Post Your Overclock!

[Britton30]

This is all quite interesting, but as a (so far) AMD user could someone explain the terms like, VCCIO, CPU PLL, Offset Volts, IntelME, VTT, LLC, etc without my wading through a bunch of OC forum posts? Thanks! Looking forward to the Intel jump in the near future.

 

[GeneO]

This is all quite interesting, but as a (so far) AMD user could someone explain the terms like, VCCIO, CPU PLL, Offset Volts, IntelME, VTT, LLC, etc without my wading through a bunch of OC forum posts? Thanks! Looking forward to the Intel jump in the near future.

Vccio is equivalent to Vtt , the memory controller voltage. Sometime you have to raise this for lots of memory or higher voltage memory or overclocking memory. It used to have a more important role before Sandy Bridge, when overclocking was achieved by varying the multiplier and the BCLK. With Sandy Bridge it is all multiplier - unless you are nuts and want to overclock your memory.

Intel ME is Intel Management engine. Some power management I thin but mostly provides custom bios config of network etc for IT shop.

Load Line Calibration: compensates for the drop in Vcore under load in an attempt to keep that voltage constant. However, with it enabled, voltage transients when load suddenly changes can exceed the chip maximum voltage

CPU PLL - Phase Locked Loop: Keeps the multiplied frequency stable and locked in-phase with BCLK.

Offsetvoltage: Instead of an absolute voltage, you provide an offset that is applied against the VID voltage. You need this for the power management to function properly so that when frequency is lowered, the voltage will be lowered with it and vice versa - instead of remaining fixed absolute voltage.

 

[Britton30]

Thanks a lot Gene! I see I'll be learning Intelese.

 

[essenbe]

Yes gary, the first thing you have to do is forget everything you ever knew about AMD. This is a new ballgame even for Intel, which is what Prof was saying. I still don't understand most of it. The only thing I know, is leave the FSB alone, unless you particularly like the color Blue.

 

[Wishmaster]

I would also like to add the Load Line Calibration may be refered to slightly diffrently depending on the Motherboard.

My eVga board for example just has the Options: With vDroop or Without Vdroop.
Where Without Vdroop would be equall to having Load Line Calibration enabled.

It should be pretty easy to figure out though even if its called something slightly diffrent.

 

[smarteyeball]

I don't overclock, what's the point.
Your machine is as fast as she is, leave it at that.

In a nutshell - The point is that you can

There's no denying that at stock speeds, these machines are plenty fast as is for the vast majority of tasks.

I personally do it primarily because I enjoy the challenge. It's a hobby. The added bonus is that through my own trials and errors, I can speak from experience when helping others, which I like to do where I can.

I also play games and I like to be able to run them at optimal IQ and speed. Some games that are either poorly coded, or single core dominant (often both) do benefit from overclocking. FPS dip vs smooth no FPS dip.

Brute force OCing can also often make up for poor coding. Even with a decent single GPU overclock, it can be the difference between 90% GPU usage vs 99% GPU utilization which results in an overall higher minimum and average framerate with higher IQ settings.

One example is GTA IV. It's notorious for being a poorly coded port. Between a GPU overclock and a hefty CPU overclock, I can actually run this game completely maxed out at close to 60FPS + vsync at all times. There is a vast difference between 60FPS and dipping below 30FPS.

Basically, it's a personal choice. If you have no desire not to, don't. If you do - then you can



*

Meh .. just barely bootable. Got into Windows yes ... but not stable enough to even get CPUz (or in mu case evga eleet) open.

But you were right. At 1.4 a x50 Multi was quite unstable for me. Turning Ht off allowed the same result but at x52

Still, all things considered its not bad. Think 4.8 was my last good stable run within my personal vcore limit. Although it was a bit much for me as a 24/7 OC.

I've dropped back to 4.7 for 24/7 because 4.8ghz would sporadically BSOD during low load transitions. Investigations continue

But 5ghz/5.2ghz, even 4.8ghz is really not worth a 24/7 pursuit (unless running multi high end GPU's)

Good for shits and giggles but the cons outweigh the pros.

