ASUS Maximus II Formula - new P45 king?

[SiGfever]

1901 offers nothing more than 1802 unfortunately
Its a stable bios but zero improvment

Do you recommend the 1802 bios over 1307?

[GeorgeVasil]

I've tried 1802 and i had unexpected hungs on restart ..(LCD Poster reports DET RAM) at the 1307's rock stable settings.

So the best BIOS for me is 1307..

[Grnfinger]

Do you recommend the 1802 bios over 1307?

I honestly find no difference with 1307, 1802 and 1901, I have 1802 on bios 1and 1901 on bios 2 on both boards.
Both chips run well with either bios

I've tried 1802 and i had unexpected hungs on restart ..(LCD Poster reports DET RAM) at the 1307's rock stable settings.

So the best BIOS for me is 1307..

Your not trying to load saved profiles are you?
If you flash to another bios always enter settings manually.

[JoeBar]

I honestly find no difference with 1307, 1802

Me too. Currently running 1802.

[andressergio]

Guys

I tested 1307 / 1802 and 1901 and vista 64 feels smoother and much more responsive than all before, i also testing memtest on the gskills at this settings:

Q9650 4080 (480x8,5) vcore idle 1,336 / load 1,3458 / vtt 1,3 / pll 1,53
CPU GTLS +10/-40/+10/-40
NB GTL +40
CPU delay 200 / NB delay 100
DRAM skews delay 100
Gskills PC29600 2x2GB 1200Mhz 5-5-4-15 vdimm 2,1 NB1,45 PL8 Ai Twister Moderate

Cheers
Sergio

[GeorgeVasil]

Your not trying to load saved profiles are you?
If you flash to another bios always enter settings manually.

No,No..

1307 is on BIOS1 Chip & 1802 on BIOS2 Chip.

My settings are manually entered.

[Gunnz Akimbo]

Chucked the 1901bios on.

Still the same performance, can't see any difference.
At least it keeps the speedingHDD raid drives alone, didnt mess that up phew!

Ive tried those ram settings before, but the thing didnt boot with bios 1307. its only value set of 8GB at DDR2-800. I can't be bothered with cmos resets, there's too many options to change all the time.

Just one thing, with 400FSB, what's the adequate voltage for

FSB
NB
SB 1.5
SB 1.1

Just take into account there is 8GB installed.
Currently its set at

FSB = 1.36xxx
NB = 1.36xxx
SB = 1.66xxx
SB = 1.36xxx
Running kinda warm though.

[CryptiK]

Well following up on my Crysis instability, it was due to my NB GTL Ref. Now, set at -80mv, it's completely stable in crysis, and other stress tests (IBT/Orthos/Memtest). If you are experiencing any similar issues it may be the setting you need to adjust to solve it.

[andressergio]

Well following up on my Crysis instability, it was due to my NB GTL Ref. Now, set at -80mv, it's completely stable in crysis, and other stress tests (IBT/Orthos/Memtest). If you are experiencing any similar issues it may be the setting you need to adjust to solve it.

Great to hear that CryptiK What NB volts you use ? I'm on +40 but maybe negative values are better, they seem to do good to my cpu on this mobo

will try

Did you see i can run the gskills at 1200 5-5-4-15 2,1vdimm fails on LinX 10 test but it's good news...I will try dif NB and GTL, i'm not sure if it's good idea to pump more than 2,1vdimm on them...

catch you later
Cheers
Sergio

[andressergio]

Chucked the 1901bios on.

Still the same performance, can't see any difference.
At least it keeps the speedingHDD raid drives alone, didnt mess that up phew!

Ive tried those ram settings before, but the thing didnt boot with bios 1307. its only value set of 8GB at DDR2-800. I can't be bothered with cmos resets, there's too many options to change all the time.

