Intel Core i5 & i7 Sandy Bridge Overclocking and Feedback

[Chosen Juan]

try disabling the C1E or C3 states in the bios.

or setting the power options within windows for full power usage. instead of balanced by default.

if its not either of those then i dont know. I havent really played around with the bios since i only put this rig together thursday just gone

C1E, C3, C6 and Package C are all disabled in the BIOS...hmm...oh well it's not a big feature for me but it would have been nice.

 

[FreedomEclipse]

Sorry, cant help you then. as soon as i built this pc i went straight into the advanced settings and OC'd this beast. didnt bother checking out all the other settings.

I had blinkers on. everything else there didnt matter other then Vcore and multiplyers.

-------

It could be something to do with with the Turbo feature. and again im drawing at straws here as i havent made any effort to have a poke around the bios extensively

 

[NdMk2o1o]

So got a brand new i5 2500k this morning, rig is up and running already, I don't waste time

Now, how do I get this baby to 4.5ghz 24/7? is it just a case of raising the multi and vcore? anything else I need to check on, disable etc.

What about turbo and power saving etc on or off?

 

[Chosen Juan]

Sorry, cant help you then. as soon as i built this pc i went straight into the advanced settings and OC'd this beast. didnt bother checking out all the other settings.

I had blinkers on. everything else there didnt matter other then Vcore and multiplyers.

-------

It could be something to do with with the Turbo feature. and again im drawing at straws here as i havent made any effort to have a poke around the bios extensively

Ahh it's alright, thanks. I'll worry about that later I guess, my main focus now is just getting it stable on low voltages. The best I have so far is 4.2Ghz on 1.2v, but that's only on 50 passes of IBT on Standard. Have yet to try Prime95 and gaming on it.

 

[FreedomEclipse]

So got a brand new i5 2500k this morning, rig is up and running already, I don't waste time

Now, how do I get this baby to 4.5ghz 24/7? is it just a case of raising the multi and vcore? anything else I need to check on, disable etc.

What about turbo and power saving etc on or off?

try for 5Ghz and if you smoke it, send it back and say it was sending smoke signals on arrival and if they fob you off...drive up to their HQ and make their receptionst eat it like shredded wheat LIKE A BOSS

Do it!

-----


as for help with overclocking I followed this guide by bit-tech It was designed for the p67 chipset but the p67 and Z68 use more or less the exact same bios. I used this guide and i hit 4.9ghz in no time. complete build of and super oc in less then 6hrs.

 

[MakubeX]

i5 2500K
5GHz 12hrs Prime Stable @ 1.368v
Batch: 3101A156
Mobo: P8Z68-V Pro
Watercooling, room temp avg: 29C

 

[cadaveca]

Nice one, Makubex. is that with or without PLL Overvoltage?

 

[MakubeX]

Nice one, Makubex. is that with or without PLL Overvoltage?

Without, I only needed to change the vcore for this OC. Every other voltage is on default (mostly Auto). You gotta love Sandy Bridge.

 

[cadaveca]

Yeah, seems you got a good chip, there are many others out there not doing so well. I'm kinda jealous, actually.

Volts are reasonable, clocks are sweet, what is your IMC volts?

 

[MakubeX]

Not sure which one is IMC. Here's a screenie of some of the voltages:

The CHB CTRL voltage not shown is also 0.5v.

 

[cadaveca]

VCCIO is the voltage I was asking about. Thanks!

 

[NdMk2o1o]

holy shit, the fun of overclocking has been brutally buttraped and murdered by Intel all I did was go into my shiny uefi bios, click on cpu multi, set to 45 and set vcore to 1.325 and voila, 4.5ghz IBT/P95 stable at less than 60c

Seriosuly though I have never had a cool chip, I had seen people with oloder i7's getting 4ghz at 70c though every older i5/i7 I have ever had has run at 80c+ OC'd but low and behold the sandybridge never goes above 58c at 4.5ghz

 

[NdMk2o1o]

holy shit, the fun of overclocking has been brutally buttraped and murdered by Intel all I did was go into my shiny uefi bios, click on cpu multi, set to 45 and set vcore to 1.325 and voila, 4.5ghz IBT/P95 stable at less than 60c

Seriosuly though I have never had a cool chip, I had seen people with oloder i7's getting 4ghz at 70c though every older i5/i7 I have ever had has run at 80c+ OC'd but low and behold the sandybridge never goes above 58c at 4.5ghz

Edit: oh and this SB beats the shit out of my old i7 860 OC to 4ghz at super pi when running stock 3.3ghz

 

[cadaveca]

Edit: oh and this SB beats the shit out of my old i7 860 OC to 4ghz at super pi when running stock 3.3ghz

Yeah, the memory performance boost IS pretty damn impressive.

