***Asus Rampage III Extreme Owners Thread***

[OC Nub]

I think what is confusing is that on other boards, like EVGA or Gigabyte raising QPI PLL is needed to achieve high bclk. My UD7 or Classified 3 will not run stable at high bclk with QPI PLL set to low. The MSI XPower doesnt have a setting for QPI PLL but it doesnt seem to effect bclk when using fast mode. I think it goes back what was posted many pages back, if you get a board with a good chipset you will be able to do higher bclk, if not your bclk limit will be much lower in fast mode..

[donmarkoni]

IMO, zoson's theory is most probably right, if not absolutely.

So everyone knows... In case it wasn't common knowledge.

QPI PLL is linked to the CPU PLL setting in BIOS on the R3E.

I asked how it was controlled to ASUS techsupport and that's the answer I got.

I've asked them for a table of values showing what QPI PLL voltage is set per different CPU PLL settings.

It seems 1.89v CPU PLL is probably between 1.14 and 1.16v QPI PLL. Raising my CPU PLL to this value gives me about the same stability/voltage settings that I got on my friend's x58a-ud3r with my chip.

I think that QPI PLL loops 2-3 times from lowest to highest in one loop of CPU PLL voltage from 1.2V to 2.1V.
One thing is sure, QPI PLL is 1.1V when CPU PLL is 1.8V and probably when CPU PLL is lowest 1.2V QPI PLL is 1.1V or even 1.0V.

Could be something like this:
CPU PLL | 1.2 | 1.3 | 1.4 | 1.5 | 1.6 | 1.7 | 1.8 | 1.9 | 2.0 | 2.1 |
QPI PLL | 1.0 | 1.1 | 1.2 | 1.3 | 1.4 | 1.0 | 1.1 | 1.2 | 1.3 | 1.4 |

or maybe not so linear, IDK

[AceBaran]

Tnx Donmarkoni for your guidance

More stable now @ lower QPI and volts

[zoson]

I'm getting a i7 970! Maybe I'll do better with 32nm.

[Utnorris]

With 32nm chips, whether it was the 980x or the 970 I have had better success with dropping CPU PLL from 1.8v to 1.42v. I couldn't get my 980x stable at 4.2ghz until I did this.

[[200]]

IMO, zoson's theory is most probably right, if not absolutely.



I think that QPI PLL loops 2-3 times from lowest to highest in one loop of CPU PLL voltage from 1.2V to 2.1V.
One thing is sure, QPI PLL is 1.1V when CPU PLL is 1.8V and probably when CPU PLL is lowest 1.2V QPI PLL is 1.1V or even 1.0V.

Could be something like this:
CPU PLL | 1.2 | 1.3 | 1.4 | 1.5 | 1.6 | 1.7 | 1.8 | 1.9 | 2.0 | 2.1 |
QPI PLL | 1.0 | 1.1 | 1.2 | 1.3 | 1.4 | 1.0 | 1.1 | 1.2 | 1.3 | 1.4 |

or maybe not so linear, IDK

A read point would be handy hey.

But having said that, I read the Gigabyte review posted in the reviews section on the forums (funny that). The reviewer stated,

Natively the X58 IOH provides 36 lanes for the pci-e bandwidth. Not only do video cards operate off these lanes, but so does the integrated USB3, SATA6G controller, as well as many other peripherals. A brief history of the X58 chipset will reveal that there are three revisions of the chipset, the most recently being revision B3/13, before this revision the main one that hit markets was B2/12, it seems that revision B1/11 was never released to the public. Older boards that have rev12/B2 chipsets seem to have a hard time reaching high base clock frequencies, but base clock is only partially dependent on the chipset.

I just quoted the whole paragraph for context and put in bold the relevant information. Perhaps the chipset is old stock if I can use that term loosely. I know other boards have better results than others which might indicate the newer revision chipset.

And here is a read point for that particular board,



I hope I am not breaking anyones balls by posting this but it would be good to get down to source of the problem.

Maybe those that can post a high bclk can take a pic of their IOH core to display the markings for reference?

Maybe i'll kill my board and hope I get a better one back?

[Dahaka]

Good bench session from last night, i use my win7 24-7 and no tweaks for the moment.





With 3DMark i have problems, my 950 isn't so good like my "oldie" 930 and requiere to much voltage for 4.5GHz, forget the QPI VTT that i'm using for that clocks...it's almost ridiculous.

Regards!

[zoson]

Wow! i7 6 core heatsinks are pretty nice! A little on the small side though.
My 970 just got here. Batch 3016B246.
Does 32nm setting UCLK to 1.5x make overclocking much easier with 12GB 2GHz mem?

[gmod]

Can anybody be so kind to give me a starting point on this board to run my 920 DO which i had on my classified (RIP) 760 at 4ghz...This is my first Asus board so the bios are a little strange...Iam running Gskill perfect Storm 2000 6gigs..