Just one thing, with 400FSB, what's the adequate voltage for

FSB
NB
SB 1.5
SB 1.1

Just take into account there is 8GB installed.
Currently its set at

FSB = 1.36xxx
NB = 1.36xxx
SB = 1.66xxx
SB = 1.36xxx
Running kinda warm though.

mate did you try dram skews ? for ex on delay 50ps to 100ps ? also NB delay 100 and NB GTL gave me better results...I'm not on 8GB just 4GB but worth trying

i'm also on bios 1901

Cheers
Sergio

[CryptiK]

Great to hear that Tim What NB volts you use ? I'm on +40 but maybe negative values are better, they seem to do good to my cpu on this mobo

will try

Did you see i can run the gskills at 1200 5-5-4-15 2,1vdimm fails on LinX 10 test but it's good news...I will try dif NB and GTL, i'm not sure if it's good idea to pump more than 2,1vdimm on them...

catch you later
Cheers
Sergio

Yeah very nice for 2 x 2GB. However I'd run ~2.00 - 2.05v max through those things as they aren't cheap. Plus dropping to 5-5-4-15 on the P45 chipset wont gain you much. I run 5-5-4-12 as I chose my desired 24/7 ram speed, and only need 1.94v to do it.

My vNB is 1.30v. Each NB chip is different just like CPU's, each will like a different GTL Ref at any given speed, mine just happens to like -ve values but others like +ve values. As with the cpu GTL's its just trial and error.

Chucked the 1901bios on.

Still the same performance, can't see any difference.
At least it keeps the speedingHDD raid drives alone, didnt mess that up phew!

Ive tried those ram settings before, but the thing didnt boot with bios 1307. its only value set of 8GB at DDR2-800. I can't be bothered with cmos resets, there's too many options to change all the time.

Just one thing, with 400FSB, what's the adequate voltage for

FSB
NB
SB 1.5
SB 1.1

Just take into account there is 8GB installed.
Currently its set at

FSB = 1.36xxx
NB = 1.36xxx
SB = 1.66xxx
SB = 1.36xxx
Running kinda warm though.

FSB = 1.36xxx --> whatever your cpu needs and you're comfortable with
NB = 1.36xxx --> 1.25 - 1.45v seems to be the normal range.
SB = 1.66xxx --> 1.5 is fine for me I've never needed to raise it
SB = 1.36xxx --> 1.2v seemed to stabilize things a little for me at high FSB/ram speeds during SPI32M runs and with gaming, although increasing to 1.3v+ didn't make any difference.

[Simps]

500FSB on 65nm quad MASTERED for 24/7 use!
DDR2-1200 5-5-5-18, PL8, STRONGER
BIOS 1901
Aircooled rig!

I am still running prime as I type. I will come back later to post full details on how to do it!

1.42v CPU
1.35v FSB
1.59v PLL
2.09v DIMM
1.44v NB
1.19v SB1.1
1.55v SB1.5



[]'s
Simps

[Grnfinger]

very nice

[WhiteFireDragon]

what memory did you use and at what vdimm?

[Simps]

very nice

Thanks! Now that 500FSB is done, I am trying to go for 24/7 508FSB on my quad. Looks doable with bios 1901. I can boot up to 530FSB now, with no extreme volts.

what memory did you use and at what vdimm?

I am using Corsair Dominator PC10000C5DF, and they are at 2.09v.

[]'s
Simps

[Furiøùs]

I sent the PM again Grnfinger.

[Simps]

I found a way to do 500FSB 24/7 stable on a quad on this board, I would like to share how I did it, and it may help everybody to reach higher FSB's too, if not 500 for quad.
I don't have a dualcore, so I am writing based on my quad experience.
I also believe, this method is the right way to fine tune your Maximus II Formula. Even if you are not going for 500FSB 24/7 on a quad, if you do the guide, you will find that you will be able to drop a lot of your volts. Just randonly setting values like GTL's, SKEWS, etc, are not the best way to find the sweet spot of your mother board.
At the end of this guide you will have no "auto" settings left on your bios. Everything is covered, and fine tunned.

