• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Sandy Bridge - My findings regarding memory and the processor

Associate
Joined
25 Nov 2002
Posts
2,218
Location
Somerset
My Sandy Bridge setup has been one of the most successful I have seen on the forum and I spent some time this weekend trying to work out why and where my limitations are.

I currently have my setup able to achieve the following:

1. 5.4GHz - Stable enough to benchmark, but not run prime95
2. 5.0Ghz - Prime95 stable (1.4v)
3. 4.8GHz - Fully stable at 1.35v - so probable 24/7 setup

I am using the Asus P8P67 Pro motherboard and a 2600K processor.

Firstly, my CPU hits a wall at 4.8GHz without the 1053 bios and the CPU PLL Overvoltage set to enabled. Then overclocking to 5.4GHz relies on Vcore and very minor tweaks to PLL voltage. Nothing else in the bios including LLC seems to make any real difference.

What is really interesting are my findings regarding RAM.
I have two sets of RAM, but have only achieved 5GHz + on the CPU using one of those sets and more interestingly that set at a single ratio and timings.

When my OCZ 1600 6-8-6-24 1T @ 1.425v ram is used with the bios manually set to 1600MHz, Vdimm set to 1.425 and timings set as stated overclocking to 5.4GHz seems very easy. I cant get anywhere at 100x55, 100x54 is the max.

My ram passes 3 passes of memtest at 1866MHz 6-8-6-24 1T at 1.575v but at the 1866ram ratio stability when overclocking the CPU above 5GHz disappears.

What is more interesting is that I get instability at any overclock above 4.5Ghz if I alter almost any RAM settings. Most notably even changing the ratio down to 1333, 1066 or even 800 with all the other settings fixed as above I get massive instability, often resulting in the board even booting.

Using my other set of RAM, Patriot 2400 9-11-9-27 1T @1.65v I get no real stability when overclocking. I also suffer the cold boot issue discussed on the board and also suffered the RED Ram light when first changing the ram resulting in needing to start with one stick, set the Bios manually and then add the second stick.

So my conclusion is that the CPU stability, at least for me and using the Asus P8P67 motherboard, is very dependent upon the RAM ratio and settings. This maybe because the Motherboard is setting other RAM timings behind the scenes badly or it maybe that the CPU memory controller is very fussy.

Anyway, for anyone struggling to overclock or get stability with their Sandy Bridge I recommend the following...

  • Enable CPU PLL Overclocking in the Bios.
  • Try your ram at different Speeds - You may need to find the sweet spot for the relationship between your CPU and Ram.

Hope this helps.

UPDATE: 18 JAN 2011
I have been looking into my findings some more and have now discovered that I can get stability with my OCZ ram at 1866MHz
The Asus board has a massive amount of memory settings in addition to the standard CAS etc. I have made a note of these settings at 1600MHz and applied them manually at 1866MHz.
When I did this first time, I manually set them all... I have now discovered that there are a few that make the system very unstable.

With the Ram settings manually set I now have as much stability as I did at 1600MHz

So in line with my findings yesterday, I can confirm that it is the Ram settings that the Asus board chooses that makes things less stable.

My next step is to explore these settings further and see if I can find which aid or reduce stability.
 
Last edited:
Thanks for your work Simon - how are you setting all your other voltages? (VCCSA, VCCIO, CPU PLL).

I'm always unsure what to do with these. People seem to tend to post only their Vcore/ram - is it an assumption that the rest should be on auto? Isn't that asking for trouble if the board tries to ramp them up too far?

I'm running an i5 2500K, 8GB Corsair Vengeance 8-8-8-24 @ 1.5V, P8P67 Pro. Currently stable at 4.4, but I want more out of it (4.6 minimum really).

My current settings are:

Load Line Calibration: Extreme
VRM Frequency: Auto
VRM Spread Spectrum: Disabled
Phase Control: Standard
Duty Control: T. Probe

CPU Voltage - 1.30
DRAM Voltage - 1.50
VCCSA Voltage -0.92500
VCCIO Voltage - 1.05000
CPU PLL Voltage - 1.80
PCH Voltage - Auto
CPU Spread Spectrum - Auto

I'm running BIOS 1201, so don't have access to the enable CPU PLL overclocking setting yet.

