is the 0.5v limit between VTT/IMC and dimm a myth?

[Chrysalis]

Simple question and if the answer is no please explain with the info I am posting.

Intel states i5 can support ram up to 1.65v
The default IMC voltage is 1.1v
That means if I set 1.65v for my ram on my motherboard and leave everything else on auto I have a 0.55v gap.

Since I am not sure I also upped my IMC to 1.15v manually and hope its safe, but if the .5v is documented then it would seem asus need to tune the auto algorithms as they allow a bigger gap.

[maxxx]

--Edit....

[Chrysalis]

I meant 0.55v gap sorry, above the limit still.

[radaja]

its actually 0.5v
so 1.65v vdimm means vtt at 1.15v
1.70v vdimm means vtt at 1.20v

[Chrysalis]

well my board isnt doing that without me overiding it.

so I could go in my bios, tell it to use the XMP profile and it will set voltage to 1.65v and leave vtt at 1.1v. Seems strange behaviour.

so intel give the message 1.65v ram is safe but to use it a user has to manually set a higher voltage.

[Johnny87au]

So it has to be .5, So i could have my dimm's @ 1.65v and say vtt 1.3? Or does it have to be .5v exact?

[OC Nub]

You want them to be within 0.5v of each other. It doesnt meam stay exactly 0.5v apart.

[Johnny87au]

Oh so anything over 0.5v is bad, So what i mentioned above i could do that?

[SimpleTECH]

Oh so anything over 0.5v is bad, So what i mentioned above i could do that?

Yes you could possibility do that.

Remember that some boards tend to undervolt. That is why Gigabyte lists their vDimm voltage at 1.66v instead of 1.65v.

[Johnny87au]

Yes you could possibility do that.

Remember that some boards tend to undervolt. That is why Gigabyte lists their vDimm voltage at 1.66v instead of 1.65v.

Yeh noticed that, i guess 1.66 is the new 1.65

[[XC] gomeler]

When do we ever have a difference of anything close to 0.5vdc though? Typically I'm running around a 0.2vdc difference between vtt and vdimm The only time you need to worry is with something like D9GTR/GTS where you are pumping 2.2vdimm with ~1.5vtt but even then I suspect everything would be a-ok.

[Chrysalis]

well I am at 0.5v diff as I dont really want to increase vtt more then I need to. So far seems stable for me with a 1.15vtt and 1.65v ram so I am happy. My post was just wanting to verify the need of keeping it within 0.5v as I had not seen any official info on it.

[Leeghoofd]

Wasn't this solely for I7 ? Or does it apply also for I5 ?

[Johnny87au]

hes saying 0.5 is the maximum gap you should use, Anything 0.5 and under is fine

[Chrysalis]

Wasn't this solely for I7 ? Or does it apply also for I5 ?

Here is what I have read.

intel state 2 things.

max 1.65v for ram
max 1.21v for vtt - (is now 1.4v correction.)

my board defaults to 1.1v for vtt and that value doesnt change if you increase the ram voltage.

I have never found a sound reason for this max 0.5v gap other than it seems to be an assumption based on that the default vtt for i7 is 0.5v below the ram voltage limit. However the i5 (at least on my board) has a default vtt that is 0.55v below the ram voltage limit.

However I am not going to experiment, I figured it safer to slightly bump my vtt so I stay within this 0.5v limit.

[Russ_64]

Here is what I have read.

intel state 2 things.

max 1.65v for ram
max 1.21v for vtt

Please can you post the source for this - it is news to me.......

[Brama]

I have never found a sound reason for this max 0.5v gap other than it seems to be an assumption based on that the default vtt for i7 is 0.5v below the ram voltage limit.

This "rule" applies to any processor with internal memory controller.

As you know, the memory controller is in the uncore region of cpu and it is powered by Vtt (or Vqpi in Asus boards).

The memories are directly connected to memory controller and they are powered by Vmem.

As Vmem is higher than Vtt, in order to protect uncore from Vmem, there are some diodes that are like open switches until there is a difference from Vmem and Vtt <0.5 V.

If this difference is > 0.5 V, the silicon junction of diodes opens and the Vmem can arrive to uncore, causing unpredictable effects.


It is not a metropolitan myth.

[donmarkoni]

Here is what I have read.

intel state 2 things.

max 1.65v for ram
max 1.21v for vtt

I would like to know where you read it too.

AFAIK, these are the Intel's specs for i7 900 series:



and these for 980X:

[massman]

Wasn't this solely for I7 ? Or does it apply also for I5 ?

Core i5, the platform where Vdimm is not deadly?

[radaja]

i think the original poster is on a P55 setup.
the original data sheets from intel had been 1.21v vtt max
they later revised it to 1.40v vtt

[Chrysalis]

Please can you post the source for this - it is news to me.......

http://www.intel.org/design/corei5/documentation.htm

radaja is correct the new documents have indeed revised it to 1.4v for vtt. scroll to 7.8 table 7-4

although the original point remains I guess and that is default voltages have a 0.55v gap between vtt and ram voltage limit.