Intel Haswell 4670k + 4770K Overclocking guide

[ProKoN]

i like the swiftech products

 

H220

 

H220X

 

Cooler Master Glacier 240L

 

the products perform well, I believe in supporting (buying from) companies you believe deliver great products value support. swiftech does it all

 

 

 

for a high end air cooler

 

Noctua nhd-12

 

I believe in supporting noctua as well!     + plus they send you free mounting hardware if a new socket comes out and you upgrade

[Myrkan]

But is it really worth buying a high end cooling system when the cpu is rubbish anyways? Cpu wont boot at 4,6 at 1,25 vcore.

[ProKoN]

But is it really worth buying a high end cooling system when the cpu is rubbish anyways? Cpu wont boot at 4,6 at 1,25 vcore.

 you dont invest into a high end cpu cooler because you want a better overclock on a particular chip

 

you invest because you will get a better overclock on your current and future builds!

 

when i go haswell 8 core will i upgrade my cooling? hell no I already did!

 

not being able to boot at 4.6 1.25V is meaningless

[Myrkan]

Thanks!

There is one more thing... Is there any question you can't answer?

Great guide and great support answering questions, keep The awesome job up!

[ProKoN]

Thanks!

There is one more thing... Is there any question you can't answer?

Great guide and great support answering questions, keep The awesome job up!

 

 

if you want to thank me heres what you do

 

update your profile signature with your system specs include your location!

 

post your validated OC in meltymoons Database

 

post your highest possible cinebench score in jumper118's thread

 

links are provided in the opening post

[Myrkan]

Ill do it in a couple of hours =)

[Henry]

Most motherboards these days have automatic one click overclocking. Thats what I do.

[whoohaaah1]

Most motherboards these days have automatic one click overclocking. Thats what I do.

 

i personally would rather do the overclocking manually so you know exactly what is going on. the automatic overclocking sometimes undervolts or overvolts and can be unstable.

[sleepy_hollow]

just applying your guide, and i got sweetspot on 4.5GHz @1.25V in my i5 4670k

it's stable to put on fixed voltage than auto as you said, but it's work fine to auto when i OC it to 4.2GHz

 

and i just replace the thermal paste and got below than 70C on max

 

and there some bother me

 

my sensor reading is :

TZ00 max 28C

TZ01 max 30C

SYSTIN max 48C
TMPIN3 max 67 C

 

which one is reading for northbirdge and which one is reading motherboard temperatute, just affraid if my motherboard got really hot over 50C

[Mjolkspam]

Hello everyone. I have now started to OC my 4670K, have done other CPUs before but this is something new.

 

Thing is, i tried something i read on another guide, to start with 4.6Ghz at 1.25V - VCCIN @ 1.9V. Memory set to 1333Mhz with 1.5V. All this to see if i got a good CPU or not, if it can boot etc.

If that didin't work, lower to 4.4Ghz - 1.25V,   VCCIN - 1.9V  and Memory at 1333Mhz with 1.5V, I then got it to boot, crashed when i had done 2 passes in Cinebench.

 

Then i started with this guide instead, because it's better and just more simple to understand. So i started by increasing cpu vcore to 1.26V,  4 passes done, crashed when I started the 5th pass in Cinebench. For some reason I went straight to 

1.29V, thought I was at that. But this time, I got 6 passes done in Cinebench without problems. I started IETU stress test, put it on 8h, 5min. It passed without any crash, good so far. (Link below).

Here is where the problem started for me. In the picture i have linked below i didn't have it on 100% cpu load, forgot I needed the picture done with cpu load etc. So i started up AIDA this time, put the stress test on to get 100% cpu load.

Then it crashed about 1min after i started AIDA stress, rebooted, started Cinbench this time to just check if that worked. Nope, it did not. Crashed when i started my 3rd pass.

 

So a summery, it did 6 passes in Cinebench and it passed 8 hours of IETU stress test @ 4.4Ghz - 1.29V, VCCIN - 1.9V and Memory @ 1333Mhz with 1.5V.  But when i started AIDA stress, it crashed after 1min, crashed in Cinebench after 2nd pass.

