4*2gb 6400 CH5 DHX stuck bit?

[IAmATeaf]

I've got 4 sticks of CH5 DHX and after the PCs been on for a few hours may get an application crash, for example the last few times my sons been playing GTA4 and GTA4 has crashed. First time this happened Windows event log confirmed that it was a memory problem so I restarted and ran memtest86+ which failed the RAM at around the 4.725gb point with what looked like a stuck bit. I tried increasing the memory voltage from 1.86 to 1.88 but this didn't help so I lowered it to 1.84 and memtest86+ tests passed. Problem solved I thought but the next day again same prob, so once again I lowered the volts to 1.82 and again memtest86+ passed. Next day, same problem so I lowered the volts to 1.80 but again the day after same problem.

Now the question I have is is it possible that after some time a bit of the RAM gets stuck? How can I RMA this as I'm sure that when it gets to Corsair HQ it will test fine and lastly how do I find out which pair is causing the problem? Can I use the 4.725gb somehow to identify which stick pair I need to remove and RMA?

OS is Vista x64 by the way.

[Yellowbeard]

On an ASUS MOBO, you need to disable Legacy USB Support to get accurate results with Memtest as there is a conflict with this setting.

Also, with 4 modules, you'll need to increase your memory controller 1-3 steps above the stock voltage to be stable with 4 up. If this does not help, you may need to reduce the memory speed to 667 due to the load on the memory controller.

[IAmATeaf]

Yellowbeard I'll check the legacy USB support, am sure it's already disabled.

I've got the NB volts set to 1.16 already, did have it set to 1.14 but after the last failure upped it another notch, think 1.16 is 3 notches above default. I've also tried upping the memory volts to 1.9 but the RAM kept on failing at exactly the same memory location. The only thing that seems to fix is it to lower the RAM voltage.

[Yellowbeard]

Originally Posted by IAmATeaf

Yellowbeard I'll check the legacy USB support, am sure it's already disabled.

I've got the NB volts set to 1.16 already, did have it set to 1.14 but after the last failure upped it another notch, think 1.16 is 3 notches above default. I've also tried upping the memory volts to 1.9 but the RAM kept on failing at exactly the same memory location. The only thing that seems to fix is it to lower the RAM voltage.

You should be safe up to at least 1.4v on the NB. Try 1.25v or on the NB and take it up from there if 1.25v does not work.

[SammEl]

What voltage you on? Try 2.1v, that seems the favourite with most people.

Also, what are your timings like? Maybe up tRFC etc?

[Yellowbeard]

Originally Posted by SammEl

What voltage you on? Try 2.1v, that seems the favourite with most people.

Also, what are your timings like? Maybe up tRFC etc?

This memory is rated at 1.8v so 2.1v is not necessary.

http://www.corsair.com/_datasheets/T...-6400C5DHX.pdf

The issue here is the memory controller voltage and the capability of the memory controller to run 4 modules at their rated speed. That's why I suggested raising the memory controller voltage, not the memory voltage.

[SammEl]

Originally Posted by Yellowbeard

This memory is rated at 1.8v so 2.1v is not necessary.

http://www.corsair.com/_datasheets/T...-6400C5DHX.pdf

The issue here is the memory controller voltage and the capability of the memory controller to run 4 modules at their rated speed. That's why I suggested raising the memory controller voltage, not the memory voltage.

Ah ok, I never knew the RAM was rated a 1.8v.

But on another note, it is known for mobo makers to undervolt RAM slighty, could this also be a slight issue with the memory? Maybe he might need to push it up to 1.9.

[Yellowbeard]

Originally Posted by IAmATeaf

I tried increasing the memory voltage from 1.86 to 1.88 but this didn't help so I lowered it to 1.84 and memtest86+ tests passed. Problem solved I thought but the next day again same prob, so once again I lowered the volts to 1.82 and again memtest86+ passed. Next day, same problem so I lowered the volts to 1.80 but again the day after same problem.

But on another note, it is known for mobo makers to undervolt RAM slighty, could this also be a slight issue with the memory? Maybe he might need to push it up to 1.9.

It's possible but, he already tried 1.88v. Since it's passing Memtest at all the voltages tried, I don't think the memory voltage is the issue. It's the load of 4 modules on the memory controller.

More info, AN806: http://www.corsairmemory.com/appnotes/default.aspx