High Failure Rates for DDR2

[gg1978]

Re: High Failure Rates for DDR2

Perhaps with the smaller process sizes that DRAM is fabbed one now, might be a partial explanation.. Anybody remember the Asus P3B-F?? One of the reasons for its legendary stablity under o/c conditions was its default 3.5V for the RAM and Northbridge, with it being jumper configureable to go to 3.65V or so.. Then again, that's only about a 6% overvolt there..

For later DDR memory, 2.6V was common for DDR400 modules, the only ones that typically ran at 2.5 were DDR266 and maybe 333..

[gdement]

Re: High Failure Rates for DDR2

Perhaps with the smaller process sizes that DRAM is fabbed one now, might be a partial explanation.. Anybody remember the Asus P3B-F?? One of the reasons for its legendary stablity under o/c conditions was its default 3.5V for the RAM and Northbridge, with it being jumper configureable to go to 3.65V or so.. Then again, that's only about a 6% overvolt there..

For later DDR memory, 2.6V was common for DDR400 modules, the only ones that typically ran at 2.5 were DDR266 and maybe 333..

JEDEC DDR400 is 2.6v +/- 0.1v.
JEDEC DDR333 and below is 2.5v +/- 0.2v.

basically the allowed voltage didn't go up, just the tolerance got more strict.

[MixMasta]

Re: High Failure Rates for DDR2

I guess my memory of my memory is slightly different than some peoples...

[NxB]

Re: High Failure Rates for DDR2

Laptop memory is all over the board. I've had crucial and infineon ram go. Mostly gateway and HP/Comcracks that were overheating. More and more errors in memtest86 until the computer stopped booting. Reflow only helped for a little bit, I guess I should try to reball a few by hand since the package is so small.

[bgavin]

Re: High Failure Rates for DDR2

I just finished a 24 hour burn-in of Corsair Dominator DDR2-1006 at JEDEC specs. It is running 1.8v at 5-5-5-15 without error.

I'm certifying the next 2x2gb pair now.
The 3rd set will be 24 hours from now.

One of those sets is going out as warranty replacement to a client machine out of state. He is comfortable with replacing his own sticks. This morning we walked through CMOS and I made sure he was set back to 1.8v and 5-5-5-15 timing. This is the SPD default for Dominator memory.

[zandrax]

Re: High Failure Rates for DDR2

@ bgavin:

Overvolting (and related overheat) usually kills ram sticks in the middle run (9-15 months) since speeds up electromigration: if overvolting even more can recover stability in the early stage, then electromigration is likely to be involved and chips are progressively damaged until they'll crash at any voltage.
I follows acstech's advice: keep memory at stock voltage or, if you need some jouice, don't overvolt more than 5% (so 1.9 V for standard DDR3). Play with timings, not with vdimm.

Zandrax

[acstech]

Re: High Failure Rates for DDR2

(so 1.9 V for standard DDR3)

You meant DDR2, right?

[gdement]

Re: High Failure Rates for DDR2

I guess my memory of my memory is slightly different than some peoples...

I'm not sure where CPU-Z is getting that voltage info from.
I'm pretty sure there's nothing in the SPD for DDR1 that says anything about a module's "intended" operating voltage. The motherboard just has to assume JEDEC values.
They might have added that with DDR2 though, haven't looked at that spec.

[bgavin]

Re: High Failure Rates for DDR2

I follows acstech's advice: keep memory at stock voltage or, if you need some jouice, don't overvolt more than 5% (so 1.9 V for standard DDR3). Play with timings, not with vdimm.

I'm not an overclocker at all.

My problem was following the manufacturer's recommendation, which I now think is bogus. I'm convinced all are so hungry for sales, they are willing to accept a 10% failure rate and RMA costs as part of doing business.

I think your explanation of electromigration is exactly correct.
The failure points in time for 2.1v DDR2 match your description.

I don't use DDR3 at all.
DDR2 spec is 1.8v

[Per Hansson]

Re: High Failure Rates for DDR2

Yes, DDR2 and DDR3 memory is very sensitive to voltage
For example the GDDR3 graphics memory on several nVidia 8800GTS 512MB cards actually benefits _under_volting the memory when overclocking
This is a first in this for me... Undervolting to gain higher clocks, and yes
That is undervolting lower than what the memory manufacturer says, not whatever the guys putting the memory on a PCB have to say about it...

I've had bad sticks of DDR2 from the factory, not seen much memory fail (without reason) after prolonged usage tho...
EDIT; As for overvolting I have seen countless reports of memory being physically damaged by high voltage (only referencing DDR2 & 3)
I'm running my current 4x2GB sticks at 1.97v (manufacturer recommends 2.1v)

[acstech]

Re: High Failure Rates for DDR2

I've seen that too, Per. In my overclocking experiments, I've seen memory that lost stability over about 1.9v. Meaning, it was actually less stable at high speeds at higher voltages. Most of the DDR2 I get recommends relatively low voltage though, 1.8-1.9v.

Edit: Oh, it seems we're not the only ones with problems:

http://www.ocforums.com/showthread.php?t=569986

[NxB]

Re: High Failure Rates for DDR2

Whenever I OC, I try to keep the memory at stock speeds/voltages. Only thing I've gotten is data corruption and instability from overclocking memory. My ddr runs at 208mhz and I'd rather it run stock but the math doesn't work out.

and my memory shows 2.5v in CPUz as well. Both kingston and samsung.

[Per Hansson]

Re: High Failure Rates for DDR2

NxB; Yea, I'm the same as you when it comes to system RAM
I don't overclock it past manufacturers recommendations

That put a bit of a snag in my overclocking endevours with my Intel Q9450 that I now run at 400x8=3200Mhz
Any higher would result in running my memory above spec, i.e. "PC2-6400" spec... (400Mhz or 800Mhz DDR whatever they call it these days )