I've tried 4.5GHz at 1.32 vcore, and the above settings, but it BSOD'd after a while in prime.

Do you have any pointers? Thanks.
 
I have noticed the cpus like PLL i set mine to low, and at 1.28v in the bios PLL provides upto 1.31 when under load, only seems to pull what it needs though, be nice if in the future cpu can fully auto adjust the volts.
 
That is really interesting, I have heard others say simalar things and wonder if it is only with UEFI bios based motherboard that this problem is happening.

From my point of view I am not having any ram issues at all. My CPU walls (hard wall) at 5.288GHz regardless of cooling method.

I have run my Geil ram at 2133MHz at that CPU frequency. WPrime 1024 stable, and all 3D benchmarks run perfectly including 3DMarkVantage performance - and that surprised me.

With my EVO2 ram I can run 6-9-6-23 again for all benchmarks at 1866 right up to 1998 not willing to set my BLCK above 1060 to achieve a higher ram frequency. I have to juggle my settings to stay this side of 5.288 even one MHz over and no boot to windows at all.

Same with my Corsair GT ram 8-9-8-24 on XMP profile all day long. Have not tested with a new kit of GT ram 7-8-7 yet... but hope to do some soon.

Basically no problems what so ever from day one with my set up.

Gigabyte UD7 running F7a offical bios.

Settings are pretty much the same as yours. Except I drop my Pll just a tab below the auto settings.
 
Thanks for your work Simon - how are you setting all your other voltages? (VCCSA, VCCIO, CPU PLL).

I'm always unsure what to do with these. People seem to tend to post only their Vcore/ram - is it an assumption that the rest should be on auto? Isn't that asking for trouble if the board tries to ramp them up too far?

I'm running an i5 2500K, 8GB Corsair Vengeance 8-8-8-24 @ 1.5V, P8P67 Pro. Currently stable at 4.4, but I want more out of it (4.6 minimum really).

My current settings are:

Load Line Calibration: Extreme
VRM Frequency: Auto
VRM Spread Spectrum: Disabled
Phase Control: Standard
Duty Control: T. Probe

CPU Voltage - 1.30
DRAM Voltage - 1.50
VCCSA Voltage -0.92500
VCCIO Voltage - 1.05000
CPU PLL Voltage - 1.80
PCH Voltage - Auto
CPU Spread Spectrum - Auto

I'm running BIOS 1201, so don't have access to the enable CPU PLL overclocking setting yet.

I've tried 4.5GHz at 1.32 vcore, and the above settings, but it BSOD'd after a while in prime.

Do you have any pointers? Thanks.

I don't recognise BIOS 1201?
I found that everything on AUTO voltage wise to get to 4.8GHz (but i'm 2600K not 2500K). I only needed to increase voltages other than vcore over 4.8GHz.

For best results I set this
MULTI - AUTO
LLC - AUTO

CPU PLL Overvoltage: ENABLED (Must have for over 4.8 for me will be less for 2500K)

CPU Voltage - 1.185v
DRAM Voltage - 1.425 (but this is based on your own ram)
VCCSA Voltage -AUTO
VCCIO Voltage - AUTO (Causes me instability if changed)
CPU PLL Voltage - AUTO
PCH Voltage - Auto
CPU Spread Spectrum - Auto

I then use the Asus windows software to do the overclock
Just step up the vcore and the Multi and run prime95 at each step... I found if it passed 15 minutes prime it is there or thereabouts ok.

I have only altered Vcore, VCCSA (+ 2 clicks) and CPU PLL (+3 clicks) for 5GHz.

I advise changing voltages then apply... then change the Turbo multi and apply again.
 
That is really interesting, I have heard others say simalar things and wonder if it is only with UEFI bios based motherboard that this problem is happening.

From my point of view I am not having any ram issues at all. My CPU walls (hard wall) at 5.288GHz regardless of cooling method.

