RAM won't POST at XMP after 6 months

[mickleby]

I'm curious to know if others have seen this issue? Is it worth the hassle to RMA? Will the issue get worse? Or maybe, can there be another explanation such as mobo/BIOS?

I have a set of G.SKILL Ripjaws Z Series 16GB (2 x 8GB) DDR3 1866 9-9-9-24 1.5V F3-1866C9D-16GZH. They are 6 months old with minimal use. When they were new they wouldn't POST much over 1866MHz, but that seemed acceptable. Recently, I left the system idling for 4 weeks--detected no issue. I made some BIOS changes to enable WOL (Wake On LAN). Now the system will not POST at XMP / 1866MHz.

To try to track down the issue, I have updated the BIOS. I have cleared the CMOS. I have reset and reconfigured the BIOS. Etc.

I have tried putting the BCLK etc back to stock and simply booting from XMP1 and XMP2. Neither will POST any longer.

However, I can POST and pass RAM test at around 1650MHz. I'm thinking the difference is minimal and not worth TOO MUCH bother, other than I have a lifetime warranty and paid a premium for speeds I cannot attain after 6mos minimal use. Another annoyance is the auto o/c software from ASUS wants to utilize XMP profiles. Again, a very minor inconvenience.

So please help: Does this happen? Often? Have I isolated the issue to the RAM? What will the G.Skill RMA process entail? Is it worth the hassle? Will the RAM continue to degrade? (Other answers to questions I haven't considered are quite welcome.)

If I decide to RMA: Are there red flags I need to be certain not to trigger? Like over-volting? Actually, I did not over-volt the RAM except in o/c testing and even so only to 1.60v for a brief period basically 6 months BEFORE the issue arose. Are there things I need to be certain NOT to say so that I have a successful RMA?

[mmettin]

In this particular case, the memory controller is on the processor.

An I7-3770k processor Max (stock) memory speed is 1600Mhz @ 1.5 volts (at the memory)

You may have damaged the processor by running at a higher voltage and clockspeed.

After you clear the cmos...If you manually set a profile to 1600 @ 1.5 volts does the rig complete the post cycle ?

[mickleby]

After you clear the cmos...If you manually set a profile to 1600 @ 1.5 volts does the rig complete the post cycle ?

Yes, it will POST up to about 1650.

In this particular case, the memory controller is on the processor.

You may have damaged the processor by running at a higher voltage and clockspeed.

If the mobo will oc the memory up to 2400, is the controller still the one on the i7?

Also, do you think 1.55v is "running at a higher voltage"? I think I pulled it back to 1.5v, but my signature says 1.55v so I may have left it there. Either way, it seems completely happy at 1660MHz 1.5v and will not post at XMP 1866 whether 1.5v, 1.55v, or 1.6.

[mmettin]

Yes, it will POST up to about 1650.

Since the controller is on the processor, you would effectively overclocking the memory controller.

If the mobo will oc the memory up to 2400, is the controller still the one on the i7?

Yes.

Also, do you think 1.55v is "running at a higher voltage"?

possibly.

[Poci]

DDR 3 memory can handle more then 1.5. In fact its at around 1.9v where it will literally fry it. I run mine at 1.8 with no problems at all, of course make sure the RAM at least has heatspreaders.

[mickleby]

DDR 3 memory can handle more then 1.5. In fact its at around 1.9v where it will literally fry it. I run mine at 1.8 with no problems at all, of course make sure the RAM at least has heatspreaders.

They sell a lot of 1.65v DRAM, and I find no warnings to avoid mixing them with Ivy Bridge. Surely by now we'd have heard something if this is an issue. This is a year-old response from Intel, though, saying NOT to do it "just in case". I note your system is AMD, btw.

Apparently Intel promises a margin of error, 1.5v +/- 5% (1.575v max). If this is true then I was still within the window.

[Poci]

AMD also has the memory controller on the cpu chip. But the voltage is labeled NB Voltage in bios.

[mickleby]

I have some other sticks that used to run at 2100MHz in the system. I'll swap them in and see what happens. If I can oc the old sticks it seems reasonable to suppose the "formerly 1866 stix" degraded after some burn-in.

[Poci]

Just run at rated speeds timings and voltage, of course that was XMP is. If its not working then time to RMA.

[mickleby]

I appreciate the responses.

I was hoping to gauge how common this is. So far hard to say, maybe not common. I also wonder how long it will take to get the RAM back from G.Skill.. I guess I will find out the old fashioned way

[Poci]

Should take about a week

[mickleby]

I swapped in other DRAM that used to oc to 2200MHz in this same system... They WON'T oc anymore, either.

So, probably not the RAM.

I guess I'm left with mobo and/or CPU. I think 1.55v is within the spec for the i7 memory controller, so I'm leaning toward the motherboard.

Any suggestions for how to isolate the issue will be appreciated. I don't see how I could RMA either the mobo or the cpu with only this data. I could live with it, and feel a bit cheated and unwilling to build quite so high end; but, darn it! People make this work, so I should be able to, as well.

[Poci]

Do you have any other boards to test on?

[mickleby]

no other ivy bridge boards at the mo... I could just hold until I get other parts.

And besides, I do have an oc'd i7 with 1600MHz+ ram. The diff between 1600MHz and 1866MHz is trivial in almost all cases (except benchmarking). If it's the mobo or the cpu, maybe other issues will arise later to provide more clues. I really only regret the extra ($20?) money for the 16GB 1866MHz vs 16GB 1600MHz.

And, this is a hobby and learning exercise. I want to understand what changed, why I got 2100MHz before and not even XMP now.