Intel Core i5 & i7 Sandy Bridge Overclocking and Feedback

[20mmrain]

Alright things I noticed....

1. If I enable Vcore LLC up to 100% I can get a higher overclock at lower voltss.
2. But the temps stay a little higher.

(So which one would you suggest doing?) I have never had a board that allowed different settings to Vcore LLC. So I am a noob with this. The boards in the past I have had either said enable or disable.

Next thing I noticed was....

CPU Core Current Limit.... that I have set to auto.
You told me to set it too 145watts but it is listed in amps. But why would this setting be important. I have looked it up and I see know one else even mentioning it. As a matter of fact I can't find one post about it.

Last the c1 states.... everywhere and even in this thread there are contradictions with this setting. Some say enable it others say don't. I have found more who say don't but maybe if some one could explain to me why it doesn't really effect much on SB CPU's.

I guess I do have one more question..... is 1.33 volts a little high for a 4.5 Ghz overclock. Like I said with experimenting I got the voltage down to 1.29 with 100% LLC and I am sure I can get even lower. But is having 100% LLC even good for these CPU's 24/7. So is the 1.33 volts for 4.5 ghz normal or not so much???

 

[Frizz]

LLC Overshoots the vcore at 100%, my mobo for example under load is rising 0.10-15 instead of dropping using max LLC. You should be able to drop down the vcore quite significantly. I personally wouldn't have it on after the CPU starts reaching 1.38volts though just to steer clear from spikes.

 

[20mmrain]

LLC Overshoots the vcore at 100%, my mobo for example under load is rising 0.10-15 instead of dropping using max LLC. You should be able to drop down the vcore quite significantly. I personally wouldn't have it on after the CPU starts reaching 1.38volts though just to steer clear from spikes.

Thanks so you would recommend lower then 100% though???? Possibly around 50% to 75%

 

[Frizz]

Thanks so you would recommend lower then 100% though???? Possibly around 50% to 75%

I would recommend to do what is comfortable, max LLC should be fine under 1.38vcore. I am currently using max LLC and it is nice to have a much lower vcore than needed. Got my chip set to 1.250vcore in the BIOS and it idles at 1.260 loads at 1.272.

 

[PaulieG]

Yeah, I'm very happy with this chip and the P8P67 Pro (B3). Been awhile since I enjoyed an Asus board this much.

 

[LAN_deRf_HA]

Yeah, I'm very happy with this chip and the P8P67 Pro (B3). Been awhile since I enjoyed an Asus board this much.

http://img.techpowerup.org/110406/2600k_4800.png

What bios? 1305?

 

[LifeOnMars]

On my P8P67 I have a primary plane adjustment which specifies CPU current (Ampage). It is done in incremements so I take it I should adjust in small increments and then run stability tests to see the difference at lower voltage?

What are the limits on this? I'm very interested but don't want to needlessly fry my little Sandy.

 

[LAN_deRf_HA]

I've asked about it before, no one seems to know for sure if the max setting is safe. I think it's red just for the sake of being the highest possible selection. Many set 110-130% just to make sure it isn't holding them back. I highly doubt it'll cause any harm.

 

[LifeOnMars]

I've asked about it before, no one seems to know for sure if the max setting is safe. I think it's red just for the sake of being the highest possible selection. Many set 110-130% just to make sure it isn't holding them back. I highly doubt it'll cause any harm.

No sorry Lan, not that setting ( I have that at 130%) but there is an option in CPU power management which specifies the current (amps) that the CPU is recieving as per Cadaveca's method I think. Just wondering what amount he would recommend using to start with?

 

[PaulieG]

What bios? 1305?

I'm at work, so I'll need to check, but I think it shipped with 1502.

 

[cadaveca]

But why would this setting be important. I have looked it up and I see know one else even mentioning it. As a matter of fact I can't find one post about it.

Think of it this way.


VOLTSx AMPS=WATTS

CPU is rated @ 95w. Look at default voltage...say 1.175v. Plug numbers in the formula, get an answer.

So, I got this handy Zalman meter, plugs inline to the cpu 8-pin. It tells me how much current, and how many watts, the VRM is pulling. Using the Intel-provided tools to auto-OC, I have found out what my cpu needs, basically, in wattage.

So, I know the wattage. Voltage, I can set to what I want. Pick one votlage setting, plug numbers into the very basic formula, and you'll know what amperage the cpu should need.



Oh, what's that? You aren't really measuring CPU power draw? Noone else is either? Oh, AND, noone is paying attention to current or OCP limits? Um, lol whut? Why are you changing LLC then? What purpose does that have?

 

[PaulieG]

Think of it this way.


