Intel Haswell 4670k + 4770K Overclocking guide

[ProKoN]

Bios is f7 

 

Cpu Multi 45x

Cpu Vcore 1.395

Cpu ring voltage 1.050V (Auto)

Cpu Vrin External Override 1.800V (Auto)

Ram Mult 18.66 

Ram Volt 1.700V

VCCIN 1.79000 (I see it in gigabyte tweaklauncher)

No i didn't go with dram over 1700V that is highest i went and it was stable on 1866

 

 

so your

 

ring ratio is at 4500MHz on stock volts with stock VCCIN?

 

thats a 1:1 ratio. Very good overclock.

 

Are you XTU stable for 8 or more hours at these settings?

 

im not sure why you havent tried to run your mem chips at 2400MHz @1.7V or 2133MHz at 1.65V. You have two XMP profiles to choose from.

[adicted]

I did try to run  2400 at 1.71, also 2133 at 1.65 and 1.71 but it crashed both times. Don't crash only 1866 - 1.7V and that is the Oc i will keep if pass the 12h test.

 

Cpu, 4.5 Ghz at 1.395V

 Ram      1866 at 1.7V

[adicted]

Something was wrong, i left it on doing the stress test last night, and when i woke up i saw that Pc restarted... I did cmos reset, and i will start from begining... Its seems to me that with old f3/f4 bios was more stable, im not sure yet... 

For now i will leave it on default settings, going to see few more times your videos and then start from begining in day or two...

[ProKoN]

Guys who did a bit of BCLK overclocking, I would like to ask, what's is the maximum stable BCLK you guys can achieve?

 

I have dabbled in Bclk Ocing  but never had any solid, stable results.

 

My retired 4770k would boot with the blck a 150-160. It wouldnt boot at 100-140 ish. wouldnt boot past 175.

 

My current 4670k seems to like bclk overclocking a bit more. Still testing at stock blck till i find best frequency\ volts, then maybe i will experiment with the blclk a bit more.

 

I usually had to sacrifice cpu frequency in order to increase the Bclk and you dont want that.

 

It is going to really depend on the individual characteristics of your CPU with regards to Bclk overclocking on the Haswell platform. some do..and some dont. honestly its adds another variable when trying to achieve stability, hence why most people stay away from it. But it is possible, granted your cpu can do it.

 

 

 

 

[ProKoN]

Unlucky dude, I've got 4.8 stable with 1.38v but have to go all the way to 1.47 to get 4.9 to even think about booting. 

 are you running that chip without an IHS? big ups if that 4770k is naked!

 

I would like to include a Haswell OC database in the second post. I would enjoy it if you submitted your overclock results, not many people get a 4.8GHz 4770K..

 

Any submissions would require a minimum of 8 hours of stress testing, and a screenshot of it stressing for 8 or more hours.

 

Truth is i need to update the guide. None of my tests have really stressed AVX \FMA instruction sets that hard. Reality is if you want to be "AVX" stable on Haswell your really only going to hit 4.3-4.5GHz. But really not to many applications really use AVX in a punishing way. Encoding x264 videos using Handbrake is a scenario where AVX is heavily utilized. I personally dont encode x264. So for now I am not pushing for "AVX" stable.

 

 

 

set your vcore for static mode and try to run this AVX cruncher

 

http://software.intel.com/en-us/articles/intel-math-kernel-library-linpack-download

[ProKoN]

Yeah I've got the block straight on die so I can keep temps in order

 

I'll give it a shot and let you know how it goes

 

 right on watch those temps.

 

i can get 20 hours stable running xtu

 

100% game stable

 

100% stable rendering 1080p with sony vegas

 

 run AVX enabled linpack and i bsod within a minute. requires way more voltage than you think you need and the temps are overwhelming as a result.

 

TTL briefly touches on this during the last live stream.

[ProKoN]

well i wouldnt waste any time trying to get AVX stable. your not going to be able to cope with the heat, even if that sexy die is naked

 

your going to need like 1.45V to get it stable and trust me at 1.45V with the AVX instructions stressing hard, thats alot of heat.

