Intel Haswell 4670k + 4770K Overclocking guide

[Doug2507]

See my previous post.

 

Once a little information is provided it will be a lot easier to help you! Due to high temps you are seeing i would strongly advise to post this information before continuing with the O/C!

 

I've just noticed you have a 80i for cooling. I've not used them myself but am aware of problems with them. Did you use washers on the back plate for it? If not there's a high chance it won't be getting good contact with the CPU IHS. I'd be willing to put money on the 80i not being seated properly as the cause of your high temps. Once that's sorted then it's time to O/C!

 

Trying not to jump ahead but to give you some pointers....

 

Install CPU-Z

Install HWmonitor

Install Bluescreen viewer

 

 

BIOS:

 

Go onto the MSI forum and download/install the latest BIOS for your mobo following the instructions and using the MSI installation tool. (if not already done)

 

Try setting the following in BIOS:

 

Base clock: 100mhz

Base clock strap: 1.0

CPU PCIE PLL: LC PLL

Filter PLL: Enabled

CPU Ratio Mode: Fixed

Legacy Tweaking: Disabled

Ring Ratio: 33

DRAM Reference clock:auto

DRAM reference frequency: 1333mhz

XMP: Disabled

Memory fast boot: Disabled

Digital All Power: Vdroop offset control: 100%

VCCIN: 1.9v

Core voltage mode: Override (can change this to adaptive later)

Core voltage: 1.2v

Ring voltage mode: Override (can change this to adaptive later)

Ring voltage: 1.15v

DRAM voltage: Set to whatever the kit is designed to run at. 1.5v

Everything else left on auto.

 

Boot with x44 core multi.

 

Run XTU Bench. (make sure 'active core count' is displayed on the graph and time set to 2 minutes). If it crashes, drop core multi by 1 and reboot. Take note of the BSOD code using Blue Screen Viewer after reboot and post here.

 

If it doesn't crash, run XTU stress for 9hrs. As soon as you've started XTU stress, use the snipping tool in the startup menu and take a full screenshot with XTU and HWM on screen, make sure both the clock down at the bottom right of the monitor and the start time of XTU are visible. If it doesn't crash then we're good. If it does, drop the core multi, reboot, open BSV, again take a note of the BSOD code but also take not of what time it crashed. Try again.

 

Follow these steps and we should be able to get you started. This will take a LOT of time to do right. (it took me the best part of 3 full weeks 24/7 to get my core to max!).

 

IMPORTANT: before doing any of this you need to sort out your cooling!!!!

 

If Prokon would rather have you try if a different way then cool, this is just a suggestion! There are many ways to skin a cat!

[LemonJoy]

There's two things that will degrade the cpu, one is heat and the other voltage. What is 'safe' is anyones guess right now as there isn't much data yet on what is failing. Having said that the general consensus seems to be about 1.4-1.45v core, 1.35-1.4v uncore and around 2.2-2.3 vrin. All depends on who you listen to. As such, if running high voltage and it doesn't boot one day well thats that with nobody to blame but yourself.

 

I'm running 1.35vcore (1.376v actual), around 1.2v uncore at the moment (bringing uncore up) and will quite happily push it to 1.35v - if needed, and vrin at just under 2.2v. If it dies in a couple of years so be it as new tech will be on the go. I think when you start failing at what 'were' stable settings and need to reduce voltage/multi it's time to tame things down if you want it to last.

 

IMO what you have is perfectly acceptable.

 

*EDIT - Even on IBT max i only go into mid 70deg. XTU under 70deg. That's with delid and custom W/C. The way i see it, no matter what voltage you're running, the lower the temps the better.

 

thank you for your opinion,

so i stay with my overclock because i`m happy with that, also i haven`t delided so my temps on stress goes to 80c, while gaming 60~70~, `im using h100i and big pc case with good ventilation

 

and also agree that if it dies within 2-3 years that would be a good reason for upgrade..

[Koori]

So ring bus voltage > vcore = bad?