VOLTSx AMPS=WATTS

CPU is rated @ 95w. Look at default voltage...say 1.175v. Plug numbers in the formula, get an answer.

So, I got this handy Zalman meter, plugs inline to the cpu 8-pin. It tells me how much current, and how many watts, the VRM is pulling. Using the Intel-provided tools to auto-OC, I have found out what my cpu needs, basically, in wattage.

So, I know the wattage. Voltage, I can set to what I want. Pick one votlage setting, plug numbers into the very basic formula, and you'll know what amperage the cpu should need.



Oh, what's that? You aren't really measuring CPU power draw? Noone else is either? Oh, AND, noone is paying attention to current or OCP limits? Um, lol whut? Why are you changing LLC then? What purpose does that have?

First, had to thank you because no one should sit at 666 thanks. LOL

Good point on randomly changing LLC without knowing power draw. I think people are simply looking at LLC as a tool to reduce vdroop, and increasing it until their system is stable. That's been the mindset over several platforms. Most don't want to know more that what fixes the problem. Maybe that's sad...but it's true.

 

[cadaveca]

First, had to thank you because no one should sit at 666 thanks. LOL

Good point on randomly changing LLC without knowing power draw. I think people are simply looking at LLC as a tool to reduce vdroop, and increasing it until their system is stable. That's been the mindset over several platforms.

OK, Paulie, look at it this way. VDroop mods have ALWAYS been about providing MOrE CURRENT. ALWAYS.

It's a function of GTL that current load lowers voltage. The switching frequency of the phases dictates how much of a drop is noticed before the voltage picks back up, too.


So, think back to 775 days, and all the P35/P45 boards blowing 12v CPU plugs...every single one was VDroop modded. Back then, you'll find numerous threads about people wondering why it blew the CPU 12v connector...yet in those same threads on XS, you'll find me there too; quite possibly, you'll find my post explaining WHY it blew.


Just because it's the mindset of the mob, doesn't mean the mob knows what it's doing...they are merely following the leader like the sheeple they are.


Meanwhile, I'm sitting here trying to "crack the code".


So yeah, my approach is different. Has been for years. I'm not one of those blindly changing settings...I have a distinct purpose in mind, and a framework of expereience to build off of. TOOT TOOT!

 

[PaulieG]

Back on topic guys.

 

[WhiteNoise]

This is not stable though. I can run IBT 24 hours straight with no errors but I can't surf the web without getting a BSOD. I'm not really sure what I'll do to make it 100% stable. As of right now I'm back at 4.7GHz which is stable for sure. I'm very new to this platform though so given time I'm sure I will work it out.

 

[cadaveca]

What waterblock are you using? Those are fairly decent temps for those volts.

 

[WhiteNoise]

What waterblock are you using? Those are fairly decent temps for those volts.

EK-Supreme HF - Full Nickel - Rev 2

 

[cadaveca]

EK-Supreme HF - Full Nickel - Rev 2

Are they shipping with 1156 hardware now? There was a time when they did not. I've been using an original DangerDen TDX from 2005, and need an update.

66c is well within what I'd consider comfortable. Normal load is gonna be high 50's? That's really nice.

Maybe ram is part of the problem? Or possibly, change you change switching frequency of CPU VRM?

 

[PolRoger]

Think of it this way.


VOLTSx AMPS=WATTS

CPU is rated @ 95w. Look at default voltage...say 1.175v. Plug numbers in the formula, get an answer.

So, I got this handy Zalman meter, plugs inline to the cpu 8-pin. It tells me how much current, and how many watts, the VRM is pulling. Using the Intel-provided tools to auto-OC, I have found out what my cpu needs, basically, in wattage.

So, I know the wattage. Voltage, I can set to what I want. Pick one votlage setting, plug numbers into the very basic formula, and you'll know what amperage the cpu should need.





Oh, what's that? You aren't really measuring CPU power draw? Noone else is either? Oh, AND, noone is paying attention to current or OCP limits? Um, lol whut? Why are you changing LLC then? What purpose does that have?

I've been following this thread off and on and I've made note of some of what you have said regarding "current" and using the supplied software utility that came with the motherboard...

How long before we see a review on one of your test p67 boards?

I believe you said that you prefer doing your o.c. from windows via the utility... why not just make the necessary adjustments via BIOS?

 

[cadaveca]

I've been following this thread off and on and I've made note of some of what you have said regarding "current" and using the supplied software utility that came with the motherboard...

How long before we see a review on one of your test p67 boards?

As soon as I get some B3 boards. I refuse to talk about the parts I have purchased so that nothing I talk about is coloured by me justifying my purchases. It wouldn't have been, really, but I don't want it to be a factor, either.

