Over-volting DIMMs merely to make them work with two per channel?

[Timothy Miller]

Hi,

This problem happened with the same DIMMs on a different motherboard. I think the problem is most likely signal integrity but possibly Vdd fluctuations causing memory retention errors.

I have a new ECS H87H3-M motherboard, and I've installed four 2GB Crucial 1333MHz DIMMs. I put everything in default configuration, and it works okay until I do something memory-intensive like compiling a Linux kernel. The whole system will lock up, with garbage on the screen and everything.

If I put in one DIMM per channel, it works fine. I have tested all 4 DIMMs this way.

As I had to do with my old MSI board (which was a flaky piece of junk in other ways), I have tried stepping up the DIMM voltage 20mV at a time. As I do that, the failures have become progressively less catastrophic, to the point that now at 1.72V, I've managed to get through one whole kernel compile, but not a second one. (The nominal voltage for these memories is 1.5V.)

Memtest86 ran over night with no errors, so there are no hard failures, although something could be borderline, I guess.

Is it normal to have to boost the voltage to get two DIMMs per channel to work? Is it normal to have to boost it by over 200mV? What are the expectations here?

Also, since I've had to increase the voltage for another motherboard, and the RAMs work fine with one per channel, does this probably rule out the new motherboard and CPU? (I'd have to try other DIMMs two per channel to be totally sure, but I'm trying to work out the likeliest causes.)

Thanks!

 

[Timothy Miller]

Stopped at 1.8V

This being a Core i5, Haswell, I was nervous about doing this, but I went as high as 1.8V before stopping my experiments. I cannot even at that voltage get the DRAM to be reliable. But all of them work as long as there's only one per channel.

Time to call Crucial, I guess.

 

[Theocnoob]

Intel would tell you to avoid RAM over 1.5V on Haswell. Obviously you can go higher but you don't want to damage your CPU/RAM.

The obvious thing to try here is a different set of RAM. Or just get 2x4GB and solve your problem completely.

 

[Timothy Miller]

I only over-volted briefly for experimental purposes. Hopefully that hasn't caused any long-term damage. To make these DIMMs work on the old motherboard, I had to boost to 1.58V, which I'm sure isn't any kind of big deal. Anyhow, I'm limping along at 4GB for the moment, until I can get a replacement or RMA or something.

Thanks.

 

[Woomack]

I don't think it's memory issue but hard to say.
New memory IC can work up to 1.8V+ ( depends from IC some have max 1.975V ). It's of course not recommended but shouldn't damage CPU or memory.
In this case I would check other board or BIOS as I also don't think it's IMC issue.

 

[Niku-Sama]

sounds like a ram setting/timing issue to me, 4 slots populated trying to run 1T instead of 2T

Edit:
http://www.ecs.com.tw/ECSWebSite/Pr...DetailID=1460&CategoryID=1&MenuID=172&LanID=0

starting at page 59 of your manual it gives the settings and timings for the ram, I hate how every manufacturer has a different name for the same command or setting but I think yours is NMode, it should only be switchable between 1 and 2, If ti goes higher than 2 than its not and and I don't know what they would have labled it.

any way if it only goes to 2 set it to 2, don't do auto because even if its autoing to 2 it doesent always work right.