[Doug2507]

Hi ProKon,

 

I'm new to this forum so first off hello and thank you for writing such an excellent guide!

 

I've just built myself a new rig having been out of the PC game since the QX9650 was on the go, back in the day when 280 GTX was king. How things have changed!

 

Anyway, i'm now running a 4770K along with a fellow MSI Z87 although just the MPower for myself. Coupled to that is a pair of Twin Frozr 780 GTX's (simply awesome) and 2x8GB of G.Skill TridentX.

 

Cooling is all done on air thanks to the AP181's running in the FT02 and a Phanteks cooler with the fans replaced with Thermalight TY147's.

 

Initial overclocking was done following the ASUS guide where 4.6ghz proved stable at 1.2v and i was able to achieve 4.8ghz quite happily at 2.5v with a little boost to VCCIN.

 

I initially stable tested with Prime 95 which was all fine and dandy with 4.6@1.2v. I realised later on that i was running an older version of P95 (25.x) so updated to 27.9.

 

Started running stability tests again and although still stable at 4.6@1.2v, max temps ended up in the high 80's as 27.9 uses AVX as you're probably more than aware.

 

With a vague idea of what the chip is capable of i de-lidded earlier on today and everything is back up and running. I ran P95 small FFT for over an hour and max temps hit 71deg, 3 out of the 4 cores wouldn't even hit 70deg. After reading various o/c guides (with varying opinions) i decided to ignore a lot of them due to the constant instruction to use the likes of IBT. I only use the rig for gaming, cd ripping, browsing etc so i figured whats the point of running IBT etc!

 

So, stumbled upon this guide and here i am now 1hr 30mins into a 24 stability test using IETU. So far so good and average temp is 53deg, highest having been 58deg so far. CPU TDP averaging at 77w. Nice! (all on air!)

 

This leads me to a couple of questions....

 

Even though i've established current settings stable through P95 for a couple of hours i'm going to run IETU for 8hrs+ anyway. After that i'll stay at 46ghz but back the vcore down to 1.18, run it overnight and see where i am in the morning. I'll keep testing till i find the lowest vcore for 46ghz.

 

After that i aim to start increasing the multiplier/vcore to try and take it back up to 4.7ghz then 4.8ghz, temps dependant.

 

Should i go through the process of finding max ring and ram for each of these speeds or should i find my max cpu clock 1st then go back to step 2?

 

All settings in BIOS are as you've specified in step 1.

 

 

[ProKoN]

Hi ProKon,

 

I'm new to this forum so first off hello and thank you for writing such an excellent guide!

 

I've just built myself a new rig having been out of the PC game since the QX9650 was on the go, back in the day when 280 GTX was king. How things have changed!

 

Anyway, i'm now running a 4770K along with a fellow MSI Z87 although just the MPower for myself. Coupled to that is a pair of Twin Frozr 780 GTX's (simply awesome) and 2x8GB of G.Skill TridentX.

 

Cooling is all done on air thanks to the AP181's running in the FT02 and a Phanteks cooler with the fans replaced with Thermalight TY147's.

 

Initial overclocking was done following the ASUS guide where 4.6ghz proved stable at 1.2v and i was able to achieve 4.8ghz quite happily at 2.5v with a little boost to VCCIN.

 

I initially stable tested with Prime 95 which was all fine and dandy with 4.6@1.2v. I realised later on that i was running an older version of P95 (25.x) so updated to 27.9.

 

Started running stability tests again and although still stable at 4.6@1.2v, max temps ended up in the high 80's as 27.9 uses AVX as you're probably more than aware.

 

With a vague idea of what the chip is capable of i de-lidded earlier on today and everything is back up and running. I ran P95 small FFT for over an hour and max temps hit 71deg, 3 out of the 4 cores wouldn't even hit 70deg. After reading various o/c guides (with varying opinions) i decided to ignore a lot of them due to the constant instruction to use the likes of IBT. I only use the rig for gaming, cd ripping, browsing etc so i figured whats the point of running IBT etc!