I have run my Geil ram at 2133MHz at that CPU frequency. WPrime 1024 stable, and all 3D benchmarks run perfectly including 3DMarkVantage performance - and that surprised me.

With my EVO2 ram I can run 6-9-6-23 again for all benchmarks at 1866 right up to 1998 not willing to set my BLCK above 1060 to achieve a higher ram frequency. I have to juggle my settings to stay this side of 5.288 even one MHz over and no boot to windows at all.

Same with my Corsair GT ram 8-9-8-24 on XMP profile all day long. Have not tested with a new kit of GT ram 7-8-7 yet... but hope to do some soon.

Basically no problems what so ever from day one with my set up.

Gigabyte UD7 running F7a offical bios.

Settings are pretty much the same as yours. Except I drop my Pll just a tab below the auto settings.

Have to say it does feel, from the Asus P8P67 issues as if the Asus does some RAM stuff that we can't see and it maybe that causing issues.

I find your comment interesting about a CPU Hard Wall... I'm similar at 5.4GHz, if i touch BCLK (up or down) or try and apply 100x55 I get nowhere. Once I overcame the wall at 4.8GHz every increase in multiplier had some kind of stability, meaning that it changed to it but only lasted a short while before BSOD if the voltage was not high enough. 55 multi is different and just causes the screen to freeze. I'm not a believer in 'Hard Wall', I think there is always something that can be done to get further. That said, I have now stopped trying. I'm quite happy with some killer benchmarks at 5.4 and the ability to do anything at 5GHz, if i choose to. Last night I wiped the hard drive and have put on Raid 0 and installed a nice clean XP Pro x64 and Windows 7 Ultimate x64 - Tonight i will actually put on some games and start using the thing LOL
 
Last edited:
I don't recognise BIOS 1201?

Sorry, that was me being dim and reading the wrong number - it was whatever the board came with basically. I've flashed to 1053 now.

I found that everything on AUTO voltage wise to get to 4.8GHz (but i'm 2600K not 2500K). I only needed to increase voltages other than vcore over 4.8GHz.

For best results I set this
MULTI - AUTO
LLC - AUTO

CPU PLL Overvoltage: ENABLED (Must have for over 4.8 for me will be less for 2500K)

CPU Voltage - 1.185v
DRAM Voltage - 1.425 (but this is based on your own ram)
VCCSA Voltage -AUTO
VCCIO Voltage - AUTO (Causes me instability if changed)
CPU PLL Voltage - AUTO
PCH Voltage - Auto
CPU Spread Spectrum - Auto

Thanks for that. I assume it's safe therefore to leave them all on AUTO? Apart from CPU/DRAM obviously.

What about all these settings?

VRM Frequency
VRM Spread Spectrum
Phase Control
Duty Control
 
Last edited:
Sorry, that was me being dim and reading the wrong number - it was whatever the board came with basically. I've flashed to 1053 now.



Thanks for that. I assume it's safe therefore to leave them all on AUTO? Apart from CPU/DRAM obviously.

What about all these settings?

VRM Frequency
VRM Spread Spectrum
Phase Control
Duty Control

I started setting those to EXTREME, but have now tested and found that AUTO / DEFAULT made no difference to my stability so they are back their default settings.
 
Simon I hear you with hard wall. I am hoping that a BIOS fix will do something good. But in this instance I have other 2600K CPU's that I was allowed to use. And of the 8 CU's I played with three walled at 54X multi 5.377 and one would not get past 5GHz at all. The others were about the same as mine. One CPU just blew past 55x multi and was snatched out of my rig before I could even react. And he calls himself a mate. :D

I think it is a BIOS issue and one that Intel needs to assist the vendors in finding a work round. I just find it incredible that I can bench Vantage at my max frequency but not up the frequency by one MHz. And that is incredible as in dismayed. Why I say that because on my 980X set up my Vantage benching frequency is always about 300-400MHz less than my other 3D benchmark frequency. So 5.4 for Vantage and 5.8 for 3Dmark 06 say.

