Intel Core i5 & i7 Sandy Bridge Overclocking and Feedback

[LAN_deRf_HA]

Isn't D2 the retail stepping? That's what they say they used. And all that Intel stuff is on by default with auto, yet everyone still seems to top out at 4.4-4.5 on auto settings. I'm just not seeing a huge gap between leaving the intel speed stuff on vs. off.

 

[cadaveca]

Of course you won't, because you are temperature-limited. but if you don't see a big differnce, then why do you have to push the "max", rather than what it was designed to do?

ASUS said:

these results will most likely represent retail CPUs.

These were ES chips, not retail.

Also:

K series overclocking and voltage range recommendations



Quick Note regarding Voltage Scaling – Internal binning of both D1 and D2 parts we discovered consistent voltage scaling patterns.

1. For K series parts, the stock voltage supplied will allow for consistent overclocking generally up to a multiplier of 43x. There is potential for the multi to be raised to 44x depending on the load induced. This default voltage range be approx 1.240 to 1.260 under load.

2. Increased range between 44 to 47x multipliers will generally require a voltage range between 1.30 to 1.375V with an LLC recommended setting of high to ultra high.

3. Increasing the range between 48 to 50x multiplier will generally require a voltage range between 1.40 to 1.500 with a LLC recommended setting of ultra high.

4. Increased range between 50 to 52 (52 generally considered peak max multiplier except for rare 54x parts) will generally require a CPU voltage range between 1.515 to 1.535V with LLC at Ultra High and potential fine adjustments to the CPU skew range.



Overall a key item to note is the best voltage to oc scaling range potential for the turbo multiplier is 1.400 to 1.425 vcore. Using this voltage range with an LLC recommendation of ultra high will generally provide the best scaling potential with proper load temperatures*. We have generally found exceeding this voltage will not provide additional scaling or will increase load temperatures to a high level with synthetic load applications ( like Prime, Linx, OCCT ). Should you use more realistic loading testing (our recommendation is a combination of AIDA64 stress test, PC Mark Vantage) then temperatures under will be considerably under the max TDP rating.

K series overclocking multiplier wall



Quick Note regarding the K series multiplier wall – Previous generations of Intel CPUs due to architectural differences had two aspects that could readily change OC scaling potential. These two factors were TDP (heat output under load) as well as core (CPU/VTT ) voltage being supplied. With previous generation cpus, improvements such as high end air cooling, water cooling or more extreme forms like LN2 in addition to extreme voltage increases could potentially provide increases in clock speeds on a continual ramp until the limits of the CPU were reached. This is not the case with K series on the P67 chipset.

In internal testing we have found that maximum multipliers in excess of 50 (up to 54) have been able to be realized with full stability under Air Cooling and with cpu vid voltages under 1.525V. In fact, the processors will actually start to downclock as they get colder with temperatures under -20C actually performing worse than the retail air cooler at times. The processors optimal range for performance is around 15~20C in our testing.

An example is noted below.

Should you have a CPU in hand that reaches a 47 multi easily ( posting/booting and running full stability tests at a set Vcore of 1.375V) you would then attempt to jump to the next multiplier. With the K series CPUs you may have reached the max multi regardless of the cooling or voltage used. Adjustments to either of these settings will not change the ability of the board and CPU to post at 48x. The only time it will continue to scale would be if the CPU in hand supports higher multipliers or can be finely tweaked to reach that next multi, in this case we suggest using the 47x multi and adjusting Bclk until you reach the limits of stability.

100BCLK + 47 Turbo Multiplier with Vcore of 1.375 and LLC ( load line calibration of ultra high = 4.7GHz

100BCLK + 48 Turbo Multiplier with Vcore of 1.400 and LLC ( load line calibration of ultra high ) = no boot

100BCLK + 48 Turbo Multiplier with Vcore of 1.425 and LLC ( load line calibration of ultra high ) = no boot

100BCLK + 48 Turbo Multiplier with Vcore of 1.450 and LLC ( load line calibration of ultra high ) = no boot

It is important to note that a post will still occur with a multi present that is at the multi wall. This is because only when the OS begins to initialize and the Turbo ratio is initialized ( kicks in and ramps up ) that Turbo Multiplier will be raised to defined multi at this point it will then cease to boot due to the multiplier wall being reached.