 

So, stumbled upon this guide and here i am now 1hr 30mins into a 24 stability test using IETU. So far so good and average temp is 53deg, highest having been 58deg so far. CPU TDP averaging at 77w. Nice! (all on air!)

 

This leads me to a couple of questions....

 

Even though i've established current settings stable through P95 for a couple of hours i'm going to run IETU for 8hrs+ anyway. After that i'll stay at 46ghz but back the vcore down to 1.18, run it overnight and see where i am in the morning. I'll keep testing till i find the lowest vcore for 46ghz.

 

After that i aim to start increasing the multiplier/vcore to try and take it back up to 4.7ghz then 4.8ghz, temps dependant.

 

Should i go through the process of finding max ring and ram for each of these speeds or should i find my max cpu clock 1st then go back to step 2?

 

All settings in BIOS are as you've specified in step 1.

 

 

do things one at a time. from my experience if you try and increase everything all at once it really is difficult to rule out and isolate an individual variable, so best to to things step-by step.

 

thanks for sharing your results!

 

seems like you have a good chip. 4.8GHz at 1.2V.... me and everyone else can only wish of getting this i7

 

there is "stable"

 

and then there's "AVX" stable

 

I wish I could elaborate more but im just off to work.

 

I plan on upgrading the guide to talk about "AVX" stability. and why you may or may not need to be AVX stable.

[Doug2507]

Indeed, slight difference with AVX stable! 4.6@1.2v, 4.8ghz required 1.25v and don't think that was stable. 4.7ghz@1.25v certainly was.

 

I understand doing one thing at a time, if not then you just end up chasing your tail constantly trying to figure out what's going on!

 

I'm wondering if i should get the full system done at 46ghz including ring/ram or start moving up the cpu multi to find max stable o/c on the cpu 1st, then turn attention to ring/ram. Thoughts?

[ProKoN]

I would go for a safe oc and a max oc In your case 4.6 seems pretty safe. Its nice to have a safe oc thats less aggressive. I always push for maximum clocks! Its up to you. Start by pushing your cpu to the max then you have a fairly good understanding of what max frequencies and voltages you can push. anything else in terms of lower clocks is gravy

[Doug2507]

Cool cool. Any idea what sort of time frame you've been getting for finding out if a clock is unstable? I'll run 8hr passes initially before 24hr but wouldn't mind knowing when most failures occur. At least that way i can go a little bigger in the drop/increase on vcore then have a fair idea of how close it is to being stable.

 

That's 4hrs in on 1.2v and you'd think the machine was having a sleep.

[ProKoN]

I would say the points most likely for me to crash during the xtu stress test is the 1.5 hr mark or 9 hour mark. Wish I could always be there to be certain.

I usually boot the computer and run 4-5 runs of cinebench. If its stable I may game or surf the web. Start xtu before bed and check on it when I wake up. if it crashed I make adjustments and run xtu again before I go to work. If it hasnt crashed I let er go till I get home from work. If I get home from work and its still running I consider those settings stable; thats typically a 18-20 hr run

[ProKoN]

Cool cool. Any idea what sort of time frame you've been getting for finding out if a clock is unstable? I'll run 8hr passes initially before 24hr but wouldn't mind knowing when most failures occur. At least that way i can go a little bigger in the drop/increase on vcore then have a fair idea of how close it is to being stable.

 

That's 4hrs in on 1.2v and you'd think the machine was having a sleep.

 

 

The MSI IETU is a fair bit dated now. It would be a good Idea to run the latest version 4.2.0.8 of IETU for a couple reasons.

 

a- displays cache frequency

 

b- you can run the benchmark, not sure why the msi version does not have the benchmark option. ( make sure to set you vcore mode to static \ override if using the ietu benchmark)

 

https://downloadcenter.intel.com/SearchResult.aspx?keyword=%22%22extreme+tuning+utility%22%22

 

 

I just had a decent look at your screenshot. Thank you for a perfectly detailed SS all the info is right there, thats what I like to see. Good job on the memory settings 1333 @ 1.5V, this is imperative for quickest results.

 

is your memory 2400MHz kit? have you tried to run it at 2400MHz?