For example my chip (4670k)

-vcore 1.25 (actual 1.252)

-ring bus 1.25 (actual 1.272)

both @ 100mhz x 46

 

is it unsafe?

currently running at 1.22v@ (46 46 44 44) and ring bus @ (43 and 42) @ 1.2v

[ProKoN]

So ring bus voltage > vcore = bad?

For example my chip (4670k)

-vcore 1.25 (actual 1.252)

-ring bus 1.25 (actual 1.272)

both @ 100mhz x 46

 

is it unsafe?

currently running at 1.22v@ (46 46 44 44) and ring bus @ (43 and 42) @ 1.2v

your good to go

 

 

keep your ring bus voltage below 1.4V

 

 

it is ok if your ring voltage is more than your cpu Vcore voltage. ( just edited that line of the guide)

[Koori]

your good to go

 

 

keep your ring bus voltage below 1.4V

 

 

it is ok if your ring voltage is more than your cpu Vcore voltage. ( just edited that line of the guide)

Thanks for the reply! 

[BaZZyL]

Hello everyone.

First of all I would like to thank the Author for this guide, but...

 

... I'm having problems with oc'ing my 4670k (batch: L307B437).

 

I've achived stable 4.2GHz with the following settings:

 

CPU Ratio: x42
CPU Cache Ratio: x38
CPU Input Voltage: 1.9
CPU Override Voltage: 1.25
CPU Cache Override Voltage: 1.25

 

Ram XMP 1.6 1600Mhz 1.5v

 

When i try to push past 4.2 with 1.25V everything becomes very unstable and i get BSODs during OS boot.

So far I've tried 1.3V with 4.4Ghz - OS boots but during stress i get BSODs.

 

If its possible I would like to achive 4.4-4.5 with this CPU.

Maybe its not possible with this early batch, I don't know.

Seeing how people OC their CPUs to 4.5 with low voltage makes me wanna cry

 

Right now i'll try to lower cache ratio to 30 and ram to 1333 and see that happends...

[ProKoN]

 

Right now i'll try to lower cache ratio to 30 and ram to 1333 and see that happends...

 

yup try those two things.

 

make sure your on the latest official bios.

 

try to raise the cpu input voltage to 2.2V, see if that helps.

 

thanks for sharing your results and welcome to the LTT community.

[BaZZyL]

My bios is 2.30 with date 7/3/2013 but on asrock's page there is bios 2.30 with 7/8/2013 date.

Should I update it?

[Doug2507]

Hello!

 

Make sure you drop your cache voltage down as well and disable XMP. 1.25vring seems quite high for x38. I wouldn't expect it to need even 1.2vring.... Try dropping it down to x33 and 1.15vring. Leave it there till the core is stable.

 

What did you use to stress to declare x42 stable? 8hr run of XTU?

 

BSOD during boot means you're miles away from being anywhere near stable.

 

What BSOD code were you getting? 101 or 124? (or neither?!)

[Hikaru12]

Wouldn't Intel Burn Test be a more reliable tool since it runs hotter than this program and Prime 95?

[Doug2507]

No, it doesn't work like that. All you are doing there is seeing how hot you can get the cpu! You'll find many instances where people have passed IBT then failed another such as P95. I personally use UTX like Prokon suggested but i also use x264 as it uses AVX. I've never had a BSOD/crash after declaring stable in both. At the end of the day it all depends on what you use your computer for with regards to which stability tests you use.

[MrQ]

Hi there. I bought a new computer a little over 2 months ago and have now decided to overclock it. It is the first time I overclock due to my old computer only had standard CPU fan.

I started with overclocking 3 days ago and have read about it a lot and checked many different guides before actually doing anything. So the first thing I did was to set the CPU ratio @ 44, CPU Core Voltage @ 1.25V, VCCIN @ 1.9V, DRAM Frequency @ 1333 and DRAM Voltage @ 1.5.

