Stability & RAM voltage

K.S. · *Last edited by K.S.; 07-10-2012 at 23:14.*

I'm using these DDR3-1866 modules, listed here: http://www.newegg.com/Product/Produc...82E16820220650

It's rated at 9-11-9-27 with 1.65V, however anything under 1.8V on my ASRock Extreme 4 x79 gives me errors on Linpack *OCCT* My question is, why would I need so much more voltage to stabilize the ram when it's listed by the manu @ 1.65V? I'm using the modules' xmp profile. I didn't overclock the cpu while testing this out. I got Linpack errors @ 3.2GHz *stock* with 1.65V, with 1.8V I got none.

Also, can 1.8V damage my memory? *I often leave my sys on 24/7.*

Agent-A01 · 07-10-2012, 23:27 | posts: 5,621 | Location: USA

with 1.8v you will burn out your CPUs memory controller in no time. You picked the wrong memory in the first place...

K.S. · 07-10-2012, 23:37 | posts: 736 | Location: Honolulu, Hawaii

Quote:

Originally Posted by Agent-A01

with 1.8v you will burn out your CPUs memory controller in no time. You picked the wrong memory in the first place...

I find that hard to believe being that's it's listed as being compatible only with "Intel X79 chipsets" Please, do tell, why is this the wrong ram?

This still fails to answer why I don't get errors with 1.8 as opposed to 1.65 with the listed SPD

Agent-A01 · 07-10-2012, 23:43 | posts: 5,621 | Location: USA

Quote:

Originally Posted by K.S.

I find that hard to believe being that's it's listed as being compatible only with "Intel X79 chipsets" Please, do tell, why is this the wrong ram?

This still fails to answer why I don't get errors with 1.8 as opposed to 1.65 with the listed SPD

Maybe you got bad ram, maybe the VCCIO voltage needs to be raised to 1.1 volts. Intel Recommends 1.5 voltage ram, with a max of 1.6 volts. 1.65 is out of intels voltage range due to risk of frying the IMC(integrated memory controller) on the CPU.

you should have got ram that is listed as 1.5 volts, like this
http://www.newegg.com/Product/Produc...82E16820231456

K.S. · 07-10-2012, 23:54 | posts: 736 | Location: Honolulu, Hawaii

Quote:

Originally Posted by Agent-A01

Maybe you got bad ram, maybe the VCCIO voltage needs to be raised to 1.1 volts. Intel Recommends 1.5 voltage ram, with a max of 1.6 volts. 1.65 is out of intels voltage range due to risk of frying the IMC(integrated memory controller) on the CPU.

you should have got ram that is listed as 1.5 volts, like this
http://www.newegg.com/Product/Produc...82E16820231456

huh, :/ well still thanks for helping.. that info helps alot!

IcE · 07-11-2012, 00:15 | posts: 6,990 | Location: Toledo, Ohio

As a temporary solution you can most likely get away with running the RAM at 1600 speed and lowering the voltage to 1.5V.

Lane · *Last edited by Lane; 07-11-2012 at 00:30.*

- Check on Patriot site, support forum, maybe the 1.8V result by a bad setting enabled (can be the 1T instead of 2T or whatever) ... if there's the same situation for some other, or if they can help you.
A kit rated at 1.65V, should not need 1.8V for be stable. In general this can mean a subtiming is too tight and the ram need more voltage ( same as when we tight the timing for overclocking and performance )

The ram are compatible XMP, so the ram should normally be plug and work with this setting. But it happend some motherboard bios dont support the ram at this moment or they need a fix... check on Asrock site the compatibility of the kit and in case, check with their support too.

K.S. · 07-11-2012, 00:37 | posts: 736 | Location: Honolulu, Hawaii

Quote:

Originally Posted by Lane

- Check on Patriot site, support forum, maybe the 1.8V result by a bad setting enabled (can be the 1T instead of 2T or whatever) ... if there's the same situation for some other, or if they can help you.
A kit rated at 1.65V, should not need 1.8V for be stable. In general this can mean a subtiming is too tight and the ram need more voltage ( same as when we tight the timing for overclocking and performance )

The ram are compatible XMP, so the ram should normally be plug and work with this setting. But it happend some motherboard bios dont support the ram at this moment or they need a fix... check on Asrock site the compatibility of the kit and in case, check with their support too.

