Overclocking my 4770k

[wrenaudrey]

i7 4770k
Sabertooth Z87
8GB Dominator 1866
EVGA GTX 780
Samsung 830 128GB
Cooling
EK Supremacy CPU Block
XSPC Titan Block
Alphacool D5
XSPC D5 Top
Phobya Balancer 150
Alphacool UT(stands for 60mm thick for those who doesnt know) 360 and 240
5x SP120 all in a PULL config
Windows 7 Ultimate 64bit SP1

I've had my system(full system with watercoolong) for couple of weeks now and the temps are not too bad.
Now i want to try overclocking it. At the moment everything is on AUTO, the only setting ive changed is to set the XMP profile for my memory.

Before i want to overclock, i want a couple of questions answering.
So i've watched TTL's overclocking guide on 1155 and he said there to establish the most base(lowest) vcore. I've been trying to do this with failure atm and have been running to some weird OS issues, i say issues because i've never encountered this before.

So the question is, DOES THIS NORMALLY HAPPEN?
I dial in a vcore eg. 0.950, press F10(save & exit). Then it POST's, tries to BOOT, get to the Window animation thing, completes it then stays there for a bit then BAMM! It restarts, gets to the same point BUT this time it boots into Windows Repair or Startup Repair or whatever its called.
Obviously i repair it, then after that it restarts but happens again.
This kept happening until i changed the BIOS into default settings, so pressuming its the vcore. Does this really happen if you fail undervolt pr is it just me?

Also as i said i needed to reset to default BIOS settings, sometimes i get into the BIOS but then if i dont move fast enough, it freezes :S
Is this normal too or just me?


I want to establish my base vcore because if i dont see a big diff when ive OC'd ill come back to this base vcore and run stock frequency.
So if you guys could help me with this problem first before helping me further with pushing my 4770k
Thanks

 

[FTLN]

Hve you dropped your memory down 1333 / 1600 ??

Faster memory speed requires more volts on the cpu in my experience...

 

[wrenaudrey]

Hve you dropped your memory down 1333 / 1600 ??

Faster memory speed requires more volts on the cpu in my experience...

No i have not. Perhaps i will later.

But what im trying to get from this first post is to establish whether what i am experiencing about the Windows Startup Repair normally happens when an OC has failed. I really dont want my SSD to get broken or for my windows to be corrupted as reinstalling windows is just to fussy.

 

[FTLN]

If windows detects a bad start / restart then its pretty normal it will launch up startup repair.. Dont worry to much about it... Make a system backup if you feel you must...

 

[NeverBackDown]

Your vcore is too low so your system is unstable. Try doing this..

set the multiplier(should be in MHZ) to stock at 35(some motherboards make you do 00 after so if so then 3500 is fine) and at 1v. If windows works then go back to bios and try the next one down in volts which is most likely .995, if that works repeat the process until windows will not boot properly or something abnormal happens. If you find that then go up 1 step (ex .995 to 1v) and then test the system for stability. If it works then that's the lowest volt possible for that clock.

 

[wrenaudrey]

If windows detects a bad start / restart then its pretty normal it will launch up startup repair.. Dont worry to much about it... Make a system backup if you feel you must...

System backup? Is this the whole windows, programs and files?

Your vcore is too low so your system is unstable. Try doing this..

set the multiplier(should be in MHZ) to stock at 35(some motherboards make you do 00 after so if so then 3500 is fine) and at 1v. If windows works then go back to bios and try the next one down in volts which is most likely .995, if that works repeat the process until windows will not boot properly or something abnormal happens. If you find that then go up 1 step (ex .995 to 1v) and then test the system for stability. If it works then that's the lowest volt possible for that clock.

Ill try this out, definitely.
But in your experience of OC'ing, have you come across my experience of windows repair?

 

[NeverBackDown]

Nope. I checked what my vcore was before overclocking and when under load. Then a made a calculated descision on the vcore.

