safe micron DDR3 voltage

[massman]

Saaya, as you're working for a memory manufacturer, is there no way for you to test this instead of relying on the reports of xtreme overclockers?

Make a quick set-up in your lab: 5 setups all equiped with the same hardware/software and make sure all memory kits are from the same batch. First setup with 1.9, second with 2.1, third with 2.3, fourth with 2.5 and fifth with 2.7 and let them run at non-maxed-out overclocked settings 24/7. Now time how long the setups last and report back. I'm quite sure you'd do the community a big favour with it.

[saaya]

My benching sessions last ~8 - 10 hours each.....
I run my CellShocks PC14400 at 2.31Volts in every benching session and I don't go higher OR lower that this since I got them (VERY FIRST BATCH - FIRST DDR3 Ram modules - when DDR3 show up)......

thanks for posting george
you forgot to mention that your using a maximizer
i dont know how stable and clean the voltage of the different boards are, and more importantly, if you set 2.31 its 2.31 and not 2.35 or 2.38

i havent heard of any memory dieing or beeing damaged at 2.3v, as long as its cooled well enough. the 2 sticks i damaged were with less than 2.3v, but they were not cooled and running very hot.

so if the memory is cooled well even 2.3v seems to be safe even for extended periods of time. Id recommend everybody to check how much your board overvolts though, some boards overvolt by 0.05 and more.

again, this is voiding the warranty of cellshock memory and most other brands. if you decide to void the warranty of your memory and use more than what the manufacturer specified then you do this at your own risk, yadda yadda yadda

[saaya]

Make a quick set-up in your lab: 5 setups all equiped with the same hardware/software and make sure all memory kits are from the same batch. First setup with 1.9, second with 2.1, third with 2.3, fourth with 2.5 and fifth with 2.7 and let them run at non-maxed-out overclocked settings 24/7. Now time how long the setups last and report back. I'm quite sure you'd do the community a big favour with it.

1. like i already posted, ive done something similar already and will do it again if i have time. without cooling already 2.1v can degrade chips, and 2.25 can damage chips severly. as i said, i think this was caused by the temps, not by the voltages.
2. the only thing this proves is that some chips of x batch on board y with z volts get damaged/killed. to get a more reliable impression you need more results from different setups. thats why i made this thread.

[massman]

1. like i already posted, ive done something similar already and will do it again if i have time. without cooling already 2.1v can degrade chips, and 2.25 can damage chips severly. as i said, i think this was caused by the temps, not by the voltages.
2. the only thing this proves is that some chips of x batch on board y with z volts get damaged/killed. to get a more reliable impression you need more results from different setups. thats why i made this thread.

1. Different environments = different results. Keep the temperatures in your tests constant and you'll know if temperature caused the degradation.

2. Different memory on different boards in different environments using different cooling solutions being stressed with different tools = extreme increase of the number of variables and the data will not be reliable enough to be compared, let alone that you can draw conclusions out of it. To test the inlfuence of one variable, you need to keep as much variables as possible constant.

[saaya]

do you honestly think i dont know that?

[kiwi]

But you can't know what will happen after half year, 1 or 2 years. If some voltage seems safe for a few weeks or months it doesn't mean it will be safe for longer period because degree of degradation (or death coming) might be much slower. The longer you intend to keep your memory the less "safe" volts run it

[fredyama]

In my case, IIRC...
I was too lazy to meaure the actual Vdimm when killed the both sticks

M/B: Maximus Extreme
Vdimm: 2.34V in BIOS -- 0.036V to 0.038V overvoltage (measured at leg of MOSFET: Bios 1.90V --> 1.938V, Bios 2.20V --> 2.236V)
LN2 benching: about 4-5 hours
Mem cooling: 2 x 8cm fan

M/B: P5E3 WS Pro
Vdimm: 2.34V in BIOS? -- 0.095V to 0.098V overvoltage (measured at leg of MOSFET: Bios 1.88V --> 1.975V, Bios 1.90V --> 1.997V, Bios 2.00V -- 2.098V)
Air cooling: Scythe Infinity + Delta TFB1212GHE x 1
Mem cooling: 2 x 8cm fan
Booted up to OS was ok, crashed/killed immediately after running 45GP (i.e. SPi8M@4.5GHz)

[M.Beier]

Thats quite a bit overvoltaging on vdimm... :\
Still sounds odd that you lost at that voltage..
But 2.43v is quite high vdimm for DDR3, I'll confess

Do you remember your bios ram settings Fredyama? Im very currious if that might have had an impact...

[fredyama]

IIRC, MEM settings in Bios were: 1:2, 1T, 7-6-6-15-4-48-7-4 Auto for the rest & FSB490@boot.