I believe you said that you prefer doing your o.c. from windows via the utility... why not just make the necessary adjustments via BIOS?

Simply, the software allows adjustments to bios that are not available in the bios itself. Things like the OCP max, and some memory timings, have proven crucial to my clocking, but the bios does not support user adjustments, so I'm left with no other chioce.

The software is pretty good too...I mean...it reboots for every single change...if it changes 5 things when using the auto-clock, it reboots 5 times, and verifies each has actaully changed before moving on to the next. I would have done the same, using bios, so I'm actually very impressed with the tools Intel has provided. Of course, bios is part of that too...the bios provides some of that support with the current version, in regards to secondary timings, that was not there in the software with a previous bios. but still, in bios, those secondary timings are not avaialble for adjustment...only in the software. Odd, but it works.

I also don't have any LLC options, so because I know that LLC affects the current provided, I've taken steps to manually do what LLC offers, but with finer adjustments. I do not know that this is possible on boards that offer LLC, but it should be.

And as an aside, no, I am not using LLC, so the gains given by adjsuting current are very noticible to me, but for those that use LLC, it might not give anywhere near the same results.

 

[WhiteNoise]

Are they shipping with 1156 hardware now? There was a time when they did not. I've been using an original DangerDen TDX from 2005, and need an update.

66c is well within what I'd consider comfortable. Normal load is gonna be high 50's? That's really nice.

Maybe ram is part of the problem? Or possibly, change you change switching frequency of CPU VRM?

It ships with a mounting plate that allows for multiple socket types. I bought mine for a Socket 775 board but it supports 1156 as well.

It could be the memory. I'm using Patriot Viper 2 Sector 5 edition DDR3 1600 -8GB- (4GBx2) and I will down clock it tonight and see what happens. Its currently running at its default timings.

Thanks for the ideas I will look into them tonight.

 

[cadaveca]

It ships with a mounting plate that allows for multiple socket types. I bought mine for a Socket 775 board but it supports 1156 as well.

Cool. I'll have to order one. Thanks!

IN regards to the memory, you might want to look at memcontroller voltage. If the mem tests fine, but you still ahve some odd stability issues, look at the memory controller.


Also, System Agent voltage can use a small(+10mv) boost sometimes, but be very careful pushing this voltage higher. SA voltage was key to getting that last multi stable for me.

 

[WhiteNoise]

I haven't touched the System Agent voltage yet. I will look into both that and the memcontroller volts.

Thanks again.

 

[PolRoger]

@cadaveca

What is the difference or advantages of two similar SB overclocks... one set to use somewhat higher voltages and 100% current vs. another set to use somewhat lower voltages and 160% current?

If a cpu needs a certain wattage to run 47x stress stable wouldn't both scenarios provide the same ~ wattage to the cpu and thus allowing for equally stable overclocks?

 

[cadaveca]

@cadaveca

What is the difference or advantages of two similar SB overclocks... one set to use somewhat higher voltages and 100% current vs. another set to use somewhat lower voltages and 160% current?

If a cpu needs a certain wattage to run 47x stress stable wouldn't both scenarios provide the same ~ wattage to the cpu and thus allowing for equally stable overclocks?

TBH, i'd say the lower-volt, higher current is more likely to experience overcurrent, and an early death.

And no, it's not just the wattage.

OK, so, CPU's transistors require so many amps to "switch", let say. Traditionally, AMD's CPU process required less amperage, and hence the higher stock voltages in AMD CPUs.

Like, for example, AMD needed .7-.8a to switch, but Intel required .9a to 1.0a, in previous generations.

So, you know, of course, there is a limit as to the current provided to the CPU..too much can lead to early electro-migration, or even burn out the circuit completely. This is similar to what happened with the 4870, and what they changed in the 4890 silicon...they doubled the power input carriers, becuase TSMC's mask lead to carriers that were not properly "exposed", to put it very basically.

So, it's a matter of finding the right balance.

GTL, Gunning Transceiver Logic, basically deals with "machine-gunning" current to the cpu. there is a specific frequency at which this occurs, and becuase we are dealing with clock signals, matching the GTL frequency with transistor frequency, of course, leads to stability, and a lack of sync can destroy stability.

Based on how these cpus work, maintaining that frequency sync is of the utmost importance...The way multis simply end...is a sure sign that signal integrity switching at whatever frequency the cpu needs to maintain that multi is off.

So, by adjsuting the voltage, you change the waveform, whereas increasing the current can help keep that waveform strong. I ahve faith that Intel's engineers have done thier best to give us the best mix, given how well these chips clock, so I increased the current.

Does that make sense? I a, trying to keep this as simple as possible, and easy to grasp, so maybe there something I missed...