Gigabyte GA X38 DQ6 Blog

[supershanks]

Hehe, sorry SuperShanks you clicked submit just before me on XtremeGamers question hahaha.

no worries, reinforces the point

2.35v appears to be when you run very tight timings, fine for a quick bench, but not 24/7. i'd feel pretty safe though running 2.2v at 1037 you might get aweay with 2.1~2.2v I'd still use active cooling whenevver using >2.0v

luck

[Cosmic]

Originally Posted by supershanks

...
... i'd feel pretty safe though running 2.2v at 1037 ...

Wow what a difference 0.05 of a volt can make ... after upping my ram to 2.2v I've now been able to drop my ram from

5,5,5,13,trfc 34,trd 8,

down to

5,5,4,12,trfc 32,trd 7!!

(with a trd of 7 before it wouldnt even boot ).

Have been running prime now for over an hour with no errors. Will give it another hr, then put it on memtest before I go to bed.

Also you said ...

+0.40v not sure on the rated voltage of your ram I'm assuming 2.2v
may let you drop Tras to 10~12

I always thought that tRAS had to be at least tRCD+CAS latency + 2 cycles (REF). Is it worth testing tRAS lower than this?

Thanks again Supershanks, you, clunk and the forum are wellspring of valuable info !

Cosmic

[supershanks]

On P35 i could only run with a tras of 12. it was only after I got stable at 400x9 that i tried tras at 10 & was suprised when i managed it with prinme stable. Since then i've been able to run it on this & the Abit IX38 Quad. Only time I was unable was with the E8400, where i had to go back to 12.

Maybe the X38 Memory controller is a bit better than the P35, which enables the tighter setting.

Thanks again Supershanks, you, clunk and the forum are wellspring of valuable info !

Cheers Cosmic

[Cosmic]

Ok so my new best stable results are below. Tested on prime95 for 6h 7m.

Also have been playing with the new F8D beta. It defiantly improves the situation with the other Memory Multiplier Straps. None of the x.xxA straps (other than 1:1) would allow me to boot on F7 but all now let me boot using F8D.

Sadly still cant break the 3.5GHz wall I've hit even when pushing vcore up to 1.42560 and FSB Voltage to +0.20V which is about as high as I can go with my current cooling setup. Still my RAM timings are much tighter now and 3.5 is pretty damn respectable for air cooling (Especially here is sunny Queensland).

Re the FSB Voltage; What components are stressed when upping this setting? I was considering pushing it a bit further but thought I'd better research it a little more before I did ?

Also anyone know a good link/reference that explains what the GTLREF and Reference Voltage and Termination Voltage settings do ?

Thanks

Cosmic.

BIOS Revision: F8D Beta

Advanced Bios Features
----------------------
   
CPU Enhanced Halt (C1E) ................... [Disabled]
CPU Thermal Monitor 2(TM2) ................ [Enabled]
CPU EIST Function ......................... [Disabled]
   

PC Health Status
----------------
   
CPU Warning Temperature ................... [70C/158F]
CPU Smart Fan Control ..................... [Disabled]


MB Intelligent Tweaker (M.I.T.)
-------------------------------
   
Robust Graphic Booster ..................... [Turbo]
CPU Clock Ratio ............................ [ 9 x]
CPU Frequency .............................. 3.51GZh (390x9) ***
CPU Host Clock Control ..................... [Enabled] 
x CPU Host Frequency(MHz) .................. 390 ***
PCI Express Frequency(MHz) ................. [Auto]
C.I.A.2 .................................... [Disabled]
Performance Enhance ........................ [Standard] ***
System Memory Multiplier ..........(SPD) ... [2.66C]
Memory Frequency(MHz) ............. 800 .... [1040] ***
DRAM Timing Selectable ........... (SPD) ... [Manual]

********** Standard Timing Control **********

CAS Latency Time .................    5  ... [5]
DRAM RAS# to CAS# Delay ..........    5  ... [5]
DRAM RAS# Precharge ..............    5  ... [4] ***
Precharge Delay(tRAS) ............   18  ... [13] ***