***Doesn't matter if you are on a 65nm quad / dual or 45nm quad / dual. The way to fine tune your motherboard will be the same. So this guide will do for everybody.



Guide for full tunning the Maximus II Formula for high FSB / MEM / CLOCK for 24/7 use

How to start?
BIOS 1901
With your bios reseted (everything on auto), enter reasonable 24/7 volts for high FSB / mem clock like the ones below:

1.xxv CPU (you know what they are)
1.30v FSB
1.55v PLL
2.00v DIMM (Respect your ram spec)
1.45v NB (This one should be 1.45v)
1.15v SB1.1 (This one you might leave at the lowest)
1.55v SB1.5 (This one you might leave at the lowest)

Now, find the highest FSB that will boot and enter windows, and run prime blend for some minutes at least.
Keep on 333 strap if possible, and set your ram to the highest possible, adjusting the Vdimm.
Enter the 1st, 2nd and 3th ram information manually, copying all the values that "auto" settings show, replacing only the Row Refresh Recycle Time to 55 DRAM Clocks.
LEAVE IT ON PL10 - MODERATE for now.
Set your multiplier to the one that will result in the closest final CPU clock you aim for 24/7 use.
Remember, this set (fsb, multiplier and ram) must be unstable, but able to run prime for about some minutes at least.

Now the fine tune begins.

PS: The guide is cumulative, so you should keep settings before starting to test others. The guide is already on the order you should fine tune the settings.



1) CPU GTL Reference 0, 1, 2 and 3

You don't need to do any calculations.
There are only a few possible CPU GTL's combinations that makes sense from 1.20vFSB to 1.44vFSB.
If you are under 1.20vFSB, then you don't need this guide, since you are performing a "junior" overclock.
Anything higher then 1.40vFSB is a NO for 45nm and higher then 1.44vFSB is NO for 65nm. So there is no way you will be out of this range from 1.20vFSB - 1.44vFSB.
The only few values that are "right" are listed below.
If you are using any CPU GTL value different then this, then, from my calculations and experience, you are doing it wrong, and don't have the optimal values for your setttings, and that is slowing your overclock potencial down.

***As you can see below, the gaps on the CPU GTL's are constant. That is why you don't need to perform any calculation! The gaps are always going to be the same, and there are only 2 gaps that makes sense! 40mV and 45mV! No mistery about CPU GTL's.

From 1.20VTT to 1.32VTT (40mV gap)
CPU GTL 0: +100mV | +90mV | +80mV | +70mV | +60mV | +50mV | +40mV | +30mV | +20mV | +10mV
CPU GTL 1: +060mV | +50mV | +40mV | +30mV | +20mV | +10mV | +00mV | -10mV | -20mV | -30mV
CPU GTL 2: +100mV | +90mV | +80mV | +70mV | +60mV | +50mV | +40mV | +30mV | +20mV | +10mV
CPU GTL 3: +060mV | +50mV | +40mV | +30mV | +20mV | +10mV | +00mV | -10mV | -20mV | -30mV

CPU GTL 0: -060mV | -50mV | -40mV | -30mV | -20mV | -10mV | +00mV
CPU GTL 1: -100mV | -90mV | -80mV | -70mV | -60mV | -50mV | -40mV
CPU GTL 2: -060mV | -50mV | -40mV | -30mV | -20mV | -10mV | +00mV
CPU GTL 3: -100mV | -90mV | -80mV | -70mV | -60mV | -50mV | -40mV


From 1.32VTT to 1.44VTT (45mV gap)
CPU GTL 0: +100mV | +90mV | +80mV | +70mV | +60mV | +50mV | +40mV | +30mV | +20mV | +10mV
CPU GTL 1: +055mV | +45mV | +35mV | +25mV | +15mV | +05mV | -05mV | -15mV | -25mV | -35mV
CPU GTL 2: +100mV | +90mV | +80mV | +70mV | +60mV | +50mV | +40mV | +30mV | +20mV | +10mV
CPU GTL 3: +055mV | +45mV | +35mV | +25mV | +15mV | +05mV | -05mV | -15mV | -25mV | -35mV

