How to set up GTL Ref Values for 45nm & 65nm

[Demo]

@Cryptik and mikeyakame

Thanks for the tips, for stock speed since the initial post here recommends 67% should i set it at 67% or leave it at 63% ?


In regards to my OC, taking the Vcore to 1.400 shoots my temps past 72c, at 1.355 im already getting 65-66c under prime and 77-78c under linpak64

[CryptiK]

Try what Mikeyakame suggested and raise vcore to about 1.39 bios, and raise Vtt (vFSB) to at least 1.3v, preferably higher to around what you set vcore to, and do some testing at 3.4GHz. As he said you will probably find that a GTL Ref of 63% - 65% will be the optimal setting, he uses a Q6600 too and has a lot of experience with it. You can always then raise it to 67% and see if it benefits stability.

If you don't want to go any higher than 3.2GHz, and you are stable now, there is little for you to gain by GTL adjustment. You may like to drop vcore one notch, raise Vtt a little, and try different GTL settings to see if you can get it stable on less vcore than you are using now. I suggested in the first post to start at 0.667x (67%) but it appears Q6600's like the 63% - 65% range better. Just try it and see.

[Demo]

@CrystiK

Got it, checked my bios now and it seems like im only limited to 1.2375VTT

I would assume that playing around with the VTT and GTLREF provides no risk or what so ever?

EDIT:

Right now ill try vtt at 1.2375 and GTLREF at 67% in hopes that i can run at lower vcore, trying 1.315vcore under prime95SmallFTT.

[CryptiK]

You're fine up to ~1.4v Vtt, some say it's fine to go up to 1.45v, although I'm hesitant to recommend that. Your GTL's are fine so if you keep Vref under about 1.0v. That means GTL Ref x Vtt = <1.0v. If you have 1.2375v Vtt and you are using 63% GTL Ref (0.63x) your Vref is 0.779v which is absolutely safe. If you use 67% you get 0.829v which is still very safe.

[Demo]

My CPU is stubborn, i get result errors under linpak64 anything under 1.355v in bios vcore

[mikeyakame]

then ur cpu needs more voltage simple! its nto stubborn, its just the quality of the chip coming off wafer, needing mroe voltage isnt always a bad thing, some chips need a little more voltage due to minor flaws in certain parts of the chip, but clock even higher than other chips which need less voltage all round. it all comes down to defect rate on the particular wafer your cpu was fabbed on, and which area/areas are effected by defects, sometimes they can render a chip useless, or other times they can effect transistors in a small area on outer edge of defect zone in area of chip that is responsible for specific functionality, which means that chip might be weaker at clocking or requires a little more voltage to be stable, but on the other hand it may excel at the other area as that part of the chip is as good as a highest binned chip and the rest of the cpu ends up missing any other nearby defect areas.

voltage means nothing, its all about how efficient your cpu is at a given voltage. Lets say u give3 one chip 1.25v and another 1.35v, if the chip which needs 1.35v runs at the same efficiency as a chip using 1.25v, they should both produce nearly identical thermal output, since heat displaced is equivalent to work done in the system. same work = same heat displacement even if it requires more voltage to do it. higher voltage for any given clock freq will only create excess heat if the cpu is inefficient in its output at that frequency. 1.30v and 1.50v makes no difference on a cpu running 4.0ghz for example, even though the difference in voltage input is huge, if they both perform at the same efficiency whether using 1.3 or 1.5 they will both put out similar temps.

only downside of needing more voltage to switch transistor gates on a cpu for example, is the mosfets need to work harder and generate more heat and waste slightly more power than would otherwise with a lesser input voltage.

[Demo]

with the CPU VTT at 1.2375 i dont get anymore result errors from Linpak64 w/ the same old 1.355v core and also reduced the GTLREF down to 65%.

[R104_for]

My setting, correct?

Cpu Vtt 1.36v

CPU GTL -45mV

[KURTZ]

interesting thread Cryptik

however that's my question:

my vFSB is 1.29 x 0.635 = 0.819
so 1.29 x o.667 = 0.860
0.819 - 0.860 = - 0.041mv

is correct? so i must set the CPU GTL 2 and 4 to this value?

and what about the GTL 1 and 3?

thx

[radaja]

@cryptik. quik question, on the EP45-UD3P it has one setting for cpu ref,is this how i should set vtt1.28x0.635=0.812,so in bios set ref at what brings ref volt closest to that ex/in ET6 after setting in bios it shows 0.80ref . is this right?

[CryptiK]

with the CPU VTT at 1.2375 i dont get anymore result errors from Linpak64 w/ the same old 1.355v core and also reduced the GTLREF down to 65%.

Good, then you have gained some stability. That Vtt is still low, so don't worry about leaving it like that, it's fine.

My setting, correct?

Cpu Vtt 1.36v

CPU GTL -45mV

Depends on what GTL Ref options your board has, if it has only one option 'CPU GTL Ref' then its fine as a starting point, but feel free to experiment with it a little to find the best value.

interesting thread Cryptik

however that's my question:

my vFSB is 1.29 x 0.635 = 0.819
so 1.29 x o.667 = 0.860
0.819 - 0.860 = - 0.041mv

is correct? so i must set the CPU GTL 2 and 4 to this value?

and what about the GTL 1 and 3?

thx

Yes thats right, to start with. Leave 1 & 3 on auto, and set -40mv on 2 & 4 (most boards actually have the GTL Refs labelled 0,1,2,3 but if yours has 1,2,3,4 thats fine.