Hmmm ... My VCCIO is at 1.05V as well. Should it be raised with 1.5v RAM?
Although Im not having stability issues was just curious.

As Gene mentioned, no - you're perfectly fine.

It would be of benefit if you were trying to really tighten your timings etc, but your kit is performing great as is and you aren't going for a high 5.0-5.2ghz OC.

The option to up it is there should you try to OC/ really tighten the timings but as it stands - it's all good J



*

The Intel DC operating spec for VCCIO is 1.05 +/- a couple of percent and t hat is with 1.5V RAM, so you should leave it where it is.

I've done further reading including posts from Asus and G.Skill techs and real world degradation reports from experienced OC'ers on OC forums. The average result is that it's still the vcore at 1.5+ that's the biggest cause for chip degradation. 1.6v is quick, 1.5v not as much.

Mid 1.4 is a mixed bag.

As previously mentioned, a VCCIO of 1.2 is the 'max safe' limit and DRAM volts of 1.7 is the max DRAM limit. My re-reading just reinforces that. Intel specs are primarily intended as a guide for 'zero' degradation issues.

Since only 'hardcore benchers' will be using DRAM volts of 1.7, a VCCIO/VTT of 1.2 is also unnecessary for the average user.

For 1.65v users, a VCCIO/VTT of 1.15 is fine.

For 1.6v users, a VCCIO/VTT of 1.10 is fine

For 1.5V users, a VCCIO/VTT of 1.0 is also fine.

As long as it's within the .5v variance, it's all good. So if you're using 1.5v DRAM and a VCCIO of 1.05, even 1.10 it's under the .5 variance

A lot of the P67, Z68 boards tend to default between 1.0 and 1.10, so it's not an issue for most. Only users of 1.65v kits should take extra care to note what their VCCIO/VTT is set at. These boards also tend to fluctuate, even with a manually applied voltage, so a double check is worthwhile.

Also as mentioned, when running more than 8GB ie 16GB, a higher VCCIO/VTT is recomended to ensure that the RAM runs as close to it's rated speed as possible.

It's the same principle on previous sockets where the more DIMM slots filled, the more NorthBridge voltage was required to maintain stability.



*

I tried some adjustments and passed IBT. Vcore just does not want to go down much it bounces between 1.172 to 1.280

View attachment 200897

But this is the problem and with this cooler should not even be close to this.

View attachment 200898

Actually, the more I think about it, the more I'm thinking it's not as badly performing as first thought.

HT on is an automatic 5c-10c temp increase, especially under IBT.

What is your current/average ambient temps?


As already mentioned, the bouncing vcore is the vdrop (voltage drop) and offset in action. IIRC, LLC 5 is similar to 'off' (following intel spec) so voltage fluctuations are quite normal, particularly when using the offset function.

Ostensibly, the offset function either + or - uses the chips stock VID (the voltage factory set to guarantee for supplying sufficient voltages for the chip to run at rated Stock speeds) as a base voltage which is then adjusted depending on the overclock/load used. A + offset 'adds' more vcore, the - offset reduces the vcore.

In conjunction with the offset setting, LLC adds another layer of fluctuation.

When set to low/off, the vcore fluctuates more. When using a higher LLC, it 'forces' the vcore to maintain a stabler voltage with less fluctuations. The downside to a higher forced LCC is that any 'spikes' can overspike the voltage which has a higher potential to cause damage.

Using a fixed vcore and high LLC helps maintain a more constant voltage. Some boards like the Asus Z68 range actually work beetter with a lower LLC and the offset function.

Depending on target OC, it's better to use a low/no LLC. The trick is managing how high the vcore spikes under high load and how low it drops during the power saving states. Often an idle x124 BSOD crash is caused by insufficient vcore as it cycles through the varying power states. (If enabled)

 

[Britton30]

Would one of you masters have a look? http://www.sevenforums.com/hardware-devices/217062-cpu-running-100-lg-f1.html#post1819387
He's an Intel user and I don't know what to tell him. His issue is that he can't get his SpeedStep to throttle and boost his CPU.