CPU GTL 0: -060mV | -50mV | -40mV | -30mV | -20mV | -10mV | +00mV
CPU GTL 1: -105mV | -95mV | -85mV | -75mV | -65mV | -55mV | -45mV
CPU GTL 2: -060mV | -50mV | -40mV | -30mV | -20mV | -10mV | +00mV
CPU GTL 3: -105mV | -95mV | -85mV | -75mV | -65mV | -55mV | -45mV

*** The CPU GTL's in green, are, from my experience, most like the candidates of the most stable CPU GTL. But feel free to try whatever you want!

*** The positive values for the CPU GTL's, are not available on the 1901 BIOS, from 5 to 5mV increments. Only 10mV to 20mV and so on. So if you need a positive CPU GTL value like "+25mV", you will have to round it to +30mV or +20mV, or better then that, leave the 5mV to the negative CPU GTL's, as they are available on the 1901 BIOS!

*** You can select on your bios, other CPU GTL's that are beyond the +100mV and -100mV showed here. Its very unlikely that those are the best CPU GTL's for you, but you can always try them too!

You should test all the possible CPU GTL's listed above, and keep of course, the most stable one.
To do this, you should try at least 2 or 3 different unstable sets of high FSB's and CPU clocks (including the one you are at now), that are about 5-10mim prime stable only.
By testing all of them, and writing the results down (how long it last before crashing on prime), you will see a patter.
That is why you need to do it with 2 or 3 sets of unstable settings, so you can really understand the patter of the results, and pic the most stable CPU GTL.
Now, you want some sets that are 5mim prime stable only, so you don't need to wait a lot before each test. Also, you need to test sets that are close to your goal FSB / CPU clock.
That is it, now you have your perfect CPU GTL.



2) NB GTL Reference

At this time, you must have your CPU GTL's defined. Keep them.
NB GTL Reference might vary from -80mV to +80mV of even more. The best way to find the optimal one, is testing them all and finding the most stable one.
For this test, you should keep the high FSB (always), and use high DDR2 clock, with PL8 and a STRONG or STRONGER clock twister. Any thing that can put some stress on the NB.
You should prime blend again, at least 2 or 3 sets of unstable settings, and write the results down, as you change your NB GTL.
You will see a patter, and you should keep the most stable one.



3) DDR2 CHA / DDR2 CHB and NB DDR2 Reference Voltages

FOR DDR2 CHA / CHB YOU SHOULD NOT GOT BEYOND -25mV - +25mV
FOR NB DDR2 REFERENCE YOU SHOULD NOT GOT BEYOND -50mV - +50mV

So there are a few values only to be tested.
Do the same test you did on Item2, and now, with these settings, you should improve them.
Let's say you got +20mV as your NB GTL on Item2, because it was the most stable one, and it could make a 10mim in prime before crashing.
Now, you repeat that same test, but with different combinations of DDR2 CHA / DDR2 CHB and NB DDR2 Reference voltages, respecting the limits.
You will keep, of course, the best combination, and if without these settings you were getting 10 mim before crashing, now with the right combination, you will improve the time.

Start testing:
+12.5
+12.5
+25

Then:
+12.5
+12.5
-25

and keep going...



4) CPU / NB SKEWS

At this point, you should have all settings from item 1, 2 and 3 done. This is a cumulative guide, so you should keep every setting before starting to test a new one.
Ok, this is VERY easy to find. SKEWS are very sensitive, so the results will appear right away, either good or bad.
YOU NEED TO TEST THESE AS A PAIR.
From my experience, I can say that only delayed skews are good, but maybe that is not true for your system. So you should try advanced too.
Start with:

CPU SKEW - NB SKEW
normal - delay 100
delay 100 - normal
delay 100 - delay 100
delay 100 - delay 200
delay 200 - delay 100
delay 200 - delay 200

and so on...