Just to get all the info in one place.


But anyway, so you are going to use ASUS's methods, these are the result you should expect. You a differnt method, you might get different results.

 

[PaulieG]

I have had parts since before christmas. I've played with more boards than you know, and more chips, too. Don't take my experience as a one-off...it isn't. I've got at least a month head-start on clocking these chips, and I'm not really saying that YOU GUYS specifically are doing it wrong, but more that the tools provided for you aren't always the best option.

Don't take it personally, and listen a bit; I'm just trying to help you out. SandyBridge is NOT like other chips..AT ALL...so traditional approaches for clocking don't work.

Temps don't matter, really, except to unlock multis. A long time ago, when these chips were first in design, if you go back to forums posts of mine...you'll see me specifically asking Intel employees that they give us 5GHz on air.

They did it. There are some caveats to getting 5GHz, sure, but as far as I can tell, every sandybridge will hit that.


Volts aren't the answer, always. Increasing volts should not be used to provide additional power, but to bring frequency swings back into "spec".

I'm slowly going through CPUs...i think quite a few of us here are already into binning chips...


I'm going slow through mine though, because I've only got one arm. It's gonna take some time still yet, as I'm not just after max multi like these guys who go for benchmarking...I awnt to paly with lots of chips to learn more about the chip itself.

That said, I know, 100%, that max multi is a function of temps. Getting the cpu to open up teh extra multis can be an issue, but when I've got it figured out, I'll post info. Maybe I won't figure nothing out, but I do not think current bioses are helping much.

Rember many boards didn't even ship with proper EFI bioses. What we have now is kinda a Hybrid bios, and it shows itself hard, if you ask me. I juat need OEMs to send me more boards to I can start going sub-zero, as I won't do that on a board I've purchased at this point.

I've played with 2 2600k's and 2 2500ks. I have 2 more 2600k's on the way, that have already been "pre screened" for high multipliers. I've found both 2600k's to respond very similar to multiplier increases and temperature. I've not had any problems with temps at all before the multi's stop scaling well with voltage increases, and I'm not taking my chips over 1.4v.

I've had a (claimed) Intel rep tell me that they are not even sure what the safe voltage limits ore on these chips, which is why they give such a low recommended safe range. I think sometimes we give Intel and AMD personnel more credit for understanding the maximum potential of their products. They focus more on basic voltage tolerance for long term use. This is another reason why they do not officially support overclocking.

I agree with you that brute force and voltage increases are not always THE answer, but they are usually a vital part of the equation. I'm really not seeing how that would be false with these 32nm chips.

Maybe you are seeing something that the rest of us are not, but I do spend some time doing research on what is working for people. I'm not finding any information from anyone else that there is software out there that is making it easier to overclock these chips without voltage increases, and until we see what you are using and try to implement the same, it's hard to accept it as anything close to fact.

I do hope that while we are all trying to help each other, we are sharing what we have learned.

 

[cadaveca]

Paulie, maybe listening to other people's results, who really, are just doing the same thing, listening to results, and then plugging in settings, isn't the right approach.


I chose to do it my own way, and seemingly, am getting better results, with what is really, just an avg chip.

 

[LAN_deRf_HA]

Of course you won't, because you are temperature-limited. but if you don't see a big differnce, then why do you have to push the "max", rather than what it was designed to do?

ASUS said:



These were ES chips, not retail.

Also:




Just to get all the info in one place.


But anyway, so you are going to use ASUS's methods, these are the result you should expect. You a differnt method, you might get different results.

I'm not sure your source is as up to date as mine? They removed D1 results. They said D2 chips, "most likely" representing retail because D2s are what's shipping to retail. It's just safe wording on their part. And I'm not even sure what you're saying about max? I'm saying I'm not seeing people able to overclock further with speedstep on. I certainly notice the performance difference on menial tasks. I'd imagine anyone who noticed the latency switch from crt to lcd can notice the sluggishness browsing and doing whatever else at 1.6 GHz. Even if it helped overclocking, I couldn't live with it not being locked in at full speed. I might as well go back to core 2 then.

http://hardforum.com/showthread.php?t=1578110

 

[PaulieG]

Paulie, maybe listening to other people's results, who really, are just doing the same thing, listening to results, and then plugging in settings, isn't the right approach.


