Core i7 Vdimm/VTT Graveyard :D

[Astratuner]

mua maybe intell knows that the lenght of the cpu's life will be shortend by a few months... Us the normal users will have a hard time knowing that info or some intel engineer should post here.

[T_Flight]

SOME boards jack up vdimm vcore and vtt automatically when you load an xmp profile or set a high memory speed and leave the voltages on auto/default!
the voltages they load are supposedly safe, but you might want to set them a little lower and manually just to be safe.

bad news, i heard of one cpu that died with 2.25v vdimm...
might have been a ES cpu or ES board, im waiting for more details and will keep you updated.

Yeah, I read this on the Corsair Forums too. Right now we're in the end user test phase...hehehe...I think it's better to set all settings manually until everybody gets to play with all the different settings and see what they do. I'm real careful with mine right now. I'm really watching myself, and trust nothing I read in BIOS. I'm checking and testing through ProbeIt manually to make sure the BIOS agrees...there ARE variances of course.

I just got mine running lastnight, so I'm controlling myself for the time being. I'm also on stock retail cooling, so I kinda have to chill with it.

[dinos22]

SOME boards jack up vdimm vcore and vtt automatically when you load an xmp profile or set a high memory speed and leave the voltages on auto/default!
the voltages they load are supposedly safe, but you might want to set them a little lower and manually just to be safe.

bad news, i heard of one cpu that died with 2.25v vdimm...
might have been a ES cpu or ES board, im waiting for more details and will keep you updated.

yup mine does

as far as i can tell the XMP profile increases uncore to 1.55 and ram to 1.65

last night i worked out that i can run much higher QPI with lower uncore than first night

i will do a thread soon to report my findings which will hopefully save a few headaches for ppl OCing GB boards at very least i hope

[Viss]

1,475 VTT and 1.78 vdimm for now. CPU needs a LOT of volt. Mem 8-8-8-18-1T.

[Benji Tshi]

There's a problem : 143x28 = 4004, not 3986. Moreover, i get 9m flat with the same settings everywhere (uncore, ram, cpu...). Perhaps my 0703 is faster than 0804 :/

[dinos22]

its not a problem

clocks seems to jump up and down like that
its weird but it happens here too

you are right about his time
it looks slow

[Viss]

Seems multi/bclck combination, with some clocks change all the time. And yes its sloooow just started to play and a lot to do for sure

[saaya]

YES.. you are right saaya, i have this experience with Patriot Viper DDR3-1600 6GB Kit. When XMP enable, suddenly vdimm set to 1.72 volt.

well thats what xmp and epp are meant for, they can set vdimm too, but what i mean is that some mainboards BIOS increase vdimm to much more than the xmp has programmed in it (up to 2.1v) if you select high bclock or memory clock settings.

yup mine does
as far as i can tell the XMP profile increases uncore to 1.55 and ram to 1.65
last night i worked out that i can run much higher QPI with lower uncore than first night
i will do a thread soon to report my findings which will hopefully save a few headaches for ppl OCing GB boards at very least i hope

huh? your xmp profile increase uncore voltage?
and to 1.55v? wee, thats a big high isnt it?
i mean everybody is running 1.55 uncore when pushing i7 it seems, but it shouldnt be the default setting!

you can run higher qpi with lower uncore, what do you mean?
higher qpi ratio? higher bclock? higher overall qpi speed?
with lower uncore voltage or lower uncore ratio?

update about the first dead i7:
it was a 965 and it died after one day of 2.2-2.4v vdimm

[bmg]

update about the first dead i7:
it was a 965 and it died after one day of 2.2-2.4v vdimm

What were the other voltages? Seems that's as important to know as vdimm.

[Benji Tshi]

And was it a box or ES ? It seems that rev B0 & C0 are quite different !

[dinos22]

well thats what xmp and epp are meant for, they can set vdimm too, but what i mean is that some mainboards BIOS increase vdimm to much more than the xmp has programmed in it (up to 2.1v) if you select high bclock or memory clock settings.


huh? your xmp profile increase uncore voltage?
and to 1.55v? wee, thats a big high isnt it?
i mean everybody is running 1.55 uncore when pushing i7 it seems, but it shouldnt be the default setting!

you can run higher qpi with lower uncore, what do you mean?
higher qpi ratio? higher bclock? higher overall qpi speed?
with lower uncore voltage or lower uncore ratio?

update about the first dead i7:
it was a 965 and it died after one day of 2.2-2.4v vdimm

at first when i started testing i found that i had to run high uncore to get stability

i've figured out how to run high BCKL without high uncore voltage now
thats what i meant

yeah when you select profile 1 in XMP profile it says that vdimm is 1.65 and QPI voltage is 1.55...i assume that is uncore voltage

it took me a while to realise that to get stable BCKL you have to run QPI clocks at 4xRAM MHz

so 900 on RAM has to be 900x4=3600 in QPI or forget about stability

i've been running around 1.42v uncore and 1.74vdimm for most benches
i've also had the vcore at 1.5-1.55 on air in the last few sessions

i'm already sick of air testing
will have to insulate this GB board for single stage....pain in the arse to insulate

saaya can you ask the guy if he ran low vcore with high vdimm
i have a feeling that you have to jack up the vcore as well when you run high vdimm like back in AMD days ...i do it as a precaution lol

[Pt1t]

And was it a box or ES ? It seems that rev B0 & C0 are quite different !

