Micron D9 and high voltage, electromigration effects

[ozzimark]

'jitter-free' - 'lower ripple', a lower ripple would in theory prolong the life of on board capacitors and to a smallish extent the ic's

why a smallish extent? after all, each DRAM bit is a transistor and a capacitor.

[Tekdemon]

Dammit, I was all set to go buy some D9 (Corsair, since it's fairly cheap now) for my next build, but after all these horror stories I'm trying to figure out what I should do.

Even if I had a board that let me adjust drive (I'm planning on getting the new DFI 965-S so I should have it) I think it's still make me paranoid that the RAM would go suicidal one day, and while I love to push the limits I also enjoy having a system that doesn't kill itself within a couple of months. I mean, so far we still have no idea exactly how much current is killing the RAM, and it seems like a huge # of motherboards are capable of doing it, so who knows if you can even lower the drive enough to keep the RAM alive long-term?

[Revv23]

it has a lifetime warranty, buy it and dont overvolt it.

[Slammin]

it has a lifetime warranty, buy it and dont overvolt it.

Thing is, you don't have to even come close to over-volting these chips to kill them.

Lifetime warranty or not, RMA is a pita.

It seems we are all in some kind of transitional period. I mean, these mobo's are set to default at 1.8v vdimm, yet the ram are rated for 2.1-2.4v. Not a good match and just plain problematic.

I'm pretty confident, and actually was at the time, that the two sticks I sent back were not defective, even though they did 'change' in that they both would post at 1.8v when I first got them, but later on, as long as I gave them enough vdimm (2.3v) they could run at spec w/o errors, but they would no longer post at 1.8v.

Oh well, I guess I could have waited for the dust to settle some, but I really thought I did. The C2D parade started several months ago! I guess I could have gone for a non-680i mobo though too.

Really, I don't think this stuff is quite ready for prime-time right now. I mean, having to remove all my ram and find a stick/slot combination that will post at 1.8v is a serious pain. I guess I could buy an emergency (spare) 'virgin' stick that's never been "abused?" by attempting to run it at mfg spec so that I can recover from a really bad OC attempt. All my sticks originally would post at 1.8v, but only one or two will now.

Oh well. At least I am now familiar enough with this setup that I have a pretty good idea of what NOT to try, and this sucks because I may never know what this rig is capable of since I'm too lazy to play the 'which stick can post in which slot' game everytime I get too aggressive looking for my max OC.

[Slammin]

Well,

One thing I can say for this ram is that it is versatile! Right now I'm running 335MHz, 3-3-3-6-16 1T and 2.0v. When running 500MHz 4-4-4-4 24 2T, I cannot even post with less than 2.2v.

Memory is running cooler now, SuperPI is slightly faster, showing 52ms latency in everest. As a side note, and for those with this mobo and quad, I am able to bench with 360 fsb when running ram 1:1 (linked/synced) but whenever I use any ratio at all it seems I'm stuck at 333fsb.

So, I've learned a few things with this ram:

Very versatile - can run low fsb with very tight timings w/low vcore, can run pretty high fsb with reasonably tight timings w/moderately high vcore, can run very high fsb with loose timings but too much vcore. The side discovery that ram divider was limiting how high I could crank up my fsb was just icing on the cake :-)

Side note: Letting this ram, and your mobo cool for awhile after a failed OC attempt may save you from having to pull all the ram so that you can get it to boot at CMOS default vdimm. My ram gets too hot and the system won't post unless I let things cool down. Of course I could reset CMOS and pull all the sticks and find a stick/slot combo that will post with the default 1.8 vdimm, but that is a PITA! Just letting things cool down works everytime for me.

[ozzimark]

Side note: Letting this ram, and your mobo cool for awhile after a failed OC attempt may save you from having to pull all the ram so that you can get it to boot at CMOS default vdimm. My ram gets too hot and the system won't post unless I let things cool down.

if your ram is getting that hot, no wonder you have instability. anything warmer than warm to the touch is too hot in my eyes

[Slammin]

I bought that Corsair ram cooler while waiting for my RMA'd ram to come back, but even with it, plus a 120m side case fan blowing in, the ram gets pretty warm above 2.3v and 500MHz.

I guess that is one huge disadvantage of using 4 sticks to go to 4GB. They are so close together that cooling is even more of a problem.

I'm good to go now though. Just definitely a learning experience with this new mobo and ram.

[Gam3Ra]

10x a lot for this thread! This explain me why my rams are burning in P5B-D with 2.25-2.3V with cooling and the chips are not hot.

But why you speak only for Micron, migration ot electrons and all things are working with Elpida/Promos/Powerchips (and all others), no?

[cadaveca]

leakage is less onthese chips..seemingly a factor that allows them to properly use higher current, as not as much is lost in heat. This in turn requires lower drivestrength, which is what really kills DIMMs.

[dinos22]

hmmm i guess you learn something every day

good read and posting by all

[adfinni]

very interesting reading, nice !!!

About to RMA my team ram which won't run at stock anymore. Im on a p5b-d and my overclock with my E6300 was 500x7 (3.5ghz) and the ram was at 4-4-4-8.

When i get a replacement set will the same thing happen again (~6 months use) with the same overclock, then again, then again. Can anything be done to actually let the RAM live, as not all of us try and run high voltage through our ram, and just want a fast 24/7 rig?

My ram was used 24/7 at 2.2v with an 800mm fan over them since day one.

