Asus P8P67 overclocking

[kidsafe]

Before you do anything else, go to Asus's website and download the latest UEFI firmware plus AI Suite II software package. Install AI Suite II first, then update the UEFI firmware from within Windows. Alternatively you can copy the ROM file to a USB thumbdrive and update via the UEFI interface itself.

Almost all of the settings we need to modify are under the AI Tweaker tab, the first being AI Overclock Tuner. Change it to either "Manual" or "X.M.P." This will enable the option to adjust your BCLK frequency manually. DO NOT TOUCH this setting, it's pointless to change the base-clock or FSB on the Sandy Bridge platform. For the purposes of finding your fastest stable CPU overclock, change your Memory Frequency to "DDR3-1333" or lower. We'll change it back to whatever is appropriate later.

The next setting is what will establish your actual CPU overclock. Enable Turbo Ratio By All Cores and then set the value to "48." On the latest beta UEFI firmware there is also a Internal PLL Overvoltage option that should be enabled.

Scroll down and choose the "Ultra High" setting for Load-Line Calibration. This will help minimize Vdroop or dips in CPU voltage under full-load. Now set both Phase Control and Duty Control to "Extreme." DRAM Voltage should be whatever your RAM kit is specified to run at...don't set this above 1.65V.

Finally, CPU Voltage should be set to Offset Mode. I used "+.030V" to start. Your system should now at least be able to boot into Windows at 4.8 GHz.

At this point I usually open a trio of software apps: CPU-Z, RealTemp and Prime95. CPU-Z will report your processor's current core voltage, current clock speed, any speedstepping, memory speeds and timings. RealTemp is as the name suggests, a temperature monitor. Prime95 is a distributing computing client that will help determine if our overclock is stable. If you were able to boot into Windows, then start a Prime95 Torture Test and select the first option: Small FFTs. At this point your CPU will be running full tilt. Your CPU fans have probably spun up to max RPM and RealTemp should be reporting at least 65C on all cores if you are using air cooling.

Let Prime95 run for 30 minutes or so. Has your system crashed? If not, then you can try increasing your turbo multiplier to 49x. There is a decent chance that your particular Sandy Bridge processor is not capable of operating at such an overclock and will lock-up or BSOD when stressed under Prime95 at this point. Repeat the Prime95 testing process to see if you make it through 30m of stress-testing. Still alive? Then bump up the multiplier again...this is very unlikely as 4.8-4.9 GHz seems to be where most Sandy Bridge CPUs are hitting the wall. If you crashed at 4.9 GHz, then bump the overclock back down to 4.8 GHz. Start priming again at 4.8 GHz, but this time leave it alone for an hour. If it lasts that long without crashing, then adjust the CPU Voltage offset down by one or two ticks. Restart your Prime95 torture test and let it run for 1 hour again. Keep doing this until your system crashes, then bump the voltage up back up by a notch or two.

I am hesitant about going higher on the CPU voltage than 1.37-1.38V under load, but whether you want to push your CPU to the limits is up to you. If you are unlucky enough to have a CPU that cannot run stably at that voltage and 4.8 GHz, then I suggest dropping your overclock to 4.7 GHz and resume testing. You might be able to drop your CPU voltage a little as well.

Once you've found a combination of clock speed and CPU voltage that passes an hour of Prime95, then let it run continuously for up to 24 hours. If your system BSODs or locks up, then you'll need to bump the CPU voltage by a notch or two again. Once you get your CPU overclock 24-hour stable, then you can adjust you can go back and set your RAM to run at the manufacturer's specifications.

I should also mention that my CPU cores get up to 72-73 degrees celsius under load with a large Noctua NH-D14 cooler. If you don't have a decent cooler, I suggest backing your overclock down to 4.5 GHz max with a lower CPU Voltage offset.

EDIT: List-form for my personal settings. Anything not listed was left at Auto/Default values. You do not need to deal with stuff like VCCSA/VCCIO/CPU-PLL unless you are going for an extreme overclock.