Thats quite a bit overvoltaging on vdimm... :\
Still sounds odd that you lost at that voltage..
But 2.43v is quite high vdimm for DDR3, I'll confess

Do you remember your bios ram settings Fredyama? Im very currious if that might have had an impact...

[KTE]

Regardless of slight variations we can still notice "trends" if people give such feedback.
For instance, the threads where above 2.3V DDR2 was mentioned as being unsafe for a normal desktop 24/7 with a year longevity in mind, actually did help us, even though many killed those Microns at 2.1V (me too) but most didn't and were safe till around 2.3V real when cooled although no more. I would then stick to advising that much maximum for it and not anymore even though I've rallied 3V normal air into them for over 3 hours once and above 2.3V regularly for benching, I would never advise it. The trend would be similar in DDR3 although different volts.

[saaya]

But you can't know what will happen after half year, 1 or 2 years. If some voltage seems safe for a few weeks or months it doesn't mean it will be safe for longer period because degree of degradation (or death coming) might be much slower. The longer you intend to keep your memory the less "safe" volts run it

yeah, but if we know at what voltage memory died after a few weeks or months or immediatly thats at least something
and it will allow us to guess pretty well what voltages are safe and what voltages arent.

thats why i made this thread, i did my own testing and received quite some feedback about what voltages killed memory which arent publicly accessible, so i figured id give you guys a headsup and share my knowledge so that everybody at least has a rough idea where the voltage "mindfield" begins

M/B: Maximus Extreme
Vdimm: 2.34V in BIOS -- 0.036V to 0.038V overvoltage (measured at leg of MOSFET: Bios 1.90V --> 1.938V, Bios 2.20V --> 2.236V)
LN2 benching: about 4-5 hours
Mem cooling: 2 x 8cm fan

thank you very much for sharing this information here fredyama

Thats quite a bit overvoltaging on vdimm... :\
Still sounds odd that you lost at that voltage..
But 2.43v is quite high vdimm for DDR3, I'll confess

it wasnt 2.43
2.34v in bios + 0.037 overvoltage = 2.377v, so roughly 2.38v

and, correct me if im wrong fredyama, but the sticks were not degrading, both sticks just suddenly stopped working completely.

from what ive seen and also heard as feedback, vdimm fluctuates the most during bootup and shutdown on most boards. i assume that something, be it the memory or whatnot, triggered a system reboot. this somehow resulted in a big fluctuation of vdimm or maybe only vtt or vrefdq (ref voltage like vtt), which then killed both sticks at the same time.

either that, or the memory died because of a voltage spike or some other issue, which then resulted in the reboot...
but a voltage spike has to be quite big to kill both sticks at the same time... damage them, ok... but so badly that none of the sticks boots anymore... at the same fraction of a second? but then again, this is how it happened with most people who lost ddr3, from one second to the next it just didnt boot anymore.

Regardless of slight variations we can still notice "trends" if people give such feedback.
For instance, the threads where above 2.3V DDR2 was mentioned as being unsafe for a normal desktop 24/7 with a year longevity in mind, actually did help us, even though many killed those Microns at 2.1V (me too) but most didn't and were safe till around 2.3V real when cooled although no more. I would then stick to advising that much maximum for it and not anymore even though I've rallied 3V normal air into them for over 3 hours once and above 2.3V regularly for benching, I would never advise it. The trend would be similar in DDR3 although different volts.

exactly... as long as you get feedback from people telling you something just blew up, you get at least a rough picture where the midfield starts.

yes, there are people who dont care about the money a respawn costs them and run around in the midfield without a care in the world, but for those people who dont want to risk anything its good to know where the mindfield begins.

[massman]

do you honestly think i dont know that?

Lose the ego, please. If I thought you knew it, I wouldn't have said it. 5 results found the way I said will be more reliable than the results of 100 unique systems/settings.

it wasnt 2.43
2.34v in bios + 0.037 overvoltage = 2.377v, so roughly 2.38v

and, correct me if im wrong fredyama, but the sticks were not degrading, both sticks just suddenly stopped working completely.

If you read Fredyama's post correctly, you'd notice how he ran his memory at 2,38v perfectly and killed them at 2.34V + 0.095V/0.098V = 2,44v, which is quite a bit higher than 2,38.

Regardless of slight variations we can still notice "trends" if people give such feedback.
For instance, the threads where above 2.3V DDR2 was mentioned as being unsafe for a normal desktop 24/7 with a year longevity in mind, actually did help us, even though many killed those Microns at 2.1V (me too) but most didn't and were safe till around 2.3V real when cooled although no more. I would then stick to advising that much maximum for it and not anymore even though I've rallied 3V normal air into them for over 3 hours once and above 2.3V regularly for benching, I would never advise it. The trend would be similar in DDR3 although different volts.