Why would it do this?  Sounds weird to me. And i needed 1.29V before it was stable at 4.4Ghz, sure is OK, as not every CPU is awesome

 

Let me know if you need anything else I might have forgot for my question. I'm not that good at English, so i can easy forget stuff etc.

**EDIT**  Maybe i should add that my PSU (OCZ ModXStream Pro 600W is pretty much from either 08/09 (2008 is the release year for it), so it's old..

 

OC settings:   4.4Ghz - 1.29V,  VCCIN - 1.9V,  Mem @ 1333Mhz - 1.5V.   (Edit* Ring ratio at Auto, so it's at 3.8Ghz with the V set at Auto aswell.)

 

 

 

I can also add that i started this OC on my 4670K to check if it's "working" properly. I have had some issues in games like BF4 when playing in Borderless, fps locked at 60.

I get some weird stuttering. CS:GO same thing but fixed after i turned Multicore Render, Off.   So i needed to rule things out. And overclocking the CPU tests if it's up to the task or not.

 

Picture with IETU stress test for 8 hours (without load). Max thermal from stress shows in the picture.

https://hostr.co/file/3gKALG72F8dJ/Namnls.jpg

[Henry]

i personally would rather do the overclocking manually so you know exactly what is going on. the automatic overclocking sometimes undervolts or overvolts and can be unstable.

Only certain apps like prime95 make it overvolt, usually

[KTM]

auto - board defines auto values

 

normal- allows you to use voltage offset

 

manual- fixed user defined voltage

 

 

for your final setting you want to set the vcore mode to normal and use a voltage offset ( this applies for both ring voltage and cpu voltage)

 

this is how to set a gigabyte board to "adaptive voltage" to allow the voltage to drop down while your idle.

Is this the so called c7 state? If so, then I can't use this. My PSU isn't haswell compatible and cannot drop the voltages while idleing.

Ok, so I decided to manual OC the ring bus.

The best it could do:

CPU freq 4.6 ghz

vcore 1.27

vccin 1.95

cpu ring 4.1 ghz

ring volt 1.26

Ring 42 @ 1.27= bsod

Ring 41 @ 1.25=bsod

Temps at stress test with Ring 41 @ 1.27 all below 76 C

Temps today (a hot day outside + my room) with ring 41 @ 1.26 all below 81 C

I know 1.26 isn't that much, but is it too much for 4.1 ghz ring? I mean if you compare it to cpu freq 4.6 @ 1.27.

4.1 ghz ring is really the max with these voltages.

[ProKoN]

Is this the so called c7 state? If so, then I can't use this. My PSU isn't haswell compatible and cannot drop the voltages while idleing.

Ok, so I decided to manual OC the ring bus.

The best it could do:

CPU freq 4.6 ghz

vcore 1.27

vccin 1.95

cpu ring 4.1 ghz

ring volt 1.26

Ring 42 @ 1.27= bsod

Ring 41 @ 1.25=bsod

Temps at stress test with Ring 41 @ 1.27 all below 76 C

Temps today (a hot day outside + my room) with ring 41 @ 1.26 all below 81 C

I know 1.26 isn't that much, but is it too much for 4.1 ghz ring? I mean if you compare it to cpu freq 4.6 @ 1.27.

4.1 ghz ring is really the max with these voltages.

4.1 @ 1.27 for the ring is perfectly normal!

 

 

your psu is most likely more compatible with Haswell than you expect.

 

I suggest enabling c-states and using adaptive voltage when you have finalized your voltages.

[ProKoN]

Hello everyone. I have now started to OC my 4670K, have done other CPUs before but this is something new.

 

Thing is, i tried something i read on another guide, to start with 4.6Ghz at 1.25V - VCCIN @ 1.9V. Memory set to 1333Mhz with 1.5V. All this to see if i got a good CPU or not, if it can boot etc.

If that didin't work, lower to 4.4Ghz - 1.25V,   VCCIN - 1.9V  and Memory at 1333Mhz with 1.5V, I then got it to boot, crashed when i had done 2 passes in Cinebench.