Then add larger gaps (this is good)

CPU SKEW - NB SKEW
normal - delay 200
delay 200 - normal
delay 100 - delay 300

and so on...

Since this is very sensitive, when you start to prime, some of the settings will crash right away. Others will last only a few seconds...
Only the few good combinations will last long. And it will be very easy to define the most stable one. Keep it.



5) DRAM SKEWS

This are not as much sensitive as the CPU and NB skews, but they are also easy to set.
Just repeat the Item4 test, and try to lower your PL, or raise your mem frequency, or add something that will put more stress to the ram, like a clock twister or something.
For me, only delayed values were good, but you might try advanced too.
Keep changing them, and writing the results down, and you will be able to tell the most stable set.

Start with:
normal / normal / normal / normal

Then:
-50 / -50 / -50 / -50

Then:
+50 / +50 / +50 / +50

Then:
-100 / -100 / -100 / -100

and so on...



6) DRAM Static Read Control

Leave it disabled



7) DRAM Read Training

Leave it disabled



8) MEM OC CHARGER

Leave it enabled



9) MEM Timings (1st, 2nd and 3th information)

If you want, this is the time to fine tune it. If you have 2x1GB you can set you Row Refresh Recycle Time to 40 DRAM Clocks. Set it to 50 for 2x2GB, and so one. From my experience, keeping 5-5-5-18-3-55-6-3 on the 1st information, is the best when going for high fsb and mem clocks.



10) PULLINS

Keep them disabled, they don't work on bios 1901.



11) CPU Configuration

Disable everything there.



12) Load line calibration

Enabled



13) CPU and PCI-E Spread Spectrum

Disabled



14) PCI-E Frequency

100



Ok, at this point, you have fine tunned your motherboard. No auto settings anymore.
But it is not stable yet, far from it.
Now, you will find and understand how the voltages affect your overclock.

vCPU - Helps most raw cpu speed
vFSB - Helps at high FSB
vPLL - Helps at high FSB
vDIMM - Helps high ram clocks.
vNB - Helps high fsb and high ram clock
vSB1.1 - Helps with raid or several HDDs / optical
vSB1.5 - Helps at PCI-E frequency (when higher then 100)

Now its up to you, to find the correct volts, to keep the system stable at a high FSB and MEM clock.

***At this point, there is no need to change, again, GTLs, SKEWS, etc... They are already fine tunned. Now you just need to find the correct volts, and you are done.

***You will now, only mess with the following:

- FSB / RAM clock
- Performance Level
- AI Clock Twisters
- And all the voltages (vCPU, VTT, etc)

Keep changing only this values, one at a time, or on pairs at most. Always test them and write results down so you can understand what is going on. After some tests, you will find the correct volts.

That is it.
Now you have mastered your motherboard, and can push your cpu and mem to its limit for 24/7 use.



Here are the bios settings I am using for 500FSB DDR2-1200CL5 PL8-STRONGER, on my 65nm quad.

CPU Skew: Delay 100
NB Skew: Delay 300
DRAM Skews: Delay 50
CPU GTL's: +10 -35 +10 -35
NB GTL: +30
CHA / CHB Reference: +12.5
NB DDR Reference: -25
DRAM Static Read Control: Disabled
DRAM Read Training: Disabled
MEM OC CHARGER: Enabled
PULLINS: Disabled
PCI-E Frequency: 100
vPLL: 1.59275
vFSB: 1.37825
vNB: 1.45775
vSB1.5: 1.55300
vSB1.1: 1.20600
CPU Configuration: All disabled
Load line calibration: Enabled
CPU and PCI-E Spread Spectrum: Disbaled



quadcore 500FSB - 24/7


quadcore 490FSB - 24/7




ADDITIONAL INFO:

Maximus II Formula is probably the best non-core i7 board out there for 24/7 usage.