Yes, of course, when using the stick 24/7, no one should ever recommend a high overvoltage (+ 0,3/0,4 at most). But here most of us care about voltage safe to bench at. The voltage is a bit higher, the stress a bit higher, but the system runs less long. I certainly wouldn't mind killing my bench memory after 1y, but I would mind killing my 24/7 rig memory after 1y.

[M.Beier]

Fredyama ->
Hmm, well the one I suspect is drive strengh... Back with DDR2 it had a major impact, and to be honest, I dont know exactly what it did, just that it wasnt worth the squize to tighten it... Too sever risk

yeah, but if we know at what voltage memory died after a few weeks or months or immediatly thats at least something
and it will allow us to guess pretty well what voltages are safe and what voltages arent.

thats why i made this thread, i did my own testing and received quite some feedback about what voltages killed memory which arent publicly accessible, so i figured id give you guys a headsup and share my knowledge so that everybody at least has a rough idea where the voltage "mindfield" begins


thank you very much for sharing this information here fredyama


it wasnt 2.43
2.34v in bios + 0.037 overvoltage = 2.377v, so roughly 2.38v

and, correct me if im wrong fredyama, but the sticks were not degrading, both sticks just suddenly stopped working completely.

from what ive seen and also heard as feedback, vdimm fluctuates the most during bootup and shutdown on most boards. i assume that something, be it the memory or whatnot, triggered a system reboot. this somehow resulted in a big fluctuation of vdimm or maybe only vtt or vrefdq (ref voltage like vtt), which then killed both sticks at the same time.

either that, or the memory died because of a voltage spike or some other issue, which then resulted in the reboot...
but a voltage spike has to be quite big to kill both sticks at the same time... damage them, ok... but so badly that none of the sticks boots anymore... at the same fraction of a second? but then again, this is how it happened with most people who lost ddr3, from one second to the next it just didnt boot anymore.


exactly... as long as you get feedback from people telling you something just blew up, you get at least a rough picture where the midfield starts.

yes, there are people who dont care about the money a respawn costs them and run around in the midfield without a care in the world, but for those people who dont want to risk anything its good to know where the mindfield begins.

2.43-2.44v I belive.. : - )
And yes, it flucrates most during boot, and that has costed me sticks
DDR2... - 3.47v + 0.12v is a no go for most sticks

[kiwi]

This has already been mentioned but also note that many Asus boards overvolt vdimm a lot, up to ~0.1V or even more depending on your board

[saaya]

Lose the ego, please. If I thought you knew it, I wouldn't have said it. 5 results found the way I said will be more reliable than the results of 100 unique systems/settings.

what do you mean with ego? 1. i already posted in this thread that i know this, 2. its just common sense, i work for a memory manufacturer and you think i dont know that results from diferent systems cant be directly compared? :P how do you think we and ocz and stt and any memory vendor bins their memory?

and again, every chip acts diferent, so even if id burn a few dozen sticks id still only be able to guess, and id only be able to guess for that batch of chips i burned a few from...
and even then, it might be related to that mainboard, and another one has cleaner more safe vdimm so those chips would have burned on slightly higher vdimm on another board...

If you read Fredyama's post correctly, you'd notice how he ran his memory at 2,38v perfectly and killed them at 2.34V + 0.095V/0.098V = 2,44v, which is quite a bit higher than 2,38.

my bad, i overlooked the part about the ws pro
daym, almost 0.1v olvervolting... thats not cool...

Yes, of course, when using the stick 24/7, no one should ever recommend a high overvoltage (+ 0,3/0,4 at most). But here most of us care about voltage safe to bench at. The voltage is a bit higher, the stress a bit higher, but the system runs less long. I certainly wouldn't mind killing my bench memory after 1y, but I would mind killing my 24/7 rig memory after 1y.

really? you dont upgrade your mem for 1 year?
for me killing the mem after 6 months would already be acceptable, id still rather run voltages that are safe to last 1 year+ tho, simply for avoiding the stress and hassle it causes to suddenly have to fix the system or replace the memory... and yeah, the idea here is for all of us to share our knowledge to make an half educated guess on what voltage is how safe for benching and 24/7.