Then I started testing and doing 5 minutes at start just to see if it could pull of 5 min at that voltage and then go up one when it failed and try it overnight. So I tried at 1.25V. No problems at all. So I lowered to 1.24 still no problems. So I lowered to 1.23 with no problems... Then 1.22 then 1.21 then 1.20 and so on. Until I reached 1.16 and still no problems... By then it was late and I needed to go to bed so I let it running in IETU @ 44x with 1.16V. Next morning after 10 hours of stress testing in IETU no problems detected. So I tested setting 1.25V and going up in Multiplier and no problems until I reached 49x. Then my computer crashed at OS boot. So I went back to 48x multiplier and ran it for 2 hours with no problems. Then I put my DRAM to 1600MHz which it should go for and got problems so I lowered my multiplier to 47x and got an error after about 50 minutes. So I increased Voltage to 1.27 and got it running for over 2 hours with no errors at all.

That is about as far as I have come. Now to my actual question. From what I have understood being able to get 44x @ 1.16V is really good. And 47x @ 1600MHz ram @ 1.27 too for that. Anyhow is this true? Have I gotten my hands on a CPU that is the Jesus Christ of CPU's? I expected to only get about 44 multiplier with my computer setup since I only got an Air fan.

Edit: I've tested a to go lower then 1.16V with the 44 multiplier settings I said first. 1 hour of testing in Prime95 @ 1.12v achieved now.

My setup btw

 

• Intel Core i5-4670K - 4 threads / 3,40GHz (3,8GHz Turbo) / 6MB / Socket 1150 (Boxed) (84w) (Unlocked)
• Gigabyte GeForce GTX 760 OC 2GB (GV-N760OC-2GD)
• Corsair Vengeance Low Profile 8GB DDR3 PC3-12800 1600MHz (CML8GX3M2A1600C9) (2x4GB)
• Samsung 840 Series SSD 250GB SATA3 Basic Kit (MZ-7TD250BW)
• WD Green 2TB (IntelliPower / 64MB Cache)
• MSI Z87-G45 GAMING - ATX / Intel Z87
• be quiet! PowerSupply (PSU) Pure Power L8 630W Modular
• Fractal Design Arc Midi R2 - Black (Ej PSU)
• SilverStone Argon CPU cooler SST-AR01

[BaZZyL]

Hello!

 

Make sure you drop your cache voltage down as well and disable XMP. 1.25vring seems quite high for x38. I wouldn't expect it to need even 1.2vring.... Try dropping it down to x33 and 1.15vring. Leave it there till the core is stable.

 

What did you use to stress to declare x42 stable? 8hr run of XTU?

 

BSOD during boot means you're miles away from being anywhere near stable.

 

What BSOD code were you getting? 101 or 124? (or neither?!)

 

It's 124:

 

BCCode:    124

BCP1:    00000000

BCP2:    85ABE01C

BCP3:    BF800000

BCP4:    00000124

OS Version:    6_2_9200

Service Pack:    0_0

Product:    256_1

 

[ProKoN]

Hi there. I bought a new computer a little over 2 months ago and have now decided to overclock it. It is the first time I overclock due to my old computer only had standard CPU fan.

I started with overclocking 3 days ago and have read about it a lot and checked many different guides before actually doing anything. So the first thing I did was to set the CPU ratio @ 44, CPU Core Voltage @ 1.25V, VCCIN @ 1.9V, DRAM Frequency @ 1333 and DRAM Voltage @ 1.5.

Then I started testing and doing 5 minutes at start just to see if it could pull of 5 min at that voltage and then go up one when it failed and try it overnight. So I tried at 1.25V. No problems at all. So I lowered to 1.24 still no problems. So I lowered to 1.23 with no problems... Then 1.22 then 1.21 then 1.20 and so on. Until I reached 1.16 and still no problems... By then it was late and I needed to go to bed so I let it running in IETU @ 44x with 1.16V. Next morning after 10 hours of stress testing in IETU no problems detected. So I tested setting 1.25V and going up in Multiplier and no problems until I reached 49x. Then my computer crashed at OS boot. So I went back to 48x multiplier and ran it for 2 hours with no problems. Then I put my DRAM to 1600MHz which it should go for and got problems so I lowered my multiplier to 47x and got an error after about 50 minutes. So I increased Voltage to 1.27 and got it running for over 2 hours with no errors at all.