IT was ES , Vdimm 2.25V , VTT 1.45v , on air

I played a full day with RAM test./

Maybe more Vdimm after some tweak it play ... or tweak it applied more than expected.

[dinos22]

thomas
are you the guy saaya is referring to
what board?
what vcore
why did you run such a high difference between VTT and vdimm
francois suggested within 0.5V difference between uncore and vdimm

[Pt1t]

thomas
are you the guy saaya is referring to
what board?
what vcore
why did you run such a high difference between VTT and vdimm
francois suggested within 0.5V difference between uncore and vdimm

Vcore 1.5v
What Board got Tweak It ^^
I was killed before i read what francois said.

[dinos22]

looks up tweak it on google

[Pt1t]

looks up tweak it on google

http://www.matbe.com/images/biblio/a...0000078955.jpg

[Kean]

looks up tweak it on google

Rampage II Extreme

[dinos22]

yeah i realised as soon as i googled

this might be a silly question but how do you know that's what killed the CPU

[Pt1t]

yeah i realised as soon as i googled

this might be a silly question but how do you know that's what killed the CPU

God knows

[dinos22]

sorry to hear
it sucks losing CPUs
first CPUs i killed were 45nm intels
they arent as tough as they used to be heh

[Benji Tshi]

Why shouldn't we set more than 0.5v difference between vtt and vdimm ? I run 1.5/2.2 for few hours without problems (i did't say that it would be no consequence)

[Pt1t]

sorry to hear
it sucks losing CPUs
first CPUs i killed were 45nm intels
they arent as tough as they used to be heh

yeah ... First cpu intel i killed was my gem E8600 ...

[Oliver]

thx for the information

and im sorry to hear about your loss Thomas

beg up and keep pushing it

[saaya]

it took me a while to realise that to get stable BCKL you have to run QPI clocks at 4xRAM MHz

so 900 on RAM has to be 900x4=3600 in QPI or forget about stability

hmmmm damn all this different terms for core i7 architecture

so you mean qpi speed has to be 3600Mhz for ddr3 1800?
4.8GT/s is 2400Mhz is 18x133
5.6GT/s is 2800Mhz is 21x133
6.4GT/s is 3200Mhz is 24x133
You mean GT/s? even that doesnt make sense... sorry, i had a few beer maybe thats it but i really dont get you m8
what do you mean with qpi has to be 3600Mhz for 900Mhz ram?

i'm already sick of air testing
will have to insulate this GB board for single stage....pain in the arse to insulate

dont expect too much man, from what i heard most chips only scale a few hundred mhz on ln2 :/
btw, i asked our engineers about giving more space around the cpu socket and they said it will definately impact how well the pwm works and how stable vcore is, so... they recommend to keep the caps relatively close to the socket. i asked for at least 1cm free space around the socket.
and you can actually unscrew the metal socket part of 1366, that really helps for insulation! the cpu wont be clamped down anymore, so pressure has to keep it down and after removing the pot itll stick to it, but thats ok
its worth the way improved insulation once the metal stuff around the socket is gone

thomas
are you the guy saaya is referring to
what board?
what vcore
why did you run such a high difference between VTT and vdimm
francois suggested within 0.5V difference between uncore and vdimm

well what do you expect, 1.7v vtt to have 2.2v vdimm?
then the cpu would probably not die from vdimm but rather vtt

if you want to kill an i7 all it takes is high vtt...
dont ask how i know that ^^

thanks a lot for sharing your information thomas!

[dinos22]

i have to do a bios template
i am still lost with all these names to be honest
keep in mind that i've only had a couple of days testing experience so my findings are fresh to say the least heheh

but yeah what i've found on the GB board is while you could run NB Frequency in a different ratio to 4:1 i am talking about for high BCKL clocks and benching speeds on RAM i found that using NB Frequency 4x bigger than RAM frequency is easiest to get stability

let me show with an example

RAM is running at 1018MHz and NB Frequency is at 4x times that >> 4072MHz
do you understand what i mean





saaya what is dangerous VTT
i've run 1.65v so far heheheh