 

[GeneO]

As previously mentioned, a VCCIO of 1.2 is the 'max safe' limit and DRAM volts of 1.7 is the max DRAM limit. My re-reading just reinforces that. Intel specs are primarily intended as a guide for 'zero' degradation issues.

You should not even go near those voltages you mention IMO. If you go out of the range of 1.05 +/- 3% then you can shorten the lifetime of your processor. I wish people would get this (not referring to you but whoever said that was safe) - staying in Intel's absolute maximums doesn't mean it is safe, it means it won't immediately croak. What is safe is staying within the DC operating ranges. The Intel DC operating range for DRAM voltage is 1.425 to 1.575 for Sandy Bridge.

In fact, where do those safe 1.2V VCCIO and 1.7 v DRAM voltage numbers come from? Intel apparently doesn't even publish absolute maximum voltages in their spec anymore (I can speculate why), just DC operating, so somebody probably pulled those numbers out of their butt or are basing them on the previous gen maximums, which were for 45nm chips (vs 32nm Sandy Bridge) and could probably stand more voltage. Since the processor hasn't been out but a year - nobody can tell you for sure what are the long term affects of operating out of the DC window on voltages.

Except for the Vcore, which has a wide range, I think it is best to stay within Intel's DC specs for everything else and to avoid using LLC if you can, because that is also out of spec.

 

[smarteyeball]

The 1.2 VCCIO and 1.7 VRAM has come from internal testing by certain companies ie G.Skill, Asus and even intel. It also includes findings from the serious benchers.

As you say, the chip is less than a year old and nobody knows for certain what the true max safe limits are. I personally wouldn't run my RAM and VCCIO that high. It's also why I'm working on lowering my DRAM volts.

If the DRAM voltage absolute max is 1.575, then there's quite a few RAM manufacturers up for some serious class action lawsuits since there are a wide variety of 1.65V kits available

staying in Intel's absolute maximums doesn't mean it is safe, it means it won't immediately croak. What is safe is staying within the DC operating ranges. The Intel DC operating range for DRAM voltage is 1.425 to 1.575 for Sandy Bridge.

I agree mate. However for most overclockers, if the intention is to keep a chip for 10+ years, then they wouldn't be OCing to begin with

somebody probably pulled those numbers out of their butt or are basing them on the previous gen maximums, which were for 45nm chips (vs 32nm Sandy Bridge) and could probably stand more voltage. Since the processor hasn't been out but a year - nobody can tell you for sure what are the long term affects of operating out of the DC window on voltages.

Also agree. Part of that is because because intel don't release the limits any more

However part of my reading has encompassed threads with users who who intentionally pump X amount of dangerous volts on various settings 'just to see' what causes what and what causes degradation the quickest.


The majority are using 1.5V kits anyway, so staying within intel specs is a moot point as they already are. For others like myself who technically fall outside of the intel spec, I'm simply trying to offer advice that helps minimize any 'quick degradation'.

and to avoid using LLC if you can, because that is also out of spec.

Apart from some playing around at 5.0ghz+, I personally don't use LLC either.



It really comes down to the "if you can't afford to overclock - don't". But if you're going to take the risk, minimize where you can.

 

[profdlp]

I'm putting in a request for smarteyeball to write a Tutorial on this. I promise to rep you 'til the cows come home if you do.

 

[smarteyeball]

lol, I'll give it some consideration mate

 

[profdlp]

I'm dead serious. You are good at explaining this stuff.

I, on the other hand, am good at volunteering other people to do things.

 

[GeneO]

The 1.2 VCCIO and 1.7 VRAM has come from internal testing by certain companies ie G.Skill, Asus and even Intel. It also includes findings from the serious benchers.