I chose to do it my own way, and seemingly, am getting better results, with what is really, just an avg chip.

I never "plug in" results. I implement methods. Some of the conversation is useless, since no one here is aware of this "different way" that you don't care to share. I implement successful methods that are being shared, and in turn I share my experience to help others. Isn't that the real purpose of threads like this?

 

[cadaveca]

I already "shared". I'm using turbo, and increasing current instead of volts. No PLL. Software, instead of bios. Using Intel-provided methods. No magic. My results are different, due to that...I don't get what you're on about here, Paulie.


You are using bios, disabling power saving features, and loadline claibration. Pretty simple, if you ask me.

I thought we moved on, clearly you have not. Please leave the attitude at the door, OK?

 

[PaulieG]

I already "shared". I'm using turbo, and increasing current instead of volts. No magic. My results are differnt, due to that...I don't get what you're on about here, Paulie. I thought we moved on, clearly you have not.

I'm referring to "software" you've mentioned in this thread that has helped you avoid higher voltages and higher clocks? Am I wrong about this? No attitude. We are all good.

 

[cadaveca]

Dude, pls, open your eyes, and drop the hate. I told what software; if it's not on your mobo CD, then your bios does not support it, and there's no point in looking for it.


Like geez, guys, WTF?!?

Here I am explaining WHY i chose that method...you guys are too much.

 

[PaulieG]

Dude, pls, open your eyes, and drop the hate. I told what software; if it's not on your mobo CD, then your bios does not support it, and there's no point in looking for it.


Like geez, guys, WTF?!?

Here I am explaining WHY i chose that method...you guys are too much.

No hate at all. Just looking for knowledge. If I missed where you specifically mentioned the software you are using, I'm sorry. I'm just not seeing it. Honestly, I'm just looking for tools.

 

[LAN_deRf_HA]

Are you referring to some universal intel software or the MB specific tools like the asus ai suite?

 

[cadaveca]

Then do me a favor, pop in your mobo cd, look for software that wasn't installed automatically, or that you haven't tried yet.


I mean, I've never said you'll get higher clocks, but yeah, you might be able to use lower votlage, and enable turbo features just the same. Save the planet a bit, you know?

I don't get the hype you guys are placing on this, but OK. You asked what I'm doing, and why my avg chip seems so nice. I'm not posting any more screenshots until they show up in reviews, no big deal.

If you don't have the software, you're SOL anyway.

Yes, LAN, like c'mon. I've mentioned many times, look on mobo cd, and you guys haven't looked yet? Why don't you just send me your rig, I'll clock it for ya, and send it back?:shadedshu

Listen, maybe you learn something, instead, so I don't have to keep doing this? Does any teacher just provide the answer? C'mon, now. Yes, I expect you to use your head a bit and figure it out. Why is that so hard, so bad?

 

[PaulieG]

Then do me a favor, pop in your mobo cd, look for software that wasn't installed automatically, or that you haven't tried yet.


I mean, I've never said you'll get higher clocks, but yeah, you might be able to use lower votlage, and enable turbo features just the same. Save the planet a bit, you know?

I don't get the hype you guys are placing on this, but OK. You asked what I'm doing, and why my avg chip seems so nice. I'm not posting any more screenshots until they show up in reviews, no big deal.

If you don't have the software, you're SOL anyway.

Yes, LAN, like c'mon. I've mentioned many times, look on mobo cd, and you guys haven't looked yet? Why don't you just send me your rig, I'll clock it for ya, and send it back?:shadedshu

Listen, maybe you learn something, instead, so I don't have to keep doing this? Does any teacher just provide the answer? C'mon, now. Yes, I expect you to use your head a bit and figure it out. Why is that so hard, so bad?

Teacher? We are all just sharing info. The sharing is what makes the community great. We all spend plenty of time thinking. We just don't always think the same way. Sigh. Let's move on to another topic...

How many people are running 8GB ram on SB? Those who are, how many people are able to run 2133 or higher? I've had 3 sets of high end ram, and only 1 set has hit these speeds, when they easily do so on 1366 and 1156.

 

[cadaveca]

Teacher?