Iam running 1005 bios which it came with
I do remember I only required 1.3v vcore on the classified

[zoson]

Another question.
I used this guide to bring my overclocking knowledge up to spec from core2 to i7:
http://www.techpowerup.com/forums/sh...d.php?t=108526
They said that 32nm chips show damage above 1.35v QPI/VTT. I've been using 1.4875v QPI/VTT with my 930 and 950... Do the 32nm chips need less vQPI because of the smaller manufacturing process?
What should I know as far as the voltage limit differences between 45nm and 32nm?
Thanks!

[zoson]

Nevermind...
Wow, it's so easy...

http://valid.canardpc.com/show_oc.php?id=1552465
1.331v core
1.343v QPI
1.656v DDR
1.815v CPU PLL
1.166v IOH
1.113v ICH

[DrSlump]

Hi guys, may you help me overclocking my cpu?
Actually i've managed to overclock my i7-950 @ 3800Mhz stable in windows7.

My actual settings:
bclk=200Mhz
Pcie=101Mhz
Qpi=3200Mhz
Ram=1600Mhz 7-8-7-20-1N
Vcpu=1.25V
Vqpi=1.25V

All other settings to auto.
If i try to set the cpu multiplier to 20x, i must set Vcpu=1.35V for the windows login, but the system is unable to pass any stress test..

[Rasmus66]

Hi guys, may you help me overclocking my cpu?
Actually i've managed to overclock my i7-950 @ 3800Mhz stable in windows7.

My actual settings:
bclk=200Mhz
Pcie=101Mhz
Qpi=3200Mhz
Ram=1600Mhz 7-8-7-20-1N
Vcpu=1.25V
Vqpi=1.25V

All other settings to auto.
If i try to set the cpu multiplier to 20x, i must set Vcpu=1.35V for the windows login, but the system is unable to pass any stress test..

Try with vqpi=1,35 and IOH 1,19 volt

[donmarkoni]

Wow! i7 6 core heatsinks are pretty nice! A little on the small side though.
My 970 just got here. Batch 3016B246.
Does 32nm setting UCLK to 1.5x make overclocking much easier with 12GB 2GHz mem?

UnCore to RAM ratio of 1.5x does make it easier, but in my experience needs more VTT (QPI/DRAM) voltage then >1.5x and tend to show random instability issues. So, try to use at least one more step over the lowest.
And of course, higher UnCore speed = higher memory bandwidth.

Another question.
I used this guide to bring my overclocking knowledge up to spec from core2 to i7:
http://www.techpowerup.com/forums/sh...d.php?t=108526
They said that 32nm chips show damage above 1.35v QPI/VTT. I've been using 1.4875v QPI/VTT with my 930 and 950... Do the 32nm chips need less vQPI because of the smaller manufacturing process?
What should I know as far as the voltage limit differences between 45nm and 32nm?
Thanks!

I see you got a grip of things pretty fast, but just take care about voltages/temperature. It's not the same using 1.6V vCore with 45nm compared to 32nm when pushing hard on air/water (what I do )
Don't get offended or something, I don't take you for ignorant (on the contrary), just a friendly advice to not get carried away. There are many dead Gulftowns reported here on XS, but just a few Bloomfields.
May your CPU live long and prosperous.

Can anybody be so kind to give me a starting point on this board to run my 920 DO which i had on my classified (RIP) 760 at 4ghz...This is my first Asus board so the bios are a little strange...Iam running Gskill perfect Storm 2000 6gigs..

Iam running 1005 bios which it came with
I do remember I only required 1.3v vcore on the classified

Why don't you just start from same settings you have been using, then if it gets hard with new BIOS type, just post settings using template and someone will help for sure.
And then, there is a thrill of overclocking on new MB, why would you skip that?

Hi guys, may you help me overclocking my cpu?
Actually i've managed to overclock my i7-950 @ 3800Mhz stable in windows7.

My actual settings:
bclk=200Mhz
Pcie=101Mhz
Qpi=3200Mhz
Ram=1600Mhz 7-8-7-20-1N
Vcpu=1.25V
Vqpi=1.25V

All other settings to auto.
If i try to set the cpu multiplier to 20x, i must set Vcpu=1.35V for the windows login, but the system is unable to pass any stress test..

Just try using 21x CPU multi, you might get pleasant surprise.
Try higher vCore (1.35-1.4V) and slightly higher VTT (QPI/DRAM) (~1.3V) to get it stable, then later you can lower it a bit if possible.

[spoolindsm127]

Just got myself a W3570 Batch# 3846A803

It's a Christmas Eve miracle

EDIT: zoson looking good brother Can't wait till the day comes when I own a hexcore lol
Also are you replying using a mobile phone by any chance? Might explain the double posts..