AI Overclock Tuner = Manual
BCLK/PEG Frequency = 100
Turbo Ratio = By All Cores = 48
Internal PLL Overvoltage = Enabled
Memory Frequency = DDR3-2133
OC Tuner = Disabled
EPU Power Saving Mode = Disabled
DRAM Timing Control = 9-11-9-24-2T
Load Line Calibration = Ultra High
Phase Control = Extreme
Duty Control = Extreme
CPU Voltage = Offset Mode = +.010V
DRAM Voltage = 1.6V
VCCIO = 1.125V

[Cilraaz]

Since many of you are I building new PCs around this series of motherboard, I thought I would offer my settings for a 4.8 GHz overclock.

It should be noted that each CPU, even from within the same batch, is different and will overclock differently. Another "identical" CPU may overclock better or worse than this, especially depending upon your cooling setup.

Once you've reached a stable overclock (~24 hours of Prime95 stress testing), then you can go back and turn up your RAM to the manufacturer's specifications.

One hour of OCCT's Linpack test will also show whether the setup is stable. In fact, it will usually find instability that Prime95 won't.

I should also mention that my CPU cores get up to 72-73C temps while priming with a large Noctua NH-D14 cooler. If you don't have a decent cooler, I suggest backing your overclock down to 4.5 GHz max with a lower CPU Voltage offset.

Note that tCase on the i5 2500K and i7 2600K is 72.6°C. This puts maximum core temp around 83°C. Usually, I try to keep my maximum core temperature around tCase, as this allows for fluctuations in ambient temperature.

[dfarrall]

Shamelessly stolen from OCUK

Sandybridge maximum safe voltages

Core Voltage - Not recommended too exceed 1.38v, doing so could kill the CPU, we therefor recommend a range of 1.325-1.350v if overclocking.
Memory Voltage - Intel recommend 1.50v plus/minus 5% which means upto 1.58v is the safe recommended limit. In our testing we have found 1.65v has caused no issues.
BCLK Base Clock - This is strictly a NO, anyone using base clock overclocking could/will cause damange to CPU/Mainboard. (Set manually to 100)
PLL Voltage - Do not exceed 1.9v!!



Processor - Basically we recommend customers not to exceed 1.35v to play it safe, all our bundles are set at 1.3250v or lower, any competitors offering bundles above 4.6GHz you should be enquiring as to what voltage they are using as we believe anything over 1.38v will limit CPU lifespan and anything over 1.42v will likely kill the CPU or severely limit its lifespan.

Memory - Intel recommend 1.50v plus/minus 5% which means 1.60v is the ideal safe maximum, but we have found in our testing all 1.65v memory is fine. We have also found most new 1.65v like Corsair XMS3 will run at its rated timings with just 1.50-1.55v which is well within Intel specifications. So people upgrading to Sandybridge you can still use your old DDR3, but we do recommend you run it at 1.60v or less. We are shipping most of our bundles which feature Corsair XMS at 1.50v-1.55v at rated timings. We've also discussed with Asus and MSI regarding voltages for memory and they also confirm in their testing 1.65v caused no issues with reliability.

Base Clock - To put it simple if you value the life of your components, do not overclock using base clock!

PLL Voltage - Again do not exceed 1.9v!

[vesseblah]

One hour of OCCT's Linpack test will also show whether the setup is stable. In fact, it will usually find instability that Prime95 won't.

Some reports say that Prime95 will actually get SNB temperatures 2-3C higher than Linpack for some reason, so running both would be advisable. Found this with quick googling when looking for 2500K overclocking info and results, but nothing confirmed.

---------- Post added 2011-01-19 at 04:32 PM ----------

Currently running 2500K on Asus P8P67 Pro board with 1.2 vcore on 42x multiplier and no other changes in BIOS, probably not gonna bother going up from here.

[Ai]

Prime95 actually puts more stress on the sandy bridge cpus than ITB unless you update the linx pack. The safe 24/7 core voltage is still unknown but going past 1.38 is probably safe as intel specs say the max voltage is 1.52. There has been much discussion on overclock.net about the max safe voltage and it has been shown that staying under 1.5 is the best option. At that voltage level it becomes a question of your cooling though.