The RAM vendors just want to sell you RAM. They don't care what voltage you run your processor at, that is your risk. And serious benchers go through processors like candy I expect - no skin off their nose. I'd like to see that quote from Intel - I stand by my case to be prudent unless you can provide it.

And I am not talking a 10 year life either - I expect it is much shorter if you overvolt too much 24x7.

But really you should not have to go out of the operating ranges except for extreme overclocking if you have the right RAM because Intel seems to have decoupled the RAM from the overclocking equation - as long as you leave BCLK alone. So there is no reason to propagate unofficial "safe" voltages for Vccio and the dram voltage when you don't really need to change them. These you mention are significantly out of the operating range in the Intel spec for Sandy Bridge.

 

[essenbe]

I'm putting in a request for smarteyeball to write a Tutorial on this. I promise to rep you 'til the cows come home if you do.

 

[smarteyeball]

I'm dead serious. You are good at explaining this stuff.

I, on the other hand, am good at volunteering other people to do things.

Thanks mate, and lol

The RAM vendors just want to sell you RAM. They don't care what voltage you run your processor at, that is your risk.

Well obviously there is an element to that. Believing otherwise would be foolish, but I'm sure by now there'd have been a massive hue and cry as thousands of customers have dead components caused by unsafe products.

And serious benchers go through processors like candy I expect - no skin off their nose.

I use benchers as an example because as you say they go through components like candy, but through their endeavours the rest of us can glean an insight of what extreme settings can do and for general use thus avoid.

I'd like to see that quote from Intel - I stand by my case to be prudent unless you can provide it.

If I'd bookmarked the thousands of posts discovered during casual browsing on the off chance of providing proof, I'd be more than happy to do so. I don't wish to sound argumentative or offensive, But since I don't bookmark everything on the off chance, you can remain prudent on your stance.

I do tend to advocate sticking as close to specs as possible you know, just as much as I tend to point out possible ramifications

And I am not talking a 10 year life either - I expect it is much shorter if you overvolt too much 24x7.

10 years was just an arbitrary number.

Even if you shorten a components life span by a few years, most overclockers are willing to do so. A lot of overclockers will have upgraded long before serious noticeable degradation has occurred.

Again, if you can't afford to replace something - don't do it in the first place.

But really you should not have to go out of the operating ranges except for extreme overclocking if you have the right RAM because Intel seems to have decoupled the RAM from the overclocking equation - as long as you leave BCLK alone. So there is no reason to propagate unofficial "safe" voltages for Vccio and the dram voltage when you don't really need to change them. These you mention are significantly out of the operating range in the Intel spec for Sandy Bridge.

No you shouldn't have to, but for many years people have, and will continue to go beyond the recommended ranges.

And like the "I've been running my XXX at blah for years with no problem and the "My system died because of blah" - there's always many other variables that factor into the whole equation.

I prefer not to propagate myths, but in the absence of conclusive definitives, erring on the side of a positive general consensus is better than nothing.

In this instance, I thinks it safe to agree to disagree wouldn't you say?

 

[Britton30]

Also serious benchers have access to chips and RAM we peons don't. either they or the vendors supply hand pick components to use. Some people with $$$ out the wazoo can but several chips to find a specific binning they know is a good speed demon.

 

[GeneO]

If I'd bookmarked the thousands of posts discovered during casual browsing on the off chance of providing proof, I'd be more than happy to do so. I don't wish to sound argumentative or offensive, But since I don't bookmark everything on the off chance, you can remain prudent on your stance.

Even if you shorten a components life span by a few years, most overclockers are willing to do so. A lot of overclockers will have upgraded long before serious noticeable degradation has occurred.

Again, if you can't afford to replace something - don't do it in the first place.

No you shouldn't have to, but for many years people have, and will continue to go beyond the recommended ranges.

I very much doubt an Intel person ever stated that, I am sure through all the multitude of reading you have done you misremember. Even if they thought it they would not state it publicly in writing.

