safe micron DDR3 voltage

[Gautam]

Reference 790i. It might be the board being retarded all things considered though. It's done 1080 8-7-6 for a month now easy almost every day, with voltage no higher than that. Then yesterday I felt like trying a 32M, failed on the first loop even down to 1026, and poof, ram seems busted.

I'm crossing my fingers that once I drop it in a P35 it'll be back to normal, but somehow I'm not hopeful considering my luck.

[dinos22]

have you backed down with volts and lower clocks?

790i boards can be funny beasts so it could be something else as well
fingers crossed they are ok
pretty sure you have a nice set of sticks

[Gautam]

Nah, I was too irritated so I just went to bed. This morning for some reason 1070 seemed okay for 3D once, but it had to be unlinked with some goofy divider like 14:30. Then after that it started acting up again, and having trouble POSTing. It used to be fine sync at 540 FSB, so either the board or ram are going south. :S Problem is that only 790i is capable of clocking the ram that high so it might hard for me to verify for sure what's going on...


Hopefully it's a false alarm. (I'd probably actually prefer a dead board lol) The suddenness is what's weirdest about this. Not gradual at all, working fine for a while and then a sudden dramatic fall in speeds.

[saaya]

i wouldnt be surprised if 790 has a higher drive strength than intel chipsets, same as with ddr2, and as a result kill memory at lower voltages?
or maybe the vdimm regulation on 790 boards is just as unstable as vcore?

did you check if both sticks are affected?
did they ever run 100% stable at the speeds you ran them at?
what spp voltage are you running?
maybe you damaged the memory controller and not the mem?
and a setting working one day and not working another... well.. thats nvidia chipsets for ya

[Gautam]

It is busted for sure. Tried it in P35 (Blitz Extreme) and it cannot even do 950 CAS 7.

Dunno about the individual sticks, good idea.

If it's 790i, then how does 3oh6 get away with running 32M's all day every day at 2.15-2.2v?

This is beyond explanation, I don't think it's even the voltage that did it, but there's no sensible cause I can pin on this whatsoever.

Either way, the moral of the story for me is, it doesn't pay to be a pansy. If DDR3 can die at 2.1v, it will die no matter what if you oc. My next set is getting 2.4v straight up.

[kiwi]

Just saw that Adata has Micron based DDR3 kits that are warranted up to 2.15V. So that is supposed to be safe

[3oh6]

Either way, the moral of the story for me is, it doesn't pay to be a pansy. If DDR3 can die at 2.1v, it will die no matter what if you oc. My next set is getting 2.4v straight up.

haha, hell yeah man! i too have been like you and going easy on the volts but now...who cares. if they are going to die, let them die. i still won't be throwing 2.4v into them for anything other than a one-off record run but i am certainly not scared with 32M benching all day at 2.15~2.20v.

and yes, i can confirm i have been running 32m/dual 32m/3d all day pretty much everyday for a couple months now at 2.15v~2.20v with a couple different kits and never had an issue on a reference 790i. the Ballistix kit doesn't like POSTing at high clocks regardless of voltage...but it never did the day i got it

i might be moving over to the Maximizer on the 790i though to see if the voltage regulation on that board is as bad as i think it is. i just have to get off my but and do the soldering.

[Gautam]

I killed two more sticks a couple of days ago. Ballistix PC3 12800 this time, also at 2.1v, in very much the same fashion. It isn't voltage that's doing it. I have other sticks that continue to run strong, even on the same board.

The common thread I found between all the dead sticks is that I was trying to push them to speeds they couldn't handle. e.g., I immediately tried 1100 8-7-6 right off the bat. They failed to enter Windows a few times, then Windows got corrupted, I come back a couple of days later and one stick won't even POST, the other won't even make it into windows at 1000 CAS 8. This is more or less how all four sticks that I killed, Kingston and Crucial alike, died. Maybe you'll think I'm crazy for suggesting this, but it seems like the cause of death for all of them was instability. :S

[dinos22]

far out man
more dead sticks

i too did a lot of big jumps like that and even used max volts in bios >> 2.275v i think for some sessions but the sticks are still running ok

maybe your board had messed up RAM circuitry or DRAM VTT is waaaaaaaaay out of spec or something

[Gautam]

I'd think the same, but I still have several sticks going strong well into the high 1000's CAS 8 with the same voltage on the same board, even mustered up the courage to do some 32M on em. A single stick breaks 1100, now I just have to find another one to match it. (Trembling yet, Jody? )

The ironic thing is that I have a gut feeling that more voltage might have actually saved the others. O.o That way they'd have been more stable. I think the only moral of the story is that assuming that a fresh untested set of ram can do 1100MHz at 2.1v is a bad idea. I don't know how instability can kill ram. I remember a few years ago that people were killing Radeon X800's by running the raw clocks of their memory too high. This might be something similar, but you're probably right in that it has something to do with the board. :S

[M.Beier]

Now we're getting somewhere
I cant say I dissagree on this, I still belive that vdimm isnt what to be blaimed...
The question is just, what does cause the killing...