There's the hate again. Yes, teacher. LoL. You asked, and plain and simple, the slight chance i might know a bit more than you isn't something your ego deals well with. That's OK, but when I already gave you the answers to teh questions you've asked, and I've got to repeat teh same thing over and over, yeah, I'm in the role of teacher. We all are, and once you remove the ego, and add some humility, things work much better, I'm sure you'll agree. We are all teaching each other, and your attitude and responses say more than you intend, it seems. So much for mutual respect, I guess.


2133 hasn't been that hard. Takes bit more volts on mem. Plain and simple, SB puts more load on the mem, so memory doesn't scale as well. Most people in 1366/1156 aren't running into memory limitations, it's more controller limitations, and that limitation has been raised, so teh mem behaves differently. There's no 1/3 increase in bandwidth from 1156 to 1366, even with the extra channel, so clearly the memory companies have been able to get away with far lower toleraces with the previous platforms.

I'm pretty sure it's gonna take "new" IC configs, like the ripjaw X, or any of the other 1.5v 2133 sticks, to get 8GB(in a 4stick x2GB) config reliably.

 

[LAN_deRf_HA]

Then do me a favor, pop in your mobo cd, look for software that wasn't installed automatically, or that you haven't tried yet.


I mean, I've never said you'll get higher clocks, but yeah, you might be able to use lower votlage, and enable turbo features just the same. Save the planet a bit, you know?

I don't get the hype you guys are placing on this, but OK. You asked what I'm doing, and why my avg chip seems so nice. I'm not posting any more screenshots until they show up in reviews, no big deal.

If you don't have the software, you're SOL anyway.

Yes, LAN, like c'mon. I've mentioned many times, look on mobo cd, and you guys haven't looked yet? Why don't you just send me your rig, I'll clock it for ya, and send it back?:shadedshu

Listen, maybe you learn something, instead, so I don't have to keep doing this? Does any teacher just provide the answer? C'mon, now. Yes, I expect you to use your head a bit and figure it out. Why is that so hard, so bad?

It just seems a little odd, the software often does nothing you can't do in the bios anyways. And you seemed to make it all sound a bit vague and mysterious by not naming names, I guess cause your review isn't done? Let me just try to get this current thing straight. Are you saying leaving on all the c-states and what not allow more current to flow through? Or are you literally talking about the vrm/current settings? Because messing with vrm and current settings is common and fairly well understood on P67, only resulting in better voltage efficiency in very specific situations to small degrees. I just crank up the current limit because there's no harm, but I haven't noticed a benefit yet.

 

[PolRoger]

I think it ultimately does come down to the individual chip. Some chips with out question are better able to maintain signal integrity as you raise the multi then others. They run higher mutli(s) with out enabling PLL Voltage overide in BIOS and they are also the ones that can boot into windows at 54x, 55x, 56x(+).

 

[cadaveca]

I just crank up the current limit because there's no harm, but I haven't noticed a benefit yet.

Maybe the value you are setting is too high. Maybe the value is current for gpu, or is not current for turbo?

Maybe the current value only takes effect with turbo?

I guess I gotta say, this isn't me trying to act smart here, these are real questions.


I adjusted not "stock" voltage, but "turbo voltage"...and Turbo current.

I mean, it's cleary working in my situation, and I know that we are clocking differntly, so all I can say is that the different methods explain it all. Frankly, like I've said, I suspect that not all bioses are fully functioning, or need some tweaking.

I... just can't say what your board is doing, really.

 

[LAN_deRf_HA]

I'm guessing you haven't used a P67 asus board yet? The current is definitely for the cpu, and asus only overclocks using turbo. You can't turn it off in the same sense as you can on other boards like a UDx or biostar. I think differences between boards are another "language" barrier here.

 

[PaulieG]

There's the hate again. Yes, teacher. LoL. You asked, and plain and simple, the slight chance i might know a bit more than you isn't something your ego deals well with. That's OK, but when I already gave you the answers to teh questions you've asked, and I've got to repeat teh same thing over and over, yeah, I'm in the role of teacher. We all are, and once you remove the ego, and add some humility, things work much better, I'm sure you'll agree. We are all teaching each other, and your attitude and responses say more than you intend, it seems. So much for mutual respect, I guess.


2133 hasn't been that hard. Takes bit more volts on mem. Plain and simple, SB puts more load on the mem, so memory doesn't scale as well. Most people in 1366/1156 aren't running into memory limitations, it's more controller limitations, and that limitation has been raised, so teh mem behaves differently. There's no 1/3 increase in bandwidth from 1156 to 1366, even with the extra channel, so clearly the memory companies have been able to get away with far lower toleraces with the previous platforms.