 

[Chosen Juan]

Hey guys, I'm in the process of overclocking my 2500K right now and I was just wondering what settings I should change and what to leave (I've a read a few guides so far).

These are my settings right now:
CPU Ratio - 46
Internal PLL Overvoltage - Enabled
Intel SpeedStep Technology - Enabled
BCLK - 100.0Mhz
Spread Spectrum - Disabled

DRAM - XMP Profile 1

Power Saving Mode - Disabled
CPU Core Voltage - 1.285
CPU Load-Line Calibration - Level 1 (for my board, Level 1 overvolts at load, while Level 2 keeps it in sync, and the higher Levels up to 5 decrease at load)

And everything else is at stock. Should I be messing with CPU PLL Voltage, VTT Voltage, VCCSA Voltage or any of that? Or should I just stick to multipliers and Core volage right now?

Thanks

EDIT: Also, I have Enhanced Halt State (C1E), CPU C3 State Support, Package C State Support, and CPU Thermal Throttling Enabled, those shouldn't affect my overclocks should they?

 

[Frizz]

Is this true?

This person from Hardforum makes a good point and gives me second thoughts about running my CPU above 1.4vcore.

http://hardforum.com/showthread.php?t=1578688&page=2

1.52v is the max VID. VID is not the same as vcore, yet people seem to somehow think it is. VID is a voltage identification register. The chips are DESIGNED under VRD 12 to not only request a VID, but also to NOT be run under loadline calibration. VDROOP is part of the VRD 12 specification. Run a CPU at 1.52v set in BIOS, without any LLC, and under max load, it will drop down to 1.39-1.40v.

This is by design and intended under Intel's thermal specifications. So no, 1.52v is NOT the maximum LOAD VOLTAGE the cpu is supposed to be run under. It's more like 1.39-1.40v.

I challenge ANYONE to prove me wrong. Be prepared to present VRD-12 documents as proof, if you think VID is the actual vcore going to the chip and 1.52v is the maximum VCORE.

When I made my original post here, I did not take loadline slope into consideration, but now it's completely blatantly obvious that VID was designed also with Loadline (vdroop) in mind. Remove vdroop and you are instantly running the cpu's out of specification once you go above 1.4v. Doesn't mean you can't do it, but a lot of people are getting degradation at these voltages now. Some aren't, as every cpu is different, but some are. Some are even getting it at 1.3'isih volts, but the difference is, the cpu's become stable and stop getting any worse, a lot sooner.

 

[NdMk2o1o]

I am running stable now at 4.5ghz with speedstep and C states enabled, drops down to 1.5ghz on the desktop and 4.5ghz when I ramp up the CPU usage, seems the easiest way to keep power consumption down and also kick ass in games/video encoding. I am running 1.3vcore in the bios, haven't touched PLL, Vdroop or anything and haven't experienced any crashes whatsoever. Runs cool at 28-30c idle and 58c with all cores loaded 100% at 4.5ghz

 

[cadaveca]

This person from Hardforum makes a good point and gives me second thoughts about running my CPU above 1.4vcore.

http://hardforum.com/showthread.php?t=1578688&page=2

Yes, of course. Why do you think I say loadline calibration should be left @ default, and Turbo should be used to OC? The amount of droop under GTL is a factor of the current consumed, so you should almost never actually be @ the votlage the VID references, due to how the VRM is designed.

When you change, or enable "loadline calibration", you are running the CPU out of spec. Of course, the documents also consider reference cooling, etc, too.

Just like how I'm one of few that hype actual default memory speeds, although boards and such offer higher or more settings; there is alot of info about how things work, that is widely ignored, because it gets users more speed/power. That's why I do not go really high in volts in my reviews, and why I use the speeds that I do for testing. I try to keep things reasonably within spec, and by doing so, I should also be posting scores and such that anyone can match, and verify.