Note: This is not absolute so always do research before overclocking.

Edit: Prime can run for 13 hours and still crash a system. It is usually after the 24 hour mark that you can say the system is 99% stable. There had also been posts about people passing 24 hours of prime and at least 50 passes of linx at max settings and they still crashed when playing a game. Sandy bridge might be easy to overclock but it seems hard to get stable.

[pullmyfinger]

I'm fining tuning my voltage right now for my 2500k on P8P67 board. It seems if i set my vcore option to manual, the voltage is constant, even when my multiplier is 16x. Is this normal for manual mode?

Only when I use offset mode, does the voltage go down to around 1v when the multiplier goes down to 16.

That aside, I'm current having my CPU at 4.8ghz (47x102.1) on only 1.28 volt (the voltage shown inside cpu-z). Everything seems rock solid. Prime95 only max the temp at 67c, 66c, 62c, and 60c respectively for each of the cores. I set my vcore on offsetmode at 0.005v

I guess i got a pretty good CPU.

[kidsafe]

You've got a "golden sample" if you really are fully stable at 4.8 GHz and 1.28V. Luck you. I might even try to go higher in that case!

And yeah, manual mode locks the core voltage at the amount you set it to other than the "vdroop" you get under load determined by your Load Line Calibration setting.

[pullmyfinger]

thanks. I put the load line calibration on auto currently. I'm not so sure about fully stable, but my prime95 actually crashed and did not respond after about 40 minutes while I was writing this post. Does that mean it's not stable? I thought normally if prime95 is not stable it would report an error right? Would it actually crash? I was running the small FFT test to mainly stress the CPU

my vcore was on offset mode with auto voltage. And my board actually overvolted my CPU by quite a bit, to 1.4v. I was pretty surprised i was even able to boot into windows with +0.005 offset, much less running prime95 for 40 minutes. I'll post my settings in a sec.

---------- Post added 2011-01-20 at 03:52 AM ----------

So these are the settings I am using for my set Intel 2500k, asus P8P67 mobo, GSkill Ripjaws 4x2GB Cas 9 1.5v rams, galaxy gtx 460

Once again, i got my computer at 4.8ghz (47x120) and everything seems running very smooth

BIOS Version: 1053 x64 Beta
AI Overclock Tuner: Manual
BCLK: 102
Turbo Ratio: 47
1632mhz RAM at default 9-9-9-24
Internal PLL Overvoltage: Auto
EPU Power Saving: Enabled
Load-line Calibration: Auto
VRM Freq: Auto
VRM Spread Spectrum: Enabled
Phase Control: Optimized
Duty Control: T.Probe
CPU Current Capability: 100%
CPU Volt: Offset Mode
Offset voltage: +0.005
DRAM Volt: 1.51875
VCCSA Volt: Auto
VCCIO: 1.15
CPU PLL: 1.88125
PCH: Auto
CPU Spread Spectrum: Auto
CPU Configuration: Everything is enabled (VT, etc) except for Limit CPUID MaxVal

---------- Post added 2011-01-20 at 03:53 AM ----------

my Superpi 1M is 7.876 and my CPU voltage maxes at 1.24 when using SuperPI

[kidsafe]

Yeah my old Q6600 would report a rounding error if it needed slightly more voltage, but that doesn't seem to be the case with Sandy Bridge processors. Prime95 will just lock you up or BSOD instead with my system when it's not getting enough voltage.

Hmm odd, how did your CPU overvolt to 1.4V with just a +.005V offset? I don't even think I can use an offset with the CPU Voltage setting set to Auto.

EDIT: hmm yeah, those settings aren't optimal for an OC. I'm raiding ATM, but I'll post my settings in list form soon so it's easier to track.

[pullmyfinger]

Yeah my old Q6600 would report a rounding error if it needed slightly more voltage, but that doesn't seem to be the case with Sandy Bridge processors. Prime95 will just lock you up or BSOD instead with my system when it's not getting enough voltage.