1) 1.25vNB is what is needed for a quad at 490FSB and DDR2-1178 PL8 STRONGER
2) 1.27vFSB is what is needed to clock a 65nm quad at 490FSB on this board for 24/7 use! 45nm values would be even lower!
3) vPLL, vSB1.1 and vSB1.5 can be at the lowest possible volts the board allows, even at 490FSB on a quad.
4) The board just don't heat up! It is insane that your NB, SB, PWM's are not going to heat up, even at the most extreme clock / volts! Idle to load delta's are arround: NB=5C, SB=3C and PWM=10C

Take a look at the FSB / MEM speed, and the voltages required to do so, all on a 65nm quad!
This is a 4h prime blend and 10 loops IBT on max stress.
Everything, CPU, MOBO, etc, is aircooled inside a closed case.

CPU Skew: Delay 100
NB Skew: Delay 300
DRAM Skews: Delay 50
CPU GTL's: +10 -35 +10 -35
NB GTL: +30
CHA / CHB Reference: +12.5
NB DDR Reference: -25
DRAM Static Read Control: Disabled
DRAM Read Training: Disabled
MEM OC CHARGER: Enabled
PULLINS: Disabled
PCI-E Frequency: 100
vPLL: 1.50000
vFSB: 1.28550
vNB: 1.25900
vSB1.5: 1.50000
vSB1.1: 1.10000
CPU Configuration: All disabled
Load line calibration: Enabled
CPU and PCI-E Spread Spectrum: Disbaled



quadcore 494FSB - 24/7




ADDITIONAL INFO (2):

500FSB and 1:1 divider (DDR2-1000 5-5-5-18 PL10)





7h prime blend + 20 IBT loops on max stress:


[]'s
Simps

[Grnfinger]

Simps you have done a very nice job here, you should be proud of your accomplishment!!

65nm quad running 500FSB ING A!!!
WELL DONE!

[andressergio]

I found a way to do 500FSB 24/7 stable on a quad on this board, I would like to share how I did it, and it may help everybody to reach higher FSB's too, if not 500 for quad.
Simps

Congrats mate nice guide great work

I think that on my personal exp it's much easier runing 2x1GB than 2x2GB to achieve those 500fsb stable, i never had dual cores always quads and when i had my Crucials old batch 2x1GB life was much easier, then i decided to go for 2x2GB sticks and things got complex. ON 2GB and XP32 will be easier than 2x2GB and Vista 64.

Cheers
Sergio

[loonym]

Simps, that is outstanding!! Thank you for your effort. I'll try this when I get time.

[andressergio]

Hey Simps

I followed your guide as i was very close on CPU GTLS (+10/-40...) for that given VTT and NB, Skews, etc

I tried this upping a notch of vcore and managed to pass LinX 10 runs on 15000 problem size, normally after this i can pass IBT 5 runs max stress




I love the bandwith of this 8,5 multi



Thanks for your guide i will keep going up

Cheers
Sergio

[CryptiK]

500FSB on 65nm quad MASTERED for 24/7 use!
DDR2-1200 5-5-5-18, PL8, STRONGER
BIOS 1901
Aircooled rig!

Very nicely done, and on 65nm too

[WhiteFireDragon]

now we just need a guide for 45nm quads since i'm assuming most people are using these chips by now. i'm guessing it would be the same, except slightly different GTL's?

[Simps]

Thanks everybody. After playing with the board for a while, I found out, that most of the settings, are interrelated. So the best way to tune the board up, if by following a logical order on setting things up. That is basically the guide I wrote. First find the CPU GTL's, then NB GTL's, and so on. But you have to test all the settings, write the results down, and then pick the most stable one. That is the most important part. No guessing or randonlly picking up values "thinking" they are the most stable ones. Just do that, get rid of all the auto settings, and you will find out that you can now drop your volts or reach higher clocks...