Yeah I had the idea some timing was too tight, trying diff values atm trying to find a sweet spot while maintaining 9-11-9-27, I'm thinking trc min might be what's too tight *time will tell as I have to wait for it to stress it past the point of error*

---TK--- · 07-11-2012, 00:40 | posts: 14,696 | Location: New Jersey, USA

raise your viccio man and try lowering the ram volts. 1.8v is going to kill your ram

K.S. · *Last edited by K.S.; 07-11-2012 at 01:07.*

Quote:

Originally Posted by ---TK---

raise your viccio man and try lowering the ram volts. 1.8v is going to kill your ram

Already did lower it back to 1.65, and what's viccio? I can't find anything by that name listed in BIOS.

EDIT: figured it out, it's named as something else in mine.

The Chubu · *Last edited by The Chubu; 07-11-2012 at 01:17.*

1.65v isnt uncommon. Even the Kingston modules that Hilbert uses (or used, I didnt checked his current setup) use 1.65v. I've read that the 1.5v value is actually a very conservative estimate from Intel, thus you have 1.65v rams in the market selling like hotcakes anyway.

But 1.8v is too high. Phenom II may not have a problem with it (G2 mems for AMD use that kind of voltages), but Sandy Bridge/Ivy Bridge do have issues.

Just try to set every value to its stock value manually and see what you can get.

If you're going to change the mems, G Skill Ripjaws work on all kind of frequencies at 1.5v (ninja'd by agent

). I actually have 1600Mhz G2 mems "for AMD". At default they work at 1.8v or something like that, so i lowered them to 1333Mhz 1.5v for now and someday i'll try to OC 'em to 1600Mhz 1.5v (I hear they OC well enough).

K.S. · *Last edited by K.S.; 07-11-2012 at 01:41.*

Quote:

Originally Posted by The Chubu

1.65v isnt uncommon. Even the Kingston modules that Hilbert uses (or used, I didnt checked his current setup) use 1.65v. I've read that the 1.5v value is actually a very conservative estimate from Intel, thus you have 1.65v rams in the market selling like hotcakes anyway.

But 1.8v is too high. Phenom II may not have a problem with it (G2 mems for AMD use that kind of voltages), but Sandy Bridge/Ivy Bridge do have issues.

Just try to set every value to its stock value manually and see what you can get.

If you're going to change the mems, G Skill Ripjaws work on all kind of frequencies at 1.5v (ninja'd by agent

). I actually have 1600Mhz G2 mems "for AMD". At default they work at 1.8v or something like that, so i lowered them to 1333Mhz 1.5v for now and someday i'll try to OC 'em to 1600Mhz 1.5v (I hear they OC well enough).

Thanks for the input, yes it's become quite clear by now 1.8 is too much, that's already been established. As for stability, I went back to defaults for everything else and I'm trying a higher VTT *default is 1.1v* being that in theory it's supposed to be .5v below the DRAM voltage I'm trying VTT @ 1.15v atm, hopefully that'll do the trick, I'd totally forgotten about this value, hopefully this combined with the xmp profile will do the trick.

Brendruis · 07-11-2012, 01:43 | posts: 1,242 | Location: Melbourne, FL

I had a set of 1.65 DDR3 1600 once that would pass memtest86 but fail prime with rounding errors. Think it was Corsair kit, but who knows..

Since then I have used Intel spec 1.5 V kits and haven't had an issue in 2 dozen or so builds.

I know this doesn't help you now but food for thought..

Can those DIMM run at 1.5 1600 or 1333? might be worth it to do so for stability. The difference between 1600 and 1866 is well not major.. You see I have 1333 it doesn't make a huge difference.. the most I've noticed it is in synthetic benches like Sandra.

---TK--- · 07-11-2012, 01:30 | posts: 14,696 | Location: New Jersey, USA

my ram is rated at 1.6v for 1600 2t. I have it oc to 2133 1t just 1.5375v. start at like 1.60 and stress test. I guess ur vccio or qpi/vtt is 1.050v? try 1.075 or so. I dont have SBE so I could be mistaken

The Chubu · 07-11-2012, 01:43 | posts: 2,534 | Location: Look out!