 

if your last platform was C2Q you got a bit to learn, and with the ring bus Haswell certainly adds a level of complexity. the other thing is xmp profiles. I have seen so many XMP values straight up fail. including 4 kits that I have owned. I have always got my memory to run at the rated speeds / timings  but either had to input the values manually or increase voltage higher, than what xmp suggested.

 

your prolly wondering where your FSB is lol. With BLCK replacing the FSB (same thing really) the later generation core series processors dont usually take well to BCLK adjustments and with all k series  having unlocked multipliers there's really no need to touch the BCLK  unless your just a MHZ Junkie like me and wish to see how far you can push things. My 4770k did not like BCLK adjustments, this lil 4670K doesnt seem to mind up to 150.00MHz.

 

Welcome to the community and thank you for joining this discussion.

[Doug2507]

Wow, i swear i just posted a reply here 5 minutes ago!

 

Anyway, short version, thanks for the heads up, now running latest.

 

1.19v failed somewhere between 4hrs & 12hrs. No record of BSOD though so not sure what it'll be. Event viewer shows XTU restarted at 03:20hrs, started running at 18:34hrs. Makes it an 8-9hr pass before crash i reckon. With no 124 would that mean vcore is good and it needs slight tweaking elsewhere?

 

Mem - G.Skill TridentX 2400 2x8GB. Ran it with XMP no problems initially but now i've taken it off due to various guides suggesting this. It even booted no problem with XMP@4.8Ghz/1.25vcore.

 

Yes, a LOT of learning to do! Where the hell am i gonna put my NB/SB waterblocks when the time comes?

 

Couple of questions regarding readings in HWMon.....

 

 

VCC?

VIN1?

VIN2?

IA?

 

*EDIT - Should i have VCCIN at 2.0v or 1.9v? Video and 2nd post contradict each other!

 

Just for a curiosity...

 

 

 

[ProKoN]

Wow, i swear i just posted a reply here 5 minutes ago!

 

Anyway, short version, thanks for the heads up, now running latest.

 

1.19v failed somewhere between 4hrs & 12hrs. No record of BSOD though so not sure what it'll be. Event viewer shows XTU restarted at 03:20hrs, started running at 18:34hrs. Makes it an 8-9hr pass before crash i reckon. With no 124 would that mean vcore is good and it needs slight tweaking elsewhere?

 

Mem - G.Skill TridentX 2400 2x8GB. Ran it with XMP no problems initially but now i've taken it off due to various guides suggesting this. It even booted no problem with XMP@4.8Ghz/1.25vcore.

 

Yes, a LOT of learning to do! Where the hell am i gonna put my NB/SB waterblocks when the time comes?

 

Couple of questions regarding readings in HWMon.....

 

 

VCC?

VIN1?

VIN2?

IA?

 

*EDIT - Should i have VCCIN at 2.0v or 1.9v? Video and 2nd post contradict each other!

 

Just for a curiosity...

 

 

 

 

 

The fact that you can even boot into windows with such low voltage @ 49x and 50x amazes me!

 

Dude that is one sexy chip!

 

You are correct about me contraindicating myself with PIV values.  I have seen varied results from myself and other users in regards to PIV.

 

example

 

on my 4770K I could reach my max stable OC of 4.7GHz with PIV set for auto (which was usually 1.78V). It wasnt until I tried to find ring bus stability that I had to increase it to 1.9V

 

on the 4670K i needed to increase the PIV to 2.2V just to find some stability  at 4.8GHz, regardless of the ring ratio.

 

at the end of the day you want to run the PIV at the lowest voltage you can.

 

I dont run Hwinfo so I cant help with your question about thoses readings you mentioned.

 

I uses the XTU \ multimeter to read \ monitor all my voltages.

[kaefers]

Hey, thanks for the guide - I am new to Intel and struggling a bit

 

I had 4x 4770K in my hands this week, L312B333 (stock VIN/Cache 1.12v / 1.15v), L314B520 (stock VIN/Cache 1.024v / 1.09v) , 2x L315B628 (stock VIN/Cache 1.056v / 1.125v) - the L314B520 was the best of the lot as far as I could tell - unfortunately no golden egg.