As for core overclockers, sure you can give advice about what you or someone you have read think are safe voltages, even though no one can know the affect they will have on the lifetime for more than about 6 months to a year. The only voltages that can be said to be "safe" are those in the Intel spec. The only thing that can be said about voltages beyond those are that the processor will run stably for some period of time. I am not suggesting people shouldn't run their processors beyond the Intel spec voltages- just that they know the risk and labeling them as "safe" just means that your processor won't roll over on you for some unknown period of time. I think the audience here is a little different than core overclockers.

Sure for many years people have overvolted, and a lot of them have cooked their processors pretty quickly too I am sure. Running at The lower voltages become more critical as the dies shrink - smaller components and traces require lower voltages. That 32 nm Sandy bridge runs at lower voltages than the previous generation is part of the reason they can overclock higher than the previous gen right out of the box -they generate less heat because of this. OTOH, that also means the last generation "safe" voltages don't apply.

I would be less worried about the DRAM voltage than increasing the Vccio. It is too bad Intel and the memory makers won't get together ion the memory voltages.

 

[essenbe]

I just think that the words safe and overclocking used in the same sentence is an oxymoron. Everyone knows, or should know, that running something faster than specs is taking a risk. The question then becomes, how big a risk are you willing to take. What amount of risk do you deem acceptable. To me, it is as simple as that. If you want to be safe, don't overclock. I've said on these forums many times, 'if you are not willing to replace it, don't overclock'. How much risk you are willing to take is a personal decision. Ram manufacturers are making ram approved by them for sandy Bridge at 1.65V. While I agree they want to sell ram, I don't believe they would open themselves up to potential litigation. Motherboard manufacturers are making boards that allow you to overclock the ram and CPU and making it easy for you to do it. I am sure they have taken these same considerations in mind. I overclock and am willing to take the risk, up to a point. Whatever that point is for everyone else is a personal decision. But, I would like to see a comprehensive guide that tells me how to do it. I already know what is stock, I'll gage the risk for myself.

 

[GeneO]

I just think that the words safe and overclocking used in the same sentence is an oxymoron. Everyone knows, or should know, that running something faster than specs is taking a risk. The question then becomes, how big a risk are you willing to take. What amount of risk do you deem acceptable. To me, it is as simple as that. If you want to be safe, don't overclock. I've said on these forums many times, 'if you are not willing to replace it, don't overclock'. How much risk you are willing to take is a personal decision. Ram manufacturers are making ram approved by them for sandy Bridge at 1.65V. While I agree they want to sell ram, I don't believe they would open themselves up to potential litigation. Motherboard manufacturers are making boards that allow you to overclock the ram and CPU and making it easy for you to do it. I am sure they have taken these same considerations in mind. I overclock and am willing to take the risk, up to a point. Whatever that point is for everyone else is a personal decision. But, I would like to see a comprehensive guide that tells me how to do it. I already know what is stock, I'll gage the risk for myself.

Well I agree with you. I guess my quibble was with the use of the words "safe". I don't know how the RAM manufacturers are open to litigation when you prob. couldn't prove what was the cause of any long-term damage.

Ditto on the risk - I try to minimize it the best I can having degraded a processor with what many declared a "safe" settings.

 

[profdlp]

... I don't know how the RAM manufacturers are open to litigation when you prob. couldn't prove what was the cause of any long-term damage...

I agree. We've probably all been through the buck-passing where the MB manufacturer blames the RAM company who blames the PSU maker who blames...

What I think the RAM companies do pay attention to is their reputation in the enthusiast community. If you were researching an upcoming purchase and saw tons of forum posts raving about RAM "A" and how stable and reliable it was, then saw lots of griping about RAM "B", which would you buy?

This doesn't help the early-adopters too much, but they usually realize that they are both paying a premium price and taking on the added risk of being the real-world guinea pigs for cutting edge technology. That's probably why we haven't seen a lot of X79 platform questions yet.