Hmm odd, how did your CPU overvolt to 1.4V with just a +.005V offset? I don't even think I can use an offset with the CPU Voltage setting set to Auto.

I meant before i started fine tuning my CPU, i was using auto offset by the motherboard, so my vcore would be at 1.4v during prime95 stress testing. So my motherboard was overvolting my CPU by more than 1 whole volt!

Yea, nm, i just got an round error during 32m superpi, so i guess i need some more offset voltage. How much you think I need?

EDIT: My 32m just passed after I set offset to 0.015

Edit2: My 32m for all 4 cores at the same time (hyperpi) just passed on the same offset, max voltage during 1 core 32m is 1.272, max voltage during 4 core 32m is about 1.312

[kidsafe]

Added my own settings in a concise list at the bottom of the original post.

[kidsafe]

Was able to improve my own personal OC a little bit with the new 1253 UEFI release. Namely my CPU Voltage offset has gone from .030V to .020V, resulting in a max of 1.368V under load instead of 1.376V. I was also able tighten up my RAM timings all the way down to 8-9-8-24-1T.

I was, however, not able to find stability at 4.9 GHz.

[Plasmon]

Thanks for this guide, I'll be assembling a system with the ASUS P8P67 PRO very soon and this will help... I'm an OC noob, but I've purchased all the sufficient hardware and cooling to push my system pretty far.

My entire OC experience consists of going into the bios and increasing the CPU multiplier by about 5% on my old AMD 64 3200+.



And for my first noob question:

Note that tCase on the i5 2500K and i7 2600K is 72.6°C. This puts maximum core temp around 83°C. Usually, I try to keep my maximum core temperature around tCase, as this allows for fluctuations in ambient temperature.

What is tCase? Sounds like some kind of important temperature maximum or threshold...

[kidsafe]

Tcase refers to the temperature of the CPU measured at the heatspreader. This is usually a value much lower than the Tjunction which is measured at the core level. The Tjunction is not mentioned on those pages, but RealTemp and CoreTemp both assume the max is 98C.

As my OC voltages seem to keep going down, I'm not going to discourage people from using slightly higher offsets. I think in general it's best to keep your max CPU Voltage under 1.4V, and even better to keep them under 1.38V. Just keep that in mind.

[pullmyfinger]

Was able to improve my own personal OC a little bit with the new 1253 UEFI release. Namely my CPU Voltage offset has gone from .030V to .020V, resulting in a max of 1.368V under load instead of 1.376V. I was also able tighten up my RAM timings all the way down to 8-9-8-24-1T.

I was, however, not able to find stability at 4.9 GHz.

Hey, what a coincidence, i also ended up settling on +0.2v. My max voltage is also around 1.368. But im still using the old beta BIOS, will try to update the BIOS later tonight.

EDIT: Where did you find the new BIOS? Im still seeing the 1053 beta as the latest

[kidsafe]

1253 beta was posted here: http://www.xtremesystems.org/Forums/...d.php?t=265477

1204 'official' was released yesterday and is basically the same.

As for higher clocks, I was able to run at 4.9 GHz on 1.4V and 5 GHz on 1.432V, but for gaming I will be sticking with 4.8 GHz and 1.368V.

[kidsafe]

Made some huge changes to the guide. My overclocking strategy no longer uses LLC, and CPU voltage offsets are now done through the Additional Turbo Voltage offset instead of the CPU Voltage offset.

[noremac]

A few things I wanted to point out...

DRAM Voltage should be whatever your RAM kit is specified to run at...don't set this above 1.65V.