********** Advanced Timing Control **********

ACT to ACT Delay(tRRD) ...........    3  ... [4]
Rank Write To READ Delay .........    3  ... [4]
Write To Precharge Delay .........    6  ... [8]
Refresh To ACT Delay .............   42  ... [32] ***
Read To Prechard Delay ...........    3  ... [4]
Static tRead Value ...............    7  ... [7] ***
Static tRead Phase Adjust ........    1  ... [Auto]
Command Rate(CMD) ................    2  ... [2T]

******* Clock Driving & Skew Control ********

CPU/PCIEX Clock Driving Control ............ [1000mV]
CPU Clock Skew Control ..................... [Normal]
(G)MCH Clock Skew Control .................. [Normal] 

********* System Voltage Optimized **********

System Voltage Control ..................... [Manual]
DDR2 OverVoltage Control ................... [+0.40V] ***
PCI-E OverVoltage Control .................. [+0.35V]
FSB OverVoltage Control .................... [+0.20V] ***
(G)MCH OverVoltage Control ................. [+0.125]
MCH Reference Voltage Control .............. [Normal]
DDR Reference Voltage Control .............. [Normal]
DDR Termination Voltage Control ............ [Normal]
CPU GTLREF1 Voltage Ratio .................. [-3.00%] *** (new default with F8D ?)
CPU GTLREF2 Voltage Ratio .................. [Normal] 
Loadline Calibration ....................... [Enabled]
CPU Voltage Control ........................ [1.41250V] ***
Normal CPU Vcore ........................... 1.28750V

*** Recently Adujusted Settings.

[supershanks]

Cosmic Cosmic
This was for the Abit but apart from the terminology being different (FSB Overvoltage Control = 1.20v +overvolt = Vtt voltage ) the rest is the same. GTL is like a fine tuning but loses it's impact at high temps.
The anandtech link is for the DFI board The VTT/FSB shown Is I've found considerably overstated. I got 8x465 Stable with FSB overvolt still on 0.20v.

Adjusting [Advanced] Gunning Transceiver Logic (A/GTL+) Voltage Levels
for Increased Front Side Bus (FSB) Signaling Margins and Overclocking

Intel rates the maximum VTT voltage for desktop processors at 1.55v

I've had problems understanding Gigabyte's GTLRef Controls. On most Bios their shown as a % roughly in the range 60~70% as described in the Anantech article (Ocing Quads section). Yet the DQ6 bios has -3% etc which is obviously relative to something but what
To be honest i never found any gain by using them.

So my experience was that FSB overvolt is a must whilst GTLref was nowt

luck

[Cosmic]

Originally Posted by supershanks

Cosmic Cosmic

Adjusting [Advanced] Gunning Transceiver Logic (A/GTL+) Voltage Levels
for Increased Front Side Bus (FSB) Signaling Margins and Overclocking

...

So my experience was that FSB overvolt is a must whilst GTLref was nowt

luck

Great Link . While I have some electrical eng in my background (many years ago) it will take me a while to get my head around all the detail there. Still even at a quick glance it gives a good concept of what the GTL settings are about though. Also why some settings of FSB just wont work ("FSB holes" as he calls them).

RE the Vtt; so with my FSB Overvoltage Control set at +0.20v adding that to 1.20v I should have another 0.15v that I can play with safely without going past the intel 1.55v sepcification?

Hmm since I am hitting a wall at 3.5Ghz (the beginning of my FSB Hole ? Possibly), I wonder if its worth trying a few FSB settings significantly higher than that ? say 3.75 to 3.8GHZ or something ? So far ive only tried from 3.55 - 3.66Ghz. Im guessing though that a stable oc at 3.75-3.8 would require higher vCore than I can afford to supply (due to heat).

Cosmic.