 

I started out with the L314B520, F5 defaults. I set my ram to 2400, 10-12-12-28, 1.65v (I actually do not care if my CPU would be going 0.1GHz faster if I had to turn down my RAM, maybe overall would be faster, don't really mind, kinda my subjective personal constraint).

So far I got all cores at 46 @1.33v, Ring 45/45 @ 1.33v, PIV 1.9v, CPU LLC 8/8 (max), Current capability 130%, CPU power phase control Optimized, Active Frequency Mode on, EIST/C-states on, Spread spectrum off, DRAM current capability 110%.

I hit 90s with IntelBurnTest, load voltage is 1.36v. Changing CPU LLC does not seem too much difference?!?

I tried 48x2/47x2 cores @ 1.375v, 45 ring @ 1.35v, 2400RAM, which boots fine, but locks up running IBT.

 

There is still so many setting left in UEFI. Apart from what you describe in your guide, what'd be the next best thing to look at?

Should I stop whining about 2400MHz RAM and go to 2133 or 1866 and go for the higher clock?

 

Happy to hear your thoughts

[ProKoN]

first thought

 

" wish i had 4 4770Ks to clock the snot outta"

 

Welcome to the forum and thanks for sharing your result!

 

All of your settings look fairly good . I dont know too much about asus bios but looks like you have things set up fairly good.

 

I think IBT is overvolting your cpu. Does it jump to 1.36V if you set your cpu to manual mode?

 

For most synthetic stress tests you need to set your cpu to manual mode to maintain the voltage you have selected .

 

XTU does NOT overvolt the cpu when stress testing.

 

http://www.intel.com/content/www/us/en/motherboards/desktop-motherboards/desktop-boards-software-extreme-tuning-utility.html

 

run the xtu stress test and see if the overvolting to 1.36V continues

 

Given your resuts I would settle for 4.6GHz with the 4500MHz ring bus and mem chips at 2400MHz. Seems to be your sweet spot. You will most likely need a significant voltage increase to get 4.7GHz stable on all cores (1.4-1.5V).

 

If you do decide to push the cpu further just find the highest frequency you can run your ram at with tighter timings.

example:

 

1866MHz 9-9-9-24 1T

 

You will always benefit more from a higher cpu clock than high mem speeds,but honestly I think you have a very well balanced OC at these settings

 

PIV 1.9

4.6GHz CPU @ 1.33V

4.5GHz ring bus @ 1.33V

2400MHz mem chips cl 10

 

 

 

I would stay away from the per\core overclocking. from my testing I score lower on every benchmark I run if I start messing with individual core frequencies. Might just be a flaw of my board. run some cinebench at 4.8\4.7 and see if it scores higher\lower than all cores at 4.7

[Doug2507]

So yesterday/last night i went from this....

 

 

...to this...

 

 

Also tried a 49 run at 1.26v but failed after 2-3 hrs. Reckon it'll need 1.27+ so 4.8Ghz will be my 24/7 clock. Stable under 1.25v now for 8hrs in XTU which gives me a little head room to increase vcore slightly if i need to. I'm gping to call it a day there and start on the ring as i'm away for two weeks on Monday with work. Once i get the ring tuned in i'll try dialling everything in for a 24hr pass on the lowest v settings i can.

 

Not 100% sure on setting the vrin in line with the vcore so i'm going to try 1.15v and just see where it's at. I'll have a think about if i'm willing to crank up the ring above 1.2v. If i can get it to x44/45 i'll be happy enough to move onto ram so i can get some gaming done before coming back to fine tune it.

[Stompy]

Hello .. First of all I wanted to thank you for the excellent guide.
 I need some help ..
I finished part 1

I am currently looking for
Part 2: CPU Bus Ring Stability
My CPU did not like 1-1 (or as close as possible)
I'm currently trying out the X40 Cpu Ring Bus now...  for 8hrs.

 

My question is ? How Much can go down and how It effective?

Normaly is X39.

 

I have to say i run

Vcore 1.2V stability at cpu multiplier X44

8hrs stable with part 1 guide.

max temp was 80 thet cool!