 

Then i started with this guide instead, because it's better and just more simple to understand. So i started by increasing cpu vcore to 1.26V,  4 passes done, crashed when I started the 5th pass in Cinebench. For some reason I went straight to 

1.29V, thought I was at that. But this time, I got 6 passes done in Cinebench without problems. I started IETU stress test, put it on 8h, 5min. It passed without any crash, good so far. (Link below).

Here is where the problem started for me. In the picture i have linked below i didn't have it on 100% cpu load, forgot I needed the picture done with cpu load etc. So i started up AIDA this time, put the stress test on to get 100% cpu load.

Then it crashed about 1min after i started AIDA stress, rebooted, started Cinbench this time to just check if that worked. Nope, it did not. Crashed when i started my 3rd pass.

 

So a summery, it did 6 passes in Cinebench and it passed 8 hours of IETU stress test @ 4.4Ghz - 1.29V, VCCIN - 1.9V and Memory @ 1333Mhz with 1.5V.  But when i started AIDA stress, it crashed after 1min, crashed in Cinebench after 2nd pass.

Why would it do this?  Sounds weird to me. And i needed 1.29V before it was stable at 4.4Ghz, sure is OK, as not every CPU is awesome

 

Let me know if you need anything else I might have forgot for my question. I'm not that good at English, so i can easy forget stuff etc.

**EDIT**  Maybe i should add that my PSU (OCZ ModXStream Pro 600W is pretty much from either 08/09 (2008 is the release year for it), so it's old..

 

OC settings:   4.4Ghz - 1.29V,  VCCIN - 1.9V,  Mem @ 1333Mhz - 1.5V.   (Edit* Ring ratio at Auto, so it's at 3.8Ghz with the V set at Auto aswell.)

 

 

 

I can also add that i started this OC on my 4670K to check if it's "working" properly. I have had some issues in games like BF4 when playing in Borderless, fps locked at 60.

I get some weird stuttering. CS:GO same thing but fixed after i turned Multicore Render, Off.   So i needed to rule things out. And overclocking the CPU tests if it's up to the task or not.

 

Picture with IETU stress test for 8 hours (without load). Max thermal from stress shows in the picture.

https://hostr.co/file/3gKALG72F8dJ/Namnls.jpg

 

 

sounds like your very close to being stable!

 

maybe try to bump up the cpu vcore a bit. you may require 1.3-1.31 volts for 100% stability! 

 

Your psu is more compatible with Haswell than you think!

[ProKoN]

just applying your guide, and i got sweetspot on 4.5GHz @1.25V in my i5 4670k

it's stable to put on fixed voltage than auto as you said, but it's work fine to auto when i OC it to 4.2GHz

 

and i just replace the thermal paste and got below than 70C on max

 

and there some bother me

 

my sensor reading is :

TZ00 max 28C

TZ01 max 30C

SYSTIN max 48C

TMPIN3 max 67 C

 

which one is reading for northbirdge and which one is reading motherboard temperatute, just affraid if my motherboard got really hot over 50C

 

 

your mother board is not overheating these are not AMD chips. your cpu is likely drawing  under 100W at full load. most boards can handle this!

 

use msi command center \ HW monitor to monitor core voltage and temps

 

motherboard temp is labeled "system " temp

 

post your system  specs in your sig please.

[Mjolkspam]

sounds like your very close to being stable!

 

maybe try to bump up the cpu vcore a bit. you may require 1.3-1.31 volts for 100% stability! 

 

Your psu is more compatible with Haswell than you think!

Yea, i'll try 1.3V tonight.  When the stress passed those 8hours last night, i thought to myself, 'Oh, nice.. It's working now.'    But then it crashed when started another stress.

So 4.4Ghz with 1.3V is "normal"?  I know CPUs have very different values when it comes to overclock. But I have seen so many with 1.27 at 4.6Ghz.  That's a lot better than my resaults

 

What about VCCIN? Should it stay on 1.9V?  Or does that only have inpact on the ring clocking?

And when talking about the ring ratio clock etc.  When i'm trying to get a stable clock on the cpu (start of the guide), Ring Ratio is on Auto, which means it's at 3.8Ghz.  Should i lower it back to 3.4? Or leave it?

Thing about the PSU was it's old and worried it could be bad, as in old bad. If it's not up to the task anymore.

