

so far stable for a few days at 1333 mhz 1.5V

any ideas ?

 

[nerp]

How long have you run memtest?

What happens at pure stock settings: voltage/speed?

What parts have you swapped out to troubleshoot?

My thought - your PSU is not up to snuff. Could be PSU.

 

[Manoa]

memtest86 1 full pass it takes around 1:30 hours
didn't tried yet stock settings but if 1333 1.5V will have problem I will
didn't swapped parts yet, everything is new so I assume it's not a failure in one of the components
power supply is also new seasonic platinum-1000
tested now for several hours in 9 concurrent instances of HCI memtest and no BSOD and no errors in RAM reported for 300-400% already


used to have stability problems due to low VRIN, but ever since I increased that to 2.050V they appear to be gone (7-zip benchmark does a good job at detecting that type of stability problem)


I have a cooling system on all the components, water block for CPU, water block for mobo that covers all the CPU MOSFETs + PLX chip + SB, water block for GPU, 2 triple radiators cooling the entire system. San Ace H1011 12 cm fan at full speed directly on the DIMMs, a fan on all capacitors near the CPU, the computer with all parts is a little over a year old




the reason I suspect it's RAM is because CMD16GX3M2A2400C10 is sold as 2x8 and not 4x8 and I read somewhere that the RAM as is sold as 2x8 will guaranteed to work at it's intended XMP but it doesn't mean that 4x8 will, that why I increased voltage to RAM to 1.7V - I thought it need more power to run XMP with 4 modules, I suspect that it may not be that simple and I may need to play timing games with it to stable it at XMP, if it's the RAM that causes the stability problem in the first place - I hope a few minor relaxations to primary timings will help

 

[nerp]

Yes, I would not be surprised if the 4 sticks is stressing the memory controller and leading to these errors. You could yank two sticks for the time being and though you might not like dealing with half the RAM for now, it could confirm the case if you have no more BSODs or errors for an extended period of time.

 

[Ketchup]

Run it stock speed and voltage. Memory at SPD ( no XMP). Is it stable?