That is about as far as I have come. Now to my actual question. From what I have understood being able to get 44x @ 1.16V is really good. And 47x @ 1600MHz ram @ 1.27 too for that. Anyhow is this true? Have I gotten my hands on a CPU that is the Jesus Christ of CPU's? I expected to only get about 44 multiplier with my computer setup since I only got an Air fan.

Edit: I've tested a to go lower then 1.16V with the 44 multiplier settings I said first. 1 hour of testing in Prime95 @ 1.12v achieved now.

My setup btw

 

• Intel Core i5-4670K - 4 threads / 3,40GHz (3,8GHz Turbo) / 6MB / Socket 1150 (Boxed) (84w) (Unlocked)

• Gigabyte GeForce GTX 760 OC 2GB (GV-N760OC-2GD)

• Corsair Vengeance Low Profile 8GB DDR3 PC3-12800 1600MHz (CML8GX3M2A1600C9) (2x4GB)

• Samsung 840 Series SSD 250GB SATA3 Basic Kit (MZ-7TD250BW)

• WD Green 2TB (IntelliPower / 64MB Cache)

• MSI Z87-G45 GAMING - ATX / Intel Z87

• be quiet! PowerSupply (PSU) Pure Power L8 630W Modular

• Fractal Design Arc Midi R2 - Black (Ej PSU)

• SilverStone Argon CPU cooler SST-AR01

Hello and welcome to the LTT community!

 

Thanks for sharing your results \ experience!

 

I think you got fairly lucky with that chip. im not sure i would call it the jesus christ of cpus but it certainly sounds like it could be one of jesus's  disciples.

 

I dont consider anything even close to stable unless you can run 8-24hrs of XTU. your 2 hr runs seem promising, however 8-24hr minimum.

 

your cooler is average. what kind of temps are you hitting at 1.27V

[MrQ]

Hello and welcome to the LTT community!

 

Thanks for sharing your results \ experience!

 

I think you got fairly lucky with that chip. im not sure i would call it the jesus christ of cpus but it certainly sounds like it could be one of jesus's  disciples.

 

I dont consider anything even close to stable unless you can run 8-24hrs of XTU. your 2 hr runs seem promising, however 8-24hr minimum.

 

your cooler is average. what kind of temps are you hitting at 1.27V

Thank you.

 

Yes I agree that 1 hour hardly can be called stable mostly just to see about how low my CPU could go. Tonight I ran XTU @ 44x multiplier with the settings you recommend in the first post. Except 1.12V, As of now it has run for 8½ hours and no signs of any problems yet. With this test max temperature seems to be 60C.

 

With 47x, 1.27V and 1600MHz DRAM I got 82C as maximum temperature in XTU, with Prime95 though I got 82C as max most of the time except during 2 of it's test when the Temperature suddenly went to 100C and it started to throttle, Only happened for about 10 minutes of a total 160 minutes of testing. What I've understood Prime95 isn't as "real world" as XTU and mostly does hardcore testing.

 

I've been thinking of trying to get a stable 1 hour test of 47x, @ 1600MHz DRAM with a higher CPU Ring bus today. I'll leave info how it's going for me as I get it.

[MrQ]

Btw in your first post when talking about CPU Ring bus stability you wrote that if unstable one should raise the VCCIN. I assume you mixed up the numbers when raising them with 0.001 between 1.9 and 1.909 and then 0.01 between 2.0 and 2.1.
"•1.9. 1.901, 1.902, 1.903, 1.904, 1.905, 1.906, 1.907, 1.908, 1.909, 2.0,  2.01,  2.02, 2.03, etc"

 

Nothing that really matters just thought I would point it out.

[BaZZyL]

Hello everyone.

First of all I would like to thank the Author for this guide, but...

 

... I'm having problems with oc'ing my 4670k (batch: L307B437).