 

More questions ..
VCCIN was at 2.0V TEST
You said once you start in the 1-1 calibration .. The VCCIN put on 1.9V
So that's what I did.
It did not work stable at 1-1
VCCIN 1.9V
VCCIN 2.0

Is it worth checking at 1-1 if the VCCIN can affect stability if I get close to V2.1 VCCIN?

 

Hope I was clear and sorry for the English Google Translator

Thanks again clearest guide I found MSI's Z87 motherboards

 

i got:

msi gd67 z87 gaming .

i7 4770k

2X4 8GB Gskill ripjaws 1600.
 

[Doug2507]

ProKon, did you find much difference in vcore/vring with the multimeter? Difference of about +.02v for both on mine. I'd need to watch the vid's again as i can't remember fully what you had said about the voltage difference. VCCIN reads as it should. Just wondering whether to base clocking of the board readings or software...

[ProKoN]

ProKon, did you find much difference in vcore/vring with the multimeter? Difference of about +.02v for both on mine. I'd need to watch the vid's again as i can't remember fully what you had said about the voltage difference. VCCIN reads as it should. Just wondering whether to base clocking of the board readings or software...

 I found the same result with a multimeter. but +\- .002V

 

thats 2 thousandth of a volt difference, nothing to be concerned with.

 

Vrin voltage stays at a consistent voltage when under load.

 

Vcore voltage may drop or gain .002V under load. ( on my board anyways)

 

VCCIN stay s a Consistent voltage.

 

I have found that software is accurate enough for me, so thats what I go by usually. plus its more convenient.

[Doug2507]

i.e, XTU reports 1.245vcore, m/m shows actual of 1.263/1.265 (+\-.002v). The difference in .02v right now is nothing to worry about but i'm curious to see if this scales with regards to an increase in vcore. Something to keep an eye one when chasing the upper limits i think.

[ProKoN]

Hello .. First of all I wanted to thank you for the excellent guide.

 I need some help ..

I finished part 1

I am currently looking for

Part 2: CPU Bus Ring Stability

My CPU did not like 1-1 (or as close as possible)

I'm currently trying out the X40 Cpu Ring Bus now...  for 8hrs.

 

My question is ? How Much can go down and how It effective?

Normaly is X39.

 

I have to say i run

Vcore 1.2V stability at cpu multiplier X44

8hrs stable with part 1 guide.

max temp was 80 thet cool!

 

More questions ..

VCCIN was at 2.0V TEST

You said once you start in the 1-1 calibration .. The VCCIN put on 1.9V

So that's what I did.

It did not work stable at 1-1

VCCIN 1.9V

VCCIN 2.0

Is it worth checking at 1-1 if the VCCIN can affect stability if I get close to V2.1 VCCIN?

 

Hope I was clear and sorry for the English Google Translator

Thanks again clearest guide I found MSI's Z87 motherboards

 

i got:

msi gd67 z87 gaming .

i7 4770k

2X4 8GB Gskill ripjaws 1600.

 

 

I would like to thank you ( and everyone else) for the feedback.

 

You can try this to try and find best ring bus results. It is the approach I decided to try with the 4670k. Basically start by maxing out vrin and piv voltages, find the best ring ratio, then begin to reduce voltages.

 

set

 

VCCIN -2.2V

 

ring voltage - 1.4V

 

start at a 4400MHz ring bus and start working your way down 43x 44x 42x 41x etc... until you get stable for 8 hours.

 

after you find stability start working on reducing your PIV and ring voltage as low as possible.

 

 

You dont want to reduce the ring bus too low. sure you can go lower than 39x but your performance\ watt ratio will suffer as a result. so its best to try and get it as close as possible to a 1:1 ratio. 

 

usually your ring frequency will be 100-300MHz lower than your core frequency but I can only verify that this is the case when I push the CPU past 4.5-4.6GHz. at 4.4GHz I was able to run at a 1:1 ratio. but that was what my 4770k achieved. your results will vary.. you just need to test..test..test

[Stompy]

Thanks again for the quick response I will try your recommendation.
And thanks for the information.