I'm pretty sure it's gonna take "new" IC configs, like the ripjaw X, or any of the other 1.5v 2133 sticks, to get 8GB(in a 4stick x2GB) config reliably.

OK. I'm thinking you and I could go rounds about ego etc. Let's agree to disagree. Moving along. Agreed on the lower tolerances. The only sticks I can get to 2133 are my Redline 1600's. My Geil 2133's and Team Xtreem 2000's will not boot 2133 despite doing so on other platforms. Good to see Mushkin stepping up with the IC's on a 1600 kit.

 

[cadaveca]

I'm guessing you haven't used a P67 asus board yet? The current is definitely for the cpu, and asus only overclocks using turbo. You can't turn it off in the same sense as you can on other boards like a UDx or biostar. I think differences between boards are another "language" barrier here.

Yeah, for sure, the board differences matter...that's why I said I gotta play with more boards before i can really come up with anything tangible for higher clocks.

However, the point remains that we all seem to basically be stuck @ 4.7ghz(I'm not, that's just on one board without pll).


If we go by ASUS's list, minimum should be 4.4ghz, and those chips should hit 4.9 or 5.0 with PLL.


How come this isn't happening? You telling me that your chip is 4.1GHz max? I don't get it. I mean, current works for me...but not for you? Why?

Good to see Mushkin stepping up with the IC's on a 1600 kit.

The redlines I got rid of are quite clealry some of the best ICs I ahve ever had my hands on, but at the same time, GSKill, Kingston and Corsair are doing just as well, it just seems that thier binning process is much tighter, so there's less overclocking available with them.

 

[LAN_deRf_HA]

I mean, current works for me...but not for you? Why?

Well that's the big questions isn't it?

What chips are those redlines that worked out so well? The usual 6-8-6 fair or the old 6-7-6 hypers? Haven't heard good things about hypers on sandybridge. I'm guessing due to sub-timings being too tight.

 

[cadaveca]

Well that's the big questions isn't it?

What chips are those redlines that worked out so well? The usual 6-8-6 fair or the old 6-7-6 hypers? Haven't heard good things about hypers on sandybridge. I'm guessing due to sub-timings being too tight.

They were definitely PSC, not hypers. They seemed to have been a limited release though. 1600 6-8-6 1.65vstock, they do 1600 6-8-6 @ 1.5v. 1900 6-8-6 1.65v, then had to go to 6-9-6. 9-11-9-27 2133 @ 1.5v.

As to the question, to me, because i got the software you guys don't, the bios is different to support that software, and hence the different results. If you check my P7P55 board review, ASUS has a switch in bios to allow other software apart from ASUS's to control bios, so i suspect that maybe this might be part of it...I cannot say for sure, but that's how it seems to me, and hence my seemingly "cryptic" coments earlier.

As far as I am concerned, there is nothing i can do that you guys cannot. It's jsut switching available options, anyway, and anyone can press keys on a keyboard. It's down to the hardware, for sure. I want to blame bios only.

 

[LAN_deRf_HA]

I'm tempted to switch to this kit http://www.newegg.com/Product/Product.aspx?Item=N82E16820231450 as people seem to manage very nice timings on them and I imagine they're designed for the platform, but I have no way to tell if that would really benefit me in my 2133 8gb quest. I mean your board and chip can limit your ability to hit 2133 just as much. My B3 comes tomorrow, maybe that will give me some different results. In theory these 2400 sticks should have a lot of potential, but increasing vccio hasn't helped me much so far. In fact I seem to get worse results going over 1.125v.

I can hit 2133mhz with 8gbs, going as low as 8-9-8-24... just isn't stable.

 

[PaulieG]

Well that's the big questions isn't it?

What chips are those redlines that worked out so well? The usual 6-8-6 fair or the old 6-7-6 hypers? Haven't heard good things about hypers on sandybridge. I'm guessing due to sub-timings being too tight.

Here are my Redlines (2x 4GB). I'm 99% sure they are PSC.

 

[LAN_deRf_HA]

2x4 eh? Isn't that a bit easier on the board? I'm sticking to 4x2 in prep for ivy and quad channel.