 

I've achived stable 4.2GHz with the following settings:

 

CPU Ratio: x42

CPU Cache Ratio: x38

CPU Input Voltage: 1.9

CPU Override Voltage: 1.25

CPU Cache Override Voltage: 1.25

 

Ram XMP 1.6 1600Mhz 1.5v

 

When i try to push past 4.2 with 1.25V everything becomes very unstable and i get BSODs during OS boot.

So far I've tried 1.3V with 4.4Ghz - OS boots but during stress i get BSODs.

 

If its possible I would like to achive 4.4-4.5 with this CPU.

Maybe its not possible with this early batch, I don't know.

Seeing how people OC their CPUs to 4.5 with low voltage makes me wanna cry

 

Right now i'll try to lower cache ratio to 30 and ram to 1333 and see that happends...

 

 

So I lowered CPU cache to 30,

cpu ratio set at 44,

set ram to 1333 (XMP off),

cpu input to 2.1,

cpu voltage 1.26,

cpu cache 1.15

 

...and got bsod at boot

 

Should I disable CPU C states?

 

Ugh, just got BSOD at 4.2Ghz.

I have no idea whats going on, going back to 4.0...

[mi3zu69]

First of all i want to thank you for the hard work you put on this thread.

After i folowed all the steps and a few days of tries i was able to make my overclock rock stable on my 4770k.

Now i keep it at 4,7 ghz core with 1.3v adaptive mode in BIOS and 4,5ghz cache with 1.25v adaptive mode in bios and 1.9v VCCIN 12 hours IEXTU stability tested and 40 runs of Lynx64 with 8gb memory used.

I also managed to keep it stable at these frequencies with my16 GB Corsair Dominator Platinum's XMP profile loaded at 1866 mhz.

I'm preety sure than i can keep it stable at 4.8 ghz if i can manage the heat, the cpu even boots to windows at 4.9 ghz 1.3v in BIOS and cache on auto at 2.0 VCCIN but when i start testing it trottles instant.

The problem i m facing are the temps at 4.7 ghz atm,in IEXTU i reached highest temp of 90 on hottest core and in Lynx arround 93.After 3 hours of Battlefield 4 gaming sesion my hottest core was arround 78.Im using atm a Thermalright Archon SB-E X2 and i dont know if i should delid it or try a diferent cooler.Deliding is cheaper since i have some Noctua NT-H1 paste arround, Liquid Pro and Liquid Ultra i heard they are hardening over time loosing the heat transfer.

 

Wich metod you used for deliding? Hammer or cutter?

[DarrenP]

Just leave the TIM. You're not going to get much tangible performance from applying new stuff. But if you must razor blade is much better then hammer. The hammer method is for morons. 

[mi3zu69]

You mean delid and put noctua tim or leave the original TIM and not delid?

[DarrenP]

Yes, just leave the TIM there is no tangible performance increase. 

[mi3zu69]

Well deliding should bring the temps at least 10,15 degrees down so i can push it to 4.8 ghz core and 4.6 cache at a decent votlage.

[iamjustme]

Hey Prokon, awesome comprehensive guide. I found ocing, for a first timer, very simple despite how time consuming it is. thanks so much...quick question, should I continue to decrease the cpu ring voltage until hitting bsod on startup or run a stress test for each decrement?

 

4.5ghz @ 1.31V

4300mhz @ 1.27V (currently stress testing)

vccin 1.9V

[TechSage]

Good Guide!

[ProKoN]

Hey Prokon, awesome comprehensive guide. I found ocing, for a first timer, very simple despite how time consuming it is. thanks so much...quick question, should I continue to decrease the cpu ring voltage until hitting bsod on startup or run a stress test for each decrement?

 

4.5ghz @ 1.31V

4300mhz @ 1.27V (currently stress testing)

vccin 1.9V

Hey thanks for the compliment and welcome to the LTT community!

 

When trying to find the best ring voltage heres what i personally do

 

set ring voltage to 1.4V

 

start at a 1:1 ring ratio

 

begin reducing ring multiplier until im stable for 20 hours

 

begin reducing voltage until stable for 20 hours

 

it really does not matter how you approach finding the best voltage. you can start "high" or start "low" its up to you, so long as you get the lowest possible voltage whilst maintaining stability.