2.43-2.44v I belive.. : - )
And yes, it flucrates most during boot, and that has costed me sticks

my bad

[saaya]

m.beier dinos22 and massman, sorry in case i sounded a bit rude, i am a bit dissapointed since i got the impression that you, instead of beeing happy about the infos i posted here, only ask for more and only try to disprove things instead of trying to work on something that would benefit all of us. if thats not your intetion then i hope you accept my apology.

i updated the first post and changed the range, 2,2v seems to be safe as well, but for 24/7 opteration over 1-2 years im not 100% sure. there might be some sticks that do fail at 2.2v after a long period of time.

about benching, 2.3v seems to be safe, even 2.35v, but above that the risk of the memory being damaged increases notably. what is your guys experience with 2.4v? has any of you run 2.4v for an extended period of time?
on what board and with what cooling?

maybe it would be useful to add a list of how mainboards overvolt vdimm.
lowest and highest reported values

P5K3 dlx 0.030 - 0.055
P5E3 non dlx 0.025 - 0.055
P5E3 dlx 0.050 - 0.120v
blitz xtreme 0.070
maximus xtreme 0.036V to 0.038
P5E3 WS Pro 0.095V to 0.098

[massman]

Naah, I'm quite happy you made this thread, really. Haven't bought me DDR3 yet (friggin' expensive ), but frankly if I ever buy ddr3, I want to know the relative safe values.

I was just wondering if you couldn't do extra tests on the subject as you're working for a memory manuf. I'd do it myself if I had too much money .

[kiwi]

Dont worry, this information is useful

Maybe it is also a good idea to add "safe" chipset voltages since at high memory speeds chipsets also need increased volts. Has anyone done research on this? Asus something mentions that 1.75V is fine with stock NB cooling on x38 chipset


You can add these mobos that overvolt vdimm:
p5e3 dlx: +0.08V
blitz extreme: +0.07V

[BenchZowner]

For P35 I'd say 1.75V, and X38/X48 1.8V
That's for totally 24/7 safe voltages.

I'd say 1.8V & 1.85V respectively, are also safe, but I haven't done any extended testing to verify my gu-estimates.

Edit: With active cooling...or at least a giant heatsink to keep them under 55C @ load

[George_o/c]

For P35 I'd say 1.75V, and X38/X48 1.8V
That's for totally 24/7 safe voltages.

But isn't that a bit low ? ...

[BenchZowner]

But isn't that a bit low ? ...

I'm talking about MCH ( NorthBridge ) voltages m8 [ not memory ]

[massman]

A bit high, to be honnest. I'm only running 1,7+ when benching.

[M.Beier]

m.beier dinos22 and massman, sorry in case i sounded a bit rude, i am a bit dissapointed since i got the impression that you, instead of beeing happy about the infos i posted here, only ask for more and only try to disprove things instead of trying to work on something that would benefit all of us. if thats not your intetion then i hope you accept my apology.

i updated the first post and changed the range, 2,2v seems to be safe as well, but for 24/7 opteration over 1-2 years im not 100% sure. there might be some sticks that do fail at 2.2v after a long period of time.

about benching, 2.3v seems to be safe, even 2.35v, but above that the risk of the memory being damaged increases notably. what is your guys experience with 2.4v? has any of you run 2.4v for an extended period of time?
on what board and with what cooling?

maybe it would be useful to add a list of how mainboards overvolt vdimm.

P5K3 dlx 0.030 - 0.055
P5E3 non dlx 0.025-0.055
maximus xtreme 0.036V to 0.038V
P5E3 WS Pro 0.095V to 0.098V

I havent myself, but I've been told that TaPaKa // Sam has tested 2.4v 1300 hours with DDR3, dont even know if he has used a fan...

Personally I'm happy with your thread, but I just dont have faith in the high risk is excisting, as I havent had first hand experience confirming so, but the direct operset, further I've hurt of ppl going even further then myself..

[BenchZowner]

I havent myself, but I've been told that TaPaKa // Sam has tested 2.4v 1300 hours with DDR3, dont even know if he has used a fan...

2.4V fanless @ continuous full load would more than likely set the kits off in less than a month...probably even less than 15 days... even a single day...

Estimated again...but chances are it's correct.
Don't look at me though, I'm not going to risk any of my current DDR3 kits...they're here for a reason

[KTE]

Hmm.. I've never fed or needed to feed high MCH volts for decent MEM speeds/timings. Even this to 648 was stock MCH volts on P35 and 2.196 vDIMM real / 2.2 vDIMM BIOS. Only if benching that would I require +0.05 MCH volts, and X38 by nature required high MCH voltage to run DDR3 because of the straps being tighter/faster at low FSB. If we could run 2800PSB stable, I highly doubt you'd need those MCH volts for DDR3 frequencies

P5E3 Dlx definitely overvolts, between +0.05 to +0.12V I've seen with stable PSU rails, nearly killed Department Managers MEM because I never checked