I mentioned that for clocking memory, for me, it seemed like skews needed adjsuting, then I found this:

Since writing this quick article, I have realised that the skews can come into play for even on air and water cooling with certain memory configurations - so it actually needs an update with some examples to reflect this. For example, in the GSkill 2133 guide I posted on this forum at 4.5GHz, the memory would not pass Memtest without the CPU skew advanced by 100ps, and that was at base of 100BCLK. Makes sense as the Intel stock layout is designed to be within functional tolerance margins up-to the supported memory speed of DDR3-1600 and at stock processor frequency.

For benchmarking scenarios over 5GHz, things become even more critical, and time spent with clock skews can give an edge.

Here's what optimal skew settings can do for Elpida Hyper when the CPU is cold (the setting will vary on a case by case basis):

http://www.asusrog.com/forums/showthread.php?2163-Overview-of-Clock-Skew

I posted about 3 weeks before Raja did. I don't think he read my post...we just came to the same conclusions.

There's alot more to clocking than simply setting voltages, and upping clocks...

 

[BababooeyHTJ]

I'm about to make the jump to sandybridge and have been doing a little research in the process and am wondering why so many people seem to be pumping these 32nm chips with the types of voltage that most people would think twice about using on a last gen 45nm cpu like lynnfield, especially after seeing clarksdales dropping like flies.

 

[cadaveca]

I'm about to make the jump to sandybridge and have been doing a little research in the process and am wondering why so many people seem to be pumping these 32nm chips with the types of voltage that most people would think twice about using on a last gen 45nm cpu like lynnfield, especially after seeing clarksdales dropping like flies.

I'm not pushing more...in fact, I'm using far less. Not too sure why others are doing exactly as you say...but I bet it's because of a combination of factors:

1. Intel says max VID is 1.525v
2. Temps prevent most users from actually going that high.
3. Many people, myself included, are recommending to not go over 1.425v for 24/7 OCs.

1+2+3= what's going on?

 

[Frizz]

Hmm I've never tried turbo ocing before but here goes nothing.

I currently have my voltages all set to auto, all I've changed are the turbo speeds for all 4 cores to 45 and then I set the turbo power limit to 135.. Everything seems to be working great as the auto voltages seem to do ok with finding the stable load voltage of my CPU without overshooting it...

LLC is disabled

Anything I'm missing?

 

[BababooeyHTJ]

I'm not pushing more...in fact, I'm using far less. Not too sure why others are doing exactly as you say...but I bet it's because of a combination of factors:

1. Intel says max VID is 1.525v
2. Temps prevent most users from actually going that high.
3. Many people, myself included, are recommending to not go over 1.425v for 24/7 OCs.

1+2+3= what's going on?

I thought that the rule of thumb was the 1.4v was about what most people would try not to exceed on air since Wolfdale and Yorkfield which have proven to withstand more abuse than intel's 32nm cpus. I've been seeing too many threads and random reports like this. It's as if people are just pumping these cpus with the types of voltages that they have been for the past three years with Intel's 45nm processors without batting an eye.

 

[cadaveca]

I think you must consider the source of your info...not that the info there is invalid, but you'd expect as much from there. People chasing the highest clocks and benchmark scores are gonna kill stuff; that's just a fact I've learned to accept over the years.


But like I said...Of course, I don't use PLL overvoltage, don't use more than 1.35v, myself, and no LLC. Both of my CPUs do 4.5 GHz @ 1.3v or less(like @ stock volts), and I don't see much need for more. I'm not posting 5.5 GHz++ scores, with 2 cores, and HT off either, so take that for what you will.

Anything I'm missing?

Not really. Looks like what I see most often, so seems good to me. You might need to up VCCIO for memory pseeds, but even then...not by much.

 

[Frizz]

You might need to up VCCIO for memory pseeds, but even then...not by much.

hmm not sure how to do that on my board

EDIT: nvm it is QPT/VTT voltage, I'll go on and do that now.

 

[cadaveca]

well, that's for mem, so if ya don't need it, don't give more! 1600MHZ mem and up only need more.

Tirbo OC's are literally that easy. Sad, but AWESOME!!!111!!!