As I told Gautam some hours ago (I probably didnt make much sence, was LATE night in Europe)
I still havent managed to kill any, and I've not been very kind to the ram.

[Infa]

Back again... sorry havent answered you saya easy to forget these PM's.

now i think they have degraded even more, havent used my comp for 2 weeks~ and now after 2 days i decided to play again..
and crysis crashed AGAIN... sure that game crashes alot, and have nothing to do with hardware, but sometimes it does..
and now it did, like before when i had to low vdimm... AND now it doesnt even pass first irretation of 32m anymore. As it did before, and now im at 2.14 at 934mhz Cas7
think thats to high, and now it want MORE , i really think they are degrading, and not the board beeing diffrent...hmmm.

[Infa]

Ok, think i made a Noob misstake... i was using to LOW vMCH!
omg , how can i not think of this

now it passed first 2 irretations without hickup....hmmmmm...

was keeping it at 1.51 for 1869mhz cas 7, with PL6!! and twister at stronger...

maybe i was to synical and thought it would work with 1.51 :P

will test more...maybe all the time it was the NB voltage :S

EDIT: failed...will try with more nb.
Edit2: 1.51 will fail 32m at first initial loop, 1.57 are better but will fail after loop 1, 1.61 will fail initial, 1.55, pushed to loop 7.
Edit3: PL7 and all channels pulled in, and 1.55vMCH passed 15 Loops, and then it hanged (ala Vista) so i take it as it worked, still same 2.1v at Vdimm, so it may not be the RAM afterall...hehe..quite tricky though...
Edit4: omg hehe.. they are a good way to learn these board.
now i tightened the ram timings even more, 7-6-5-16 , trfc 48! at 1869mhz PL6 fsb 467! keeping vMCH at 1.55, and i tried even to lower the vdimm, and imagine, i stopped it at Loop 9 myself... STRANGE, now im at 2.08vdimm.
i learn stuff everyday hehe. will comeback with some complete tests later.

regards.

maaaaaan these boards are weird.

[saaya]

gautam, the x800 mem corruption thing killed the gpus imc, not the memory from what ive been told. i have a feeling that if you feed the chips more volts than they can "process" they start to create errors, the more vdimm, the more errors. and if you continue running the mem that way it seems to degrade or kill the memory. thats my gut feeling about micron ddr3...

infa, yes, vmch is very important to get high mem clocks, from what ive seen.
more vmch will let you reach the same mem clocks with less vdimm in some cases, funny eh? usually intel nbs seem to scale up to 1.7v above that your barely get any higher memory clocks. but if you run tight PL then more chipset voltage than 1.7v actually does scale... thats why some crazy people like hipro went up to 2v and above and run PL5 and PL4

[hipro5]

I just lost by feeding ~ 2.35Vdimm:

a. One of my CellShock sticks
b. My retail Foxconn BlackOps
c. My ES QX9650 CPU (working at 3.6GHz with 1.25VCore)....

I honestly don't know what happent!....
My CPU on my other mobo now boots up and stops at "C3" code.....

Though the other stick is OK......

[dinos22]

maybe it was just a bad board that "took" them considering you lost a CPU as well donno

[saaya]

sounds like it... mobo or psu problem id say :/

[Stelios]

That's bad George ...

Did esdee drop by your place recenlty ??

[Infa]

Hmm reconfigured my bios settings, now they only need 2.06 for 1870mhz 7-6-5-16 TRFC 48!!

looks like ME was the mistake, and ddr2000 booted and loaded win at under 2.20 ... seems like no degradation then... just me degrading in knowledge :P

[saaya]

i dont blame ya bro, ddr3 boards are tricky to tame

[5010]

Setup the rig in my sig at 790i Launch day.... Been running the speeds in my sig since then, yesterday was VERY HOT and humid and lo-and-behold my RAM has degraded to the point of running 1777 8-8-8-24-2T instead of 7-7-7-21-1T like Ive had it since I bought it.

2.0V set in Bios 24/7 since March '08 (about 4 months)

[saaya]

hmmmmm so... all the sudden you can only run 888 2t and no longer 777 1t?
thats... weird...

whats the max 777 1t you can run now with the same voltage?
are you sure its not the chipset degrading or maybe some BIOS issue?

[5010]

hmmmmm so... all the sudden you can only run 888 2t and no longer 777 1t?
thats... weird...

whats the max 777 1t you can run now with the same voltage?
are you sure its not the chipset degrading or maybe some BIOS issue?

Haven't bothered to check highest 7-7-7 1T but I'm 99% sure its not the chipset since I'm only running 1.35V SPP.

I've also dropped my RAM voltage to 1.95V Bios so as hopefully not KILL my RAM completely while I shop for a replacement / RMA to Patriot.

My guess is the degredation is only very slight, still I thought worth reporting.

[saaya]

well 777 1t to 888 2t is quite a big degradation imo!
thats the difference between high end memory and value memory that costs much less

[Punisher!]

Are Micron the only known DDR3 chips suffering of degradation?