[Buckster]

I mentionned this problem to a guy at work -he thought whether it could be related to the capactitor type nature of the cells.

Is it possible that they don't fully discharge (something to do with the high Vdimm + the doubly insulated nature of DDR2 cell makeup) - and then you need higher voltage to get same result from within the charge space remaining.

He thought electro-migration seemed unlikely as people have said that if they leave the memory for months on end after failing - it re-works again - electron migration would not work this way.

Adfinni - 2.2V you sure ? (my P5-B Deluxe only does 2.15 and 2.25V ?)

?

Mark.

[HKPolice]

Any way to change the drive strength with a prog like memset? No such options for P5B?

[n-sanity]

For the record (iow for the not knowing), voltage regulation refers to the regulation, or stabilization of voltage supplied. The motherboard doesn't actually supply the ram with a pure steady 2.3v. The actual voltage jitters around and may range from 2.2-2.4 or in bad cases (cheap pos mobos) way more. If your ram is rated @ 2.3-2.5v and you used 2.4v, at which point your ram died - your mobo may not be up to the task: the voltage that really killed your ram may be as high as 2.6-7 v.

Other factors that increase as the quality of your mobo decreases may be current and (don't quote me directly on this one) induction of sorts; Also - if you have a high power water pump or fan right next to your mobo... or a 110v compressed of course , the field it creates introduces bigger ripples in the electronics all over your motherboard, even though it is somewhat shielded.

Rule of thumb - keep things away from your mobo, preferably keep it shielded in case and dont buy pos mobos/pos psu's (<<< important - primary source of the ripple in your voltage).

if your ram is getting that hot, no wonder you have instability. anything warmer than warm to the touch is too hot in my eyes

The melting point of silicon is 1414 &#176;C (2577 &#176;F)... @ 45&#176; all your base are not belong to me, trust me Of course different chips handle heat differently... but seeing as how the chips ARE sold to africa, i can't imagine them failing anything below OWWTFHOT to the finger.

[zsamz_]

i just woke up after a pc dream so bear with me
i been doing some thinking and i wonder how many of us have so many things plugged on same electrical line "not same plug"
i got 4 computers in the house and they all on same line
imagine you using your pc n someone starts a vacum or a pc when at startup lots of electricity is drawn so after initial draw when its released you should get a spike
i think the best solution would be to install a seperate line for these highpower pcs
i think this might be bigest cause of dead microns

[GFORCE100]

OK guys but no one here seems to mention how Drive strength, called AI Booster in the newer Asus boards improve memory bandwidth/latency. Turning it off takes a few 100's MB/s throughput away.

So it is just about increasing the amps?

[Rock&Roll]

Yes I'm curious about the Asus Booster setting as well.

In my BIOS it's called "Transaction Booster". It's not really fully explained in the manual or anywhere. Am I right to see this as being the "Drive" setting?

Just for testing, I disabled this feature and set the value to 3. I had to bump up ram to 2v to sabilize. But while priming, my ram is running a little cooler than 1.8v with the booster enabled. (At least as well as I can tell by feeling with my fingers)

[STEvil]

AI Booster is not drive strength

AI Booster changes timings and sometimes dividers.

[RatusNatus]

3 kills on my P5B dlx

Crucial Ballistix 6400 - 2.25v
Crucial Tracer 6400 -2.25v
G.Skill 6400 HZ - 2.2v

I have a killer mobo

I have 1 crucial RMA to go...

Adfinni - 2.2V you sure ? (my P5-B Deluxe only does 2.15 and 2.25V ?)

?

Mark.

The AUTO value is the 2.2v.

[saaya]

geez, what board and psu are you using?
and did you meassure real vdimm?
is it really only 2.25v?

[zsamz_]

geez, what board and psu are you using?
and did you meassure real vdimm?
is it really only 2.25v?

some boards are just killers saya lol
i had 3 gigabyte p35c-ds3r they were all ram killers 2.24-2.26v
i lost 8 stix thanks to gigabut never went more than 2 months per set
i been on dfi almost 5 months 2.1--2.27v not 1 dead stick same psu same everything

[saaya]

yeah, ive heard of that "special edition" gigabyte boards that give +0.2 to +0.3 vdimm you set 2.2v in bios and get 2.5v real...

Besides those gigabyte boards the asus and intel 975x boards and nvidia and asus 680 boards seem to kill a lot of ram. im doing the rmas for cellshock and its hard not to notice that those boards keep showing up way way more than they should, considering that rather few people use those boards.

[zsamz_]

yeah, ive heard of that "special edition" gigabyte boards that give +0.2 to +0.3 vdimm you set 2.2v in bios and get 2.5v real...

Besides those gigabyte boards the asus and intel 975x boards and nvidia and asus 680 boards seem to kill a lot of ram. im doing the rmas for cellshock and its hard not to notice that those boards keep showing up way way more than they should, considering that rather few people use those boards.

+4 you get 2.34-2.38 lol n most think its 2.2v on auto you get 2.38 lol
i tried for 6 months talkin to idiots @ gigabyte but lol i just gave up
when i said 2.24-2.26v i measured with dmm

[kiwi]

Don't use ddr2 D9xxx 24/7 lol

[Jaco]

My Gigabyte P35-DS4 overvolts too.
2.2v in bios gives 2.33vdimm real . That's too much for 24/7 operation, imo.