[DrSlump]

Just try using 21x CPU multi, you might get pleasant surprise.
Try higher vCore (1.35-1.4V) and slightly higher VTT (QPI/DRAM) (~1.3V) to get it stable, then later you can lower it a bit if possible.

But, if i use x21 multiplier i would raise the speed to 4200Mhz. Are you sure i should raise to 21?
The main problem is that with such voltages, my temps raises a lot. If i try a linx test with my cpu @ 4Ghz, have a rapid raise even to 91°C

P.s. i'm using a corsair h70 as heatsink

[tistou77]

Nevermind...
Wow, it's so easy...

http://valid.canardpc.com/show_oc.php?id=1552465
1.331v core
1.343v QPI
1.656v DDR
1.815v CPU PLL
1.166v IOH
1.113v ICH

by curiosity, you are how many GFlops with your 970 at this frequency (with LinX)?

[tistou77]

this new 970 (and R3E, socket LOTES) is really strange

with my old 970, to be stable and 4.25GhZ and 2025MhZ for memory (Uncore 3645), it was

Vcore: 1.337v
vQPI: 1.30v

with the new, I am stable for the same frequency with:

Vcore: 1.311v
vQPI: 1.26v
cpu-pll: 1.81v

I have not yet tested under voltage ...
I tested bios 0011, 0095 and 1102 and it's the same thing

this new 970 heats a little more (~6° higher than the other) but requires less voltage

[tistou77]

I have a question,

I'm not the NB LED that lights on the motherboard (those for DDR, CPU and SB are OK)
and the DDR LED is yellow (crazy) while VDIMM is 1.65V

with bios 1102

why?

thank you

EDIT: for LED NB, it must be hidden by the Koolance waterblock on the motherboard, I think

[zoson]

UnCore to RAM ratio of 1.5x does make it easier, but in my experience needs more VTT (QPI/DRAM) voltage then >1.5x and tend to show random instability issues. So, try to use at least one more step over the lowest.
And of course, higher UnCore speed = higher memory bandwidth.

Thanks for the info! I will up the UnCore ratio x1 and re-test!

I see you got a grip of things pretty fast, but just take care about voltages/temperature. It's not the same using 1.6V vCore with 45nm compared to 32nm when pushing hard on air/water (what I do )
Don't get offended or something, I don't take you for ignorant (on the contrary), just a friendly advice to not get carried away. There are many dead Gulftowns reported here on XS, but just a few Bloomfields.
May your CPU live long and prosperous.

I've been overclocking for many years! However, it is a completely new game from Core2/AthlonXP days. I feel fairly safe with the volts I'm using right now. I used the info from this thread to set up my BIOS! Very good info.

Just got myself a W3570 Batch# 3846A803

It's a Christmas Eve miracle

EDIT: zoson looking good brother Can't wait till the day comes when I own a hexcore lol
Also are you replying using a mobile phone by any chance? Might explain the double posts..

I got a friend who works retail to sign up for Intel Retail Edge so I could buy the i7 970 for $215!!!! There's still time if you know someone!!!

I'm posting with Chrome... Not sure what's up with the double posts... It usually doesn't do it.

by curiosity, you are how many GFlops with your 970 at this frequency (with LinX)?

82GFlops I have a bunch of services running through, so maybe a little bit more.

[fl0w3n]

IMO, zoson's theory is most probably right, if not absolutely.



I think that QPI PLL loops 2-3 times from lowest to highest in one loop of CPU PLL voltage from 1.2V to 2.1V.
One thing is sure, QPI PLL is 1.1V when CPU PLL is 1.8V and probably when CPU PLL is lowest 1.2V QPI PLL is 1.1V or even 1.0V.

Could be something like this:
CPU PLL | 1.2 | 1.3 | 1.4 | 1.5 | 1.6 | 1.7 | 1.8 | 1.9 | 2.0 | 2.1 |
QPI PLL | 1.0 | 1.1 | 1.2 | 1.3 | 1.4 | 1.0 | 1.1 | 1.2 | 1.3 | 1.4 |

or maybe not so linear, IDK

This makes sense, I believe 1.6 CPU_PLL was used for 266 BCLK on my board, meaning 1.4 QPI_PLL (It's highest)

[zoson]

I think I have two lazy temp sensors... My core #2 and core #5 are 10C less at both idle and load than all others...

[PytonOrm]

Merry Christmast all friends

[tistou77]

82GFlops I have a bunch of services running through, so maybe a little bit more.

ok thanks

I have the same thing

you have the CPU with watercooling? what temperature with LinX?
and what bios do you use?

thank you very much

[gmod]

Well got it going and actually turned out with a better OC than the Classified 760...Now got my 920 @ 4418 210 x 21...1.34vcore....
Score 41174 in vantage with my 2 480's...NEEED to get my waterblocks for them SOOON..

That dam EK Chipset block was a pain to install..When is EK gonna start using standoffs???
But it does look good on the board..