 

Side question:   Is it possible to know if Motherboard is "damaged" with an easy check?  As i mentioned, i have performance problems in BF4 with my setup (signature). Have some weird stutter

With my thinking it's either CPU, Motherboard or PSU that's bad.  GPU is fine (tested in another computer).  If there is any problem at all.

[ProKoN]

Yea, i'll try 1.3V tonight.  When the stress passed those 8hours last night, i thought to myself, 'Oh, nice.. It's working now.'    But then it crashed when started another stress.

So 4.4Ghz with 1.3V is "normal"?  I know CPUs have very different values when it comes to overclock. But I have seen so many with 1.27 at 4.6Ghz.  That's a lot better than my resaults

 

What about VCCIN? Should it stay on 1.9V?  Or does that only have inpact on the ring clocking?

And when talking about the ring ratio clock etc.  When i'm trying to get a stable clock on the cpu (start of the guide), Ring Ratio is on Auto, which means it's at 3.8Ghz.  Should i lower it back to 3.4? Or leave it?

Thing about the PSU was it's old and worried it could be bad, as in old bad. If it's not up to the task anymore.

 

Side question:   Is it possible to know if Motherboard is "damaged" with an easy check?  As i mentioned, i have performance problems in BF4 with my setup (signature). Have some weird stutter

With my thinking it's either CPU, Motherboard or PSU that's bad.  GPU is fine (tested in another computer).  If there is any problem at all.

 

Have a look at the Haswell OCDB in the original post. many people require similar or higher voltage for 4.4ghz. by no means is it a good chip but it is "normal\average"

 

its usually fine to leave the ring to auto for both frequency and voltages, unless finding cpu stability is really difficult.   BSOD X124 is typically ring bus and X101 is vcore.

 

your bf4 issue hmm, does it happen when cpu is at stock clocks?

 

fps locked at 60fps typically means vsync is enabled either in game or in the driver suite

 

please list your bf4 video settings

 

most likely driver\gforce experience related

 

have you used gforce experiance to optimize your games?

 

i would suggest re-installing gfx driver (perform clean install) and optimize settings yourself

 

are you overclocking your gfx card?

[Mjolkspam]

Have a look at the Haswell OCDB in the original post. many people require similar or higher voltage for 4.4ghz. by no means is it a good chip but it is "normal\average"

 

its usually fine to leave the ring to auto for both frequency and voltages, unless finding cpu stability is really difficult.   BSOD X124 is typically ring bus and X101 is vcore.

 

your bf4 issue hmm, does it happen when cpu is at stock clocks?

 

fps locked at 60fps typically means vsync is enabled either in game or in the driver suite

 

please list your bf4 video settings

 

most likely driver\gforce experience related

 

have you used gforce experiance to optimize your games?

 

i would suggest re-installing gfx driver (perform clean install) and optimize settings yourself

 

are you overclocking your gfx card?

About BSOD, I didn't get that when it crashed, i forgot to mention that. It just restarted,  died to black "start" screen before UEFI screen and booted like every normal day.

 

 

Battlefield 4,  I have reinstalled my OS 3 times i think, was on 8.1 when it started, installed Win7 and Win7 again. Latest drivers every time.

I never use vsync, i lock the fps with the 'user.cfg' file (console command). I get massive screentearing if i'm in Fullscreen, stutter if using Windowed-Fullscreen. I even get stutter when i

don't have the fps locked at all, running about 90-110 fps. Everything feels stuttery, weird, not smooth at all. Funny thing aswell, it's not all the time. 

If i'm at A on Zavod 311 (might not say you much), I can run around looking in circles. The stutter can be when looking into the map or even when looking at nothing at the edge of the map.

When it's not stuttering it's smooth as silk.  

 

I have tried to use it to optimize BF4, but with the standard settings it will run crap, low fps etc. But have never realy tried to change the optimize bar for better performance before i push Optimize.

Could give that a try tomorrow.

One thing that is weird..  It's not showing on my recorded videos at all. On my recorded videos at 60fps it's smooth.  So it could be the screen or DVI cable, I have tested 5 screens in total, 5 cables, 3 of them was dual-link DVI.

 

I have a