[Zenoth]

Concerning the CPU GTLREF1 and CPU GTLREF2 options, I've found the following (that's in French, at about the middle of the very first introduction post): Gigabyte X38-DQ6 // X38-DS5 et (E)X38-DS4 [Topic Unique v2] - CPU, Mobo, Ram - Hardware - FORUM HardWare.fr

Quoting:

Dans le bios F6 (beta ou non) comme le F7, il existe deux lignes cachées dans le M.I.T. (menu des paramètres d'overclocking) seulement accessibles en appuyant sur Ctrl+F1 dans la page d'accueil du bios, ce sont:

CPU GTLREF1
CPU GTLREF2

http://img404.imageshack.us/img404/9594/pict0002jg8.jpg
http://img404.imageshack.us/img404/8074/pict0008gi5.jpg

Il semblerait que ces deux éléments du bios servent à controler le Vdrop en augmentant le voltage CPU (Vcore) lorsqu'il est en charge, ce qui permet de réduire drastiquement le Vdrop (qui n'est pas le Vdroop car celui ci est un choix délibéré d'Intel)

Il y a deux types de "drops" totalement différents :

-le "vdrop", qui est un voltage plus bas (généralement lu sous windows par un logiciel de monitoring) que celui sélectionné dans le bios, est différent pour chaque type de carte mère ;

-le "vdroop" est une spécification d'Intel qui a pour effet de modifier le voltage à la baisse lorsque le CPU est en charge ; cette différence de voltage est conçue pour protéger le CPU lorsqu'il passe d'un état de charge à celui de repos (idle).

La fonctionnalité cachée du bios de la GA-X38-DQ6 agit directement et seulement sur le vdrop.

Plus d'informations dans cet article :
AnandTech: Overclocking Intel's New 45nm QX9650: The Rules Have Changed

Basic translation (pardon my basic translation talents):

---------------
Since BIOS F6 (any version), along with F7 and now the F8 series, those two options were included (only seen when pressing Ctrl + F1 in the main BIOS menu to unveil hidden options in sub-menus such as M.I.B). So far it seems that those two options have the purpose of controlling the Vdrop by augmenting the CPU voltage when under load, which in turn helps drastically decrease the actual Vdrop. And we're indeed talking about Vdrop here, not Vdroop which is quite different (a deliberate choice from Intel).

Vdrop is a lower voltage (generally read under Windows by a monitoring software) then the one selected in the BIOS, and it is different for each types of motherboards.

Vdroop is a specification by Intel which modifies the voltages (usually lowering it) when the CPU is under load; that voltage difference is made to protect the CPU when it goes from an "under heavy load" status to an idle status.

The hidden features of the GA-X38-DQ6 only and directly act upon the Vdrop.

For more information on the subject visit link (article) at: AnandTech: Overclocking Intel's New 45nm QX9650: The Rules Have Changed
---------------

Also, according to what I've read about it, the only thing that has changed with BIOS F8C and F8D with those two features is that the first one, namely CPU GTLREF1 received a new default setting and is now set to Auto with the Optimized Defaults. But from the various posts I've read from that discussion forum there were reports that it should be set to Normal instead of Auto to ensure maximum stability, and to use the Auto setting only if experiencing (after all necessary stability tests) instability issues that wouldn't fit the usual "stable" settings (in other words, if your settings should be stable but aren't, then try to set it to Auto instead, like the new default suggests).

That is what I've found about those two hidden features. I myself am currently setting them both to Normal and everything, so far, looks absolutely stable.

[supershanks]

RE the Vtt; so with my FSB Overvoltage Control set at +0.20v adding that to 1.20v I should have another 0.15v that I can play with safely without going past the intel 1.55v sepcification?

Well that's my reading of it yes

It might be worth revisiting my 8x465 settings 8x465 Stable with the additional knowkledge we now have.

I don't think that there's a wall there though I've not run the later bios.
Just noticed

(G)MCH OverVoltage Control ................. [+0.125]

looking back on my settings I used +0.175v almost always, certainly on 9x400 , so upping that might help as well ??

What a team uncovering the bios secrets great find Zenoth.
I saw that article the other week but missed that little gem.

So that really only leaves the reference voltages to sort out ??

[Cosmic]

Interesting... I only updated to F8D last night and after I did I found that my CPU GTLREF1 Voltage Ratio was set to [-3.00%]. So I'm assuming that under F7 its auto or default setting was -3.00% and has changed to something else now under F8D.

BTW great info there Zenoth Good to know we have someone who can read the french sites on the forum

Cosmic

[Cosmic]

HOLY MACKREL !!! ... It was a FSB Hole !!! (I'm running at 3.7GHz now !) ... wouldnt boot between 3.58 and 3.66 but its booting at 3.7 !!! WOOT!!!

Well a whole lot more playing now needs to be done ... Info soon ....