Here is this..http://www.youtube.com/watch?v=mBEeXajbG2o

Edit: same applies for 1150

 

[wrenaudrey]

Your vcore is too low so your system is unstable. Try doing this..

set the multiplier(should be in MHZ) to stock at 35(some motherboards make you do 00 after so if so then 3500 is fine) and at 1v. If windows works then go back to bios and try the next one down in volts which is most likely .995, if that works repeat the process until windows will not boot properly or something abnormal happens. If you find that then go up 1 step (ex .995 to 1v) and then test the system for stability. If it works then that's the lowest volt possible for that clock.

Nope. I checked what my vcore was before overclocking and when under load. Then a made a calculated descision on the vcore.
Edit: same applies for 1150

So i've followed your method and watched TTL's vid again.
I've found that my system was stable at 0.945v manual. I ran OCCT for atleast 30min.

I've ran out of time and have to go to work and i want to still run it at stock at the moment with my newly found stable undervolt vcore.
Now i've watched several vids on Haswell and they say to put it on adaptive mode after an OC. However, if i change it to adaptive mode i cant seem to enter a 'maximum' vcore like in manual. Is this correct?
Should i set a manual vcore first (0.945v) and then set it to adaptive and it follows that 'maximum' vcore inputted when on manual mode?


Edit:
I've set 0.945v on manual and then set it to adaptive. Now booted into windows, just ran OCCT for a confirmation on the vcore and it reached 1.172v at one point. Here's a screeny:

Is this meant to be happening? Cause i have set a manual voltage and i dont really want it going over that, hence the point of undervolting.



Ediiiiiiiiit:
I've been reading more into adaptive mode and people have said and i have tested that due to the AVX on OCCT it is forcing the voltages to go crazy above what i set. However apparently if you're not putting the CPU into heavier than normal loads(stress tests), you're suppose to get what you set (0.945v) BUT NO(in my case). One of the example i saw was that if you ran like rendering and those stuff you would apparently get the volts that you inputted but like i said in my case, it does not.
I used WinX Video Converter and tried to convert this random film i have into a H264 as this would 'simulate' rendering. And what i found was that max voltage is reach was 1.056v, that's 0.111v above the supposedly maximum that i entered which is 0.945v

Can someone help make my max voltage on load the same as what i entered?

 

[NeverBackDown]

I don't know much about Haswell OC simply because i have no exp with them. If adap mode makes your volts go higher than whats the point in using it? If you can keep it at .945 24/7 stable then mine as well go for that.

 

[Master&Puppet]

But in your experience of OC'ing, have you come across my experience of windows repair?

Yea, I see that quite often - if you set the voltage too low and windows fails to boot then the next successful boot will often automatically send you to windows repair since it thinks something has broken. Just restart and carry on.

There are 3 different types of overclocking you can do on Haswell,

Offset,
Adaptive,
Manual.

The one that tom illustrates in the video is for manual where you simply tell the CPU what voltage to work from.

Offset and Adaptive work from Intel's default voltage curve which defines the amount of power supplied to the CPU for any given speed. An offset OC simply increases the voltage curve across the whole range. So if you set +0.05v then the whole curve will shift up in power supplied. Adaptive is a refinement of offset where it will change the power supplied but only at the righthand/higher clock speed end of the voltage curve. That means you get extra power where you need it but you don't add volts at the idling end.

That's a very quick introduction which might not help too much. what I will add though is that I'll be writing a Haswell OC guide this week I hope so you might want to look out for that

 

[ahpadt]

Only do OCCT when you've set the voltage to fully manual. I don't know all the details regarding Adaptive, but it's recommended only to enable Adaptive once you've done all your stability tests and finally have found a satisfactory voltage, because as you have seen, with Adaptive voltage, the voltage may fluctuate way, way above what you've actually set if you run stress tests.

When I enable Adaptive I set the offset to 0 and just stick the voltage I want as the additional/total.

I also tried to follow the 1155 OC guide in the beginning, but after reading up at a few other places, it sort of brought me to a conclusion that there are some things that had to be enabled or disabled with previous chipsets that are no longer required. For example, I played around a lot with load line calibration and to be honest the voltage fluctuation is already tiny with load line calibration set to auto, so I dont see the need to enable it.