Intel spec states that this voltage should be kept closer to 1.5V rather than 1.65V.
Note this article at Bit-Tech: "As with LGA1366 and LGA1156 CPUs, keep this value to within 0.5V of the VCCIO voltage to prevent long term damage to the CPU. By default, this is 1.1V, which means the 1.65V used by previous Intel DDR3 memory is still acceptable. However, more recent memory will be rated at 1.5V (or even 1.35V if you choose a low-voltage kit). Increasing the VCCIO voltage obviously gives you more overhead on your memory voltage (remember, add +0.5V at most or risk damaging your CPU)."
So while 1.65v may be "acceptable," that is probably the absolute max and the lower away from that number the better. Since the P67 platform launch there have been many more 1.5v memory kits surfacing on the market, so I would assume that Intel expects that to be the standard for this platform.

If you were able to boot into Windows, then start a Prime95 Torture Test and select the first option: Small FFTs. At this point your CPU will be running full tilt.

While this may max your CPU at 100% load, it has been shown that with the new Sandy Bridge architecture that using the Blend mode on Prime 95 is better than any other option. Many people have said that their overclock was stable for hours and hours of large FFT or small FFT but failed using blend.

I should also mention that my CPU cores get up to 72-73 degrees celsius under load with a large Noctua NH-D14 cooler. If you don't have a decent cooler, I suggest backing your overclock down to 4.5 GHz max with a lower CPU Voltage offset.

Those temps sound pretty bad for that cooler. That is probably the best air cooler money can buy and it even does better than some cheap water cooling solutions, but my temps don't even get that high on a $30 Hyper 212+ at 5GHz and if memory serves me right the DH14 goes for about $80. Unless your ambient temperature is like 30c I would try to remount the cooler for better results.

[kidsafe]

A few things I wanted to point out...

Intel spec states that this voltage should be kept closer to 1.5V rather than 1.65V.
Note this article at Bit-Tech: "As with LGA1366 and LGA1156 CPUs, keep this value to within 0.5V of the VCCIO voltage to prevent long term damage to the CPU. By default, this is 1.1V, which means the 1.65V used by previous Intel DDR3 memory is still acceptable. However, more recent memory will be rated at 1.5V (or even 1.35V if you choose a low-voltage kit). Increasing the VCCIO voltage obviously gives you more overhead on your memory voltage (remember, add +0.5V at most or risk damaging your CPU)."
So while 1.65v may be "acceptable," that is probably the absolute max and the lower away from that number the better. Since the P67 platform launch there have been many more 1.5v memory kits surfacing on the market, so I would assume that Intel expects that to be the standard for this platform.

While this may max your CPU at 100% load, it has been shown that with the new Sandy Bridge architecture that using the Blend mode on Prime 95 is better than any other option. Many people have said that their overclock was stable for hours and hours of large FFT or small FFT but failed using blend.

Those temps sound pretty bad for that cooler. That is probably the best air cooler money can buy and it even does better than some cheap water cooling solutions, but my temps don't even get that high on a $30 Hyper 212+ at 5GHz and if memory serves me right the DH14 goes for about $80. Unless your ambient temperature is like 30c I would try to remount the cooler for better results.

Intel has not stated specifically that VCCIO and DRAM voltage have to be within .5V for the Sandy Bridge IMC. If you are worried about it, then bump up up your VCCIO to 1.15V. I currently have mine set to 1.125V with my DRAM voltage set to 1.6V.

And the blend test fails because it is putting an ever so slightly lower load on your cores. In offset mode, this might cause your voltage regulation to dip and cause instability. This is also why I suggest against using any sort of Load Line Calibration outside of an extreme overclock. Try setting a manual non-offset CPU voltage and see if your Blend test fails vs small-FFTs...it won't.

My temps are quite standard for those voltages. If you're seeing less at 5GHz on a Hyper 212+, you must have a wind-tunnel for a case or a golden sample running at extremely low voltages. My case fans do run at 500 RPM 24/7, but the CPU fans can spin up to 1500 RPM. Google: "NH-D14" 2600K 4.8GHz

[kidsafe]

My replacement board from Asus came today and there are some oddities. The same exact settings I was using before now report much lower CPU voltages and a couple degree lower temps at load. I'm seeing 1.344V now instead of 1.376V. Still seems to be stable. Going to have to stress test it overnight to be sure.