XMP it's just a way for the motherboards to adjust mem values that are outside JEDEC spec (mostly anything that isnt 1.5v 1333Mhz 9-9-9), when the motherboard doesnt finds an XMP profile, it defaults back to any of the JEDEC standard profiles. Its nothing you can't actually set up by yourself.

K.S. · *Last edited by K.S.; 07-11-2012 at 02:52.*

Quote:

Originally Posted by The Chubu

XMP it's just a way for the motherboards to adjust mem values that are outside JEDEC spec (mostly anything that isnt 1.5v 1333Mhz 9-9-9), when the motherboard doesnt finds an XMP profile, it defaults back to any of the JEDEC standard profiles. Its nothing you can't actually set up by yourself.

um... yeah... I'm using the xmp profile already oO... and modifying a few additional values because the ram isn't stable obviously... anyways I'll see how higher VTT & VCCSA values work out... just read something on an asrock forum about that...

EDIT: clarification, I'm using the timings from the XMP profile, in BIOS I'm just using 'standard' so I can customize the voltages for VTT and VCCSA *loading XMP overrides VCCSA*. So far looks like raising the VTT & VCCSA allowed me to run the RAM @ 1866 MHZ with the timings from the XMP and it appears to be stable, now just working on lowering the voltages right bfr point of error so I'm not using more than needed.

Agent-A01 · 07-11-2012, 02:56 | posts: 5,621 | Location: USA

VCSSA voltage should be left alone. Defualt is .925v

K.S. · *Last edited by K.S.; 07-11-2012 at 03:10.*

Quote:

Originally Posted by Agent-A01

VCSSA voltage should be left alone. Defualt is .925v

EDIT: ok this is weird, it looks like when the XMP profile is loaded the VCCSA goes up to 1.2v, give or take... weird. I've got it lowered to mobo default now. Going to mess with some timings and VTT...

Agent-A01 · 07-11-2012, 08:02 | posts: 5,621 | Location: USA

Quote:

Originally Posted by K.S.

EDIT: ok this is weird, it looks like when the XMP profile is loaded the VCCSA goes up to 1.2v, give or take... weird. I've got it lowered to mobo default now. Going to mess with some timings and VTT...

thats not good.. you can fry the chip that controls the vcssa with that high of voltage. intel says it should not be touched under any circumstances.

K.S. · *Last edited by K.S.; 07-11-2012 at 12:35.*

Quote:

Originally Posted by Agent-A01

thats not good.. you can fry the chip that controls the vcssa with that high of voltage. intel says it should not be touched under any circumstances.

Yeah confirmed it, using my xmp profile bumps VCSSA up to 1.201v... wtf... anyways I loaded the xmp, wrote down most of the timings it sets, *i.e. 9-11-9-27, tras min etc...* then I unloaded xmp replacing it with standard, put in all the values I wrote down, set VTT to 1.15v and DRAM to 1.65v, all other voltages excluding vcore are at stock settings now.

stress-tested 1866MHZ @ 1.65v *9-11-9-27* 1T under 6 hours OCCT Linpack, with cpu @ 4.2GHz, and no errors! ugh, Finally! All I know is my ram's finally working at advertised settings it just took a lot of tinkering.. :/ anyways, thanks to all for the help, esp. Agent-A01, much appreciated!

edit: damn I'm lucky my mobo held out through this mess...nothing's fried

*phew*

Agent-A01 · 07-11-2012, 16:01 | posts: 5,621 | Location: USA

Glad to be of help. some RAM kits will run at less voltage than what they are specified. Id try 1.6 volts to be on the safe side, if its stable leave it at that. my 2133 ram says 1.6 volts, but it runs fine on 1.5 volts, 100% stable.

K.S. · *Last edited by K.S.; 07-11-2012 at 16:37.*

Quote:

Originally Posted by Agent-A01

Glad to be of help. now in the future, some RAM kits will run at less voltage than what they are specified. Id try 1.6 volts to be on the safe side, if its stable leave it at that. my 2133 ram says 1.6 volts, but it runs fine on 1.5 volts, 100% stable.

Good to know! Thanks again! :_)

god everything's running like butter now... ahh