 

[wrenaudrey]

Yea, I see that quite often - if you set the voltage too low and windows fails to boot then the next successful boot will often automatically send you to windows repair since it thinks something has broken. Just restart and carry on.

As i watched TTL's guide again I noticed that he just skipped the repair so i just did that when it happened again on mine.

That's a very quick introduction which might not help too much. what I will add though is that I'll be writing a Haswell OC guide this week I hope so you might want to look out for that

And yes, i look forward to reading it when you publish it

Only do OCCT when you've set the voltage to fully manual. I don't know all the details regarding Adaptive, but it's recommended only to enable Adaptive once you've done all your stability tests and finally have found a satisfactory voltage, because as you have seen, with Adaptive voltage, the voltage may fluctuate way, way above what you've actually set if you run stress tests.

When I enable Adaptive I set the offset to 0 and just stick the voltage I want as the additional/total.

I also tried to follow the 1155 OC guide in the beginning, but after reading up at a few other places, it sort of brought me to a conclusion that there are some things that had to be enabled or disabled with previous chipsets that are no longer required. For example, I played around a lot with load line calibration and to be honest the voltage fluctuation is already tiny with load line calibration set to auto, so I dont see the need to enable it.

That's what i've read on adapative too, that just set it once you've got your stable voltage. But my problem is that i set 0.945v on adaptive i play a game like bf3 and it rises to 1.072v and my temps go crazy. 50-55C on manual on 0.945v, 1.072v on adaptive and it reaches up to 70C :mellow:



Anyway an update. I set 0.945v manual, it was stable 30min on OCCT so i thought it was okay. The next day, was playing bf3 for about 7hrs straight and then suddenly, BOOM! a BSOD. So i increased it to 0.950v and its been okay since.
Im currently 0.950v on adaptive and its okay so far.

My question now is that, does the CPU degrade/die much faster when you're feeding 0.950v constantly(manual mode) OR feeding, say 1.4v but adaptively; has the ability to go down on volts when idle and maxes 1.4v when loaded?

 

[NeverBackDown]

It only dies faster if you over volt or push it to its limits all the time. 1.4 v for any CPU an extremely high.. seriously stick to manuel if adapt is hitting those volts.

 

[wrenaudrey]

Okay maybe not 1.4v Let me clear it out.
Will the CPU die faster when ran 0.950v(manual mode) for 14hrs straight, 6 days a week OR in adaptive mode when ran 1.072v for 7-8hrs(when gaming) and 0.704v(idle,browsing,music) for 5-6hrs, 6 days a week?

 

[NeverBackDown]

Difference will not be noticeable. Anything above 1.25ish volts for most chips will make an impact.

 

[wrenaudrey]

Difference will not be noticeable. Anything above 1.25ish volts for most chips will make an impact.

Can anybody else confirm this? Before i run my CPU on manual mode 24/7
Dont really want it dying, as my planned upgrade cycle is atleast 3 years.

 

[NeverBackDown]

It will last 3 years.

 

[ahpadt]

In very general terms, no electronics will like significant undervolting or overvolting (from an electrical engineering student standpoint). I wonder what the theoretical minimum Vcore Intel recommends for Haswell is.

In terms of how low you can go; my motherboard (see sig) won't even allow me to set a manual voltage below 1V. This begs a question of whether 1V maybe could be a safety net for guaranteeing stable performance for Haswell? I don't know.

The main advantage, as I see it, with adaptive voltage is that the CPU will undervolt itself automatically every time you have low CPU load, which ultimately helps the temperatures and could to some degree aid longevity.

EDIT: With adaptive voltage at 1.185V, mine occasionally jumps up to 1.32V when I load it. I just set LLC to level 8 but it seemingly made no difference. I'm confused about how the voltage regulation works too.

 

[NeverBackDown]

When we say undervolt we mean below stock volts at same stock clocks.. Not bad to do and will actually improve the life of the cpu.