Come on Intel share the goodness and let the vendors unlock the awesome potential of these CPU's :)
 
Just enabled it and its booted into windows ok with ram set to 1600 ill run prime and report back.My ram is not on the qvl list in the manual and if i select xmp it sets the timings wrong so i set it all manually.
 
It's BSOD'ing in P95 at 4.5GHz:

Internal PLL Overvoltage: Enabled

Load Line Calibration: Extreme
VRM Frequency: Auto
VRM Spread Spectrum: Disabled
Phase Control: Standard
Duty Control: T. Probe

CPU Voltage - 1.32v
DRAM Voltage - 1.5
VCCSA Voltage -AUTO
VCCIO Voltage - AUTO
CPU PLL Voltage - AUTO
PCH Voltage - AUTO
CPU Spread Spectrum - AUTO

What options do I have other than just upping Vcore? With LLC on it's pushing 1.34V in CPU-Z
 
Just enabled it and its booted into windows ok with ram set to 1600 ill run prime and report back.My ram is not on the qvl list in the manual and if i select xmp it sets the timings wrong so i set it all manually.

My Patriot is on the QVL and it does not work well at all... my OCZ is not on the list and works brilliantly but only at 1600
 
It's BSOD'ing in P95 at 4.5GHz:

Internal PLL Overvoltage: Enabled

Load Line Calibration: Extreme
VRM Frequency: Auto
VRM Spread Spectrum: Disabled
Phase Control: Standard
Duty Control: T. Probe

CPU Voltage - 1.32v
DRAM Voltage - 1.5
VCCSA Voltage -AUTO
VCCIO Voltage - AUTO
CPU PLL Voltage - AUTO
PCH Voltage - AUTO
CPU Spread Spectrum - AUTO

What options do I have other than just upping Vcore? With LLC on it's pushing 1.34V in CPU-Z

Have you...
a) tested you ram using memtest4.2 via a bootCD
b) Try setting LLC to auto rather than EXTREME
c) VRM Spread Spectrum: disabled.... not by my machine, but is there not an Auto?

Here is what I did have set for 5GHz... but as I say i switched most back to Auto.
P8P67BiosFor5Ghz-voltages.jpg
 
WOW I never realised how lucky I am that all the ram I have has just worked with my Giga UD7 - have not even looked at the QVL - think I am too scared to. Might jinx things :D
 
b) Try setting LLC to auto rather than EXTREME
c) VRM Spread Spectrum: disabled.... not by my

It BSOD's even earlier with LLC auto.

VRM Spread Spectrum is either disabled or enabled.

Also, the last few times I've made changes in BIOS (or after a BSOD), I've had to go through the following:

Save BIOS settings (or after BSOD).
It turns itself off, then turns on again.
Blank screen.
I turn it off, then turn it on again.
It turns itself off, then turns on again.
I get the Overclocking Failed message.
I turn it off, then turn it on again.
It turns itself off, then turns on again.
Then I get into windows.

I take it that's not normal behaviour :p
 
It BSOD's even earlier with LLC auto.

VRM Spread Spectrum is either disabled or enabled.

Also, the last few times I've made changes in BIOS (or after a BSOD), I've had to go through the following:

Save BIOS settings (or after BSOD).
It turns itself off, then turns on again.
Blank screen.
I turn it off, then turn it on again.
It turns itself off, then turns on again.
I get the Overclocking Failed message.
I turn it off, then turn it on again.
It turns itself off, then turns on again.
Then I get into windows.

I take it that's not normal behaviour :p

I get this a lot when altering things, but it is normally when the RAM is not working in sync very well.
 
I get this a lot when altering things, but it is normally when the RAM is not working in sync very well.

Is there any way to fix that, or should I just live with it until I find stable settings?

What I'm after is a stable 24/7 overclock, as high as is safe, for running FSX. I work for a company producing video products for the sim, and it eats CPU like you wouldn't believe! I ran it stable at 4.4 which was VERY smooth, however more would be nice, as there was a big difference between 4.0 and 4.4.
 
Back
Top Bottom