Do try different settings though, for example you could set +10 for 1 & 3 and -30 for 2 & 4 etc just keep the resultant Vref around the same by using the formula. You can also experiment with adjusting them separately too.

@cryptik. quik question, on the EP45-UD3P it has one setting for cpu ref,is this how i should set vtt1.28x0.635=0.812,so in bios set ref at what brings ref volt closest to that ex/in ET6 after setting in bios it shows 0.80ref . is this right?

That's fine, ET6 is reading out the actual GTL Ref voltage (0.80v) after you set the multiplier (0.635x) in bios.

1.28v x 0.635 = 0.812v, the actual value real by the ET6 software is 0.80v, so slightly under the calculated value.

[Demo]

@CryptiK

Thats the highest VTT selection Abit IP35-E offers though.

[CryptiK]

Yeah that's ok, you'd be safe up to ~1.4 but I was just saying leave it where you have it set now (max board allows). Although it's the boards max, it's still comparatively low, on most boards can set it up to ~1.7 - 2.0v.

[KURTZ]

Do try different settings though, for example you could set +10 for 1 & 3 and -30 for 2 & 4 etc just keep the resultant Vref around the same by using the formula. You can also experiment with adjusting them separately too.

+10 and -30? not +10 and -50 Cryptik? (if my result is -40mv)

[CryptiK]

You should have options that have the default multi's

GTL 0 = 0.635x
GTL 1 = 0.667x
GTL 2 = 0.635x
GTL 3 = 0.667x

so you could have

auto
-40
auto
-40

and your Vref would be close to all the same

or:

+10
-30
+10
-30

or:

+20
-20
+20
-20

Or:

+30
-10
+30
-10

or:

+40
auto
+40
auto

or what I use with 1.2 Vtt:

+50
+10
+50
+10

See the resultant Vrefs increase you just keep the ratio between the 0.635x and 0.667x GTL Refs the same.

[spoof]

OK, since i'm with P5E-Deluxe and i don't want to flash with Rampage BIOS (so i can read the voltages), what should i set for GTLREF? I have from 51x to 65x. And which value determinate the VTT ? CPU PLL or FSBt? So, increasing GTLREF, what should i decrease?

[WhiteFireDragon]

how do i set GTL for a 65nm dual core? when i stress test both core 0 and 1, it consistently fails on core 1 first. in my M2F bios, it shows the following equation for GTL:

core-0: vtt x .635 + item
core-1: vtt x .667 + item

what i don't understand is that since it's a 65nm chip, then shouldn't i change the "Y" value for core-0 to match the .667? in my case, why is core-1 less stable? then do i need to change my core-0 or core-1? my vtt is at 1.37v, so i should set the "Y" value of core-0 should be about +43mV correct?

[ante_ante]

Sitting on a Rampage Extreme and a E8600. The board has CPU GTL Ref 0,1,2,3

VVT = 1.20

So i calcute this

1.20 x 0.635 = 0,762
1.20 x 0.667 = 0,800

0,762 - 0,800 = 0,038

= ~ -40mV

I´m i on the right track ?

CPU GTL Ref 0 = Auto
CPU GTL Ref 1 = -40mV
CPU GTL Ref 2 = Auto
CPU GTL Ref 3 = -40mV

How can i calculate NB GTL ?

[Razzz]

I have the MSI P45 Diamond and I cant get it to post at 400FSB. Im using a Q9550.
My BIOS looks like this:

I can adjust the CPU GTL and the MCH GTL in small steps and If I have understood this right, I should let it be somewhere between 0,8 and 1,0? I don't have to do the calculation.
I have tried to just adjust the MCH GTL to 0,6369v and then left everything else on auto, that works ok up until about 375FSB. If I want it to boot at 380 I have to change that setting to auto and put a very high vMCH and VTT.
Any ideas what I should set the CPU GTL and MHC GTL to make it boot at 400FSB if that's even possible? I know it's not an extreme motherboard and its a quad, but 400FSB? I'm looking for bench-settings.

[hoskic]

Which GTL Ref adjustment should I use on my M2F with q6600?
+ or - ond how much mV on 0,1,2,3 ?

[spoof]

I'm with P5E-Deluxe@Rampage Formula. When i set 1.42VTT in BIOS, Everest shows, that VTT is 1.34. Which one should i trust - sensoar reading or bios setup? My GTLREF setting is 65x.

[Demo]

Omg, those are killer voltages!

[spoof]

I know that, that's why i'm asking, which one should i trust, because i can't the CPU working ot 465x8.5 with 1.4VTT. I just keep freezing and freezing in idle, no matter what i do! Please, help me.

[carepolice]

I know that, that's why i'm asking, which one should i trust, because i can't the CPU working ot 465x8.5 with 1.4VTT. I just keep freezing and freezing in idle, no matter what i do! Please, help me.

Lower your VTT and increase CPU voltage, should make difference.

And razz you really know now what your doing? Those volts for 400FSB... Sigh.
You should be really happy if you havent blowed something up already.

[Razzz]

Lower your VTT and increase CPU voltage, should make difference.

And razz you really know now what your doing? Those volts for 400FSB... Sigh.
You should be really happy if you havent blowed something up already.

The print-screen is only to show the max-voltage for a review