Hey Simps

I followed your guide as i was very close on CPU GTLS (+10/-40...) for that given VTT and NB, Skews, etc

I tried this upping a notch of vcore and managed to pass LinX 10 runs on 15000 problem size, normally after this i can pass IBT 5 runs max stress

I love the bandwith of this 8,5 multi
Thanks for your guide i will keep going up

Cheers
Sergio

Nice going there! I have good news for you.
From my experience, I can tell that you can reach at least 490FSB stable 24/7. It will be up to you, deciding if the extra volts required for the extra mem clock / cpu clock are a deal or not. But if you are at 483FSB already, 490FSB is just a matter of time. I am saying this, because before I had my 500FSB going 24/7, I was on 490FSB and it was very easy on the board. 490FSB is doable 24/7 on this board, with the same settings you have to use to stay stable on 475FSB. Keep pushing it =/

I see that you are on the +10 -35 +10 -35 CPU GTL's. That is what I use on my 500FSB
So, your chip / mobo might like the same values mine do.
This is what I use for 500FSB on my quad, you might give it a go

vFSB: 1.37825
CPU GTL's: +10 -35 +10 -35
NB GTL Reference: +30mV

Also, try vPLL arround 1.57v for your 45nm. I use 1.59v but mine is a 65nm.

EDIT: I just went to my bios and downgraded from 500FSB to 490FSB, leaving all the rest as I had it on 500FSB, so you can see the volts that I use on 500FSB. Just did this to show what you would be getting at 490FSB, 333STRAP and DDR2-1176 5-5-5-15. This is a 2h prime95 blend + 10 IBT loops on max stress and an everest mem to show the bandwidth / lantecy.

now we just need a guide for 45nm quads since i'm assuming most people are using these chips by now. i'm guessing it would be the same, except slightly different GTL's?

I am not sure about the CPU GTL's for 45nm really... But I don't see how can they be any different from the 7 possible CPU GTL's I have shown?
Can someone explain if there will be any CPU GTL different from this 7, that are possible on 45nm?

From 1.20VTT to 1.32VTT
CPU GTL Reference (0): +30mV | +20mV | +10mV
CPU GTL Reference (1): -10mV | -20mV | -30mV
CPU GTL Reference (2): +30mV | +20mV | +10mV
CPU GTL Reference (3): -10mV | -20mV | -30mV

From 1.32VTT to 1.44VTT
CPU GTL Reference (0): +40mV | +30mV | +20mV | +10mV
CPU GTL Reference (1): -05mV | -15mV | -25mv | -35mV
CPU GTL Reference (2): +40mV | +30mV | +20mV | +10mV
CPU GTL Reference (3): -05mV | -15mV | -25mv | -35mV

[]'s
Simps

[seban]

I was playing a lot with GTLs as this was the first thing for which I found a good guide for this mobo (thanks Cryptik).

And for sure I can tell that there is not such thing as one good GTLrefs for range of vtts and also there is not such thing as good one GTL for everyone.... there might be CPUs which dont mind it and they will be stable on same GTLrefs for range of VTTs but I still believe that fine tunning GTLs would just be an improvement for those CPUs.

Your guide is impressive and I am going to follow it to see what differences it will give me but to prove your theory on GTLs you would need at least more CPUs and probably more MF2.

When I was playing with GTLs I was doing exact same thing testing different GTL and writing results then uping fsb or some voltages and testing again range of GTLs - GTLs r very sensitive and very random in effects even on minimal changes at least on my board. Of course I got bracket of stable GTLrefs but performance wise and OS feeling they are different.

And finally I am using 90/50 on 1.365vtt as they r just the best for my setting (and I found them same way you suggest to look in your guide - tho I was repeating testing of them each fsb / vtt /vcpu / vnb change - this is why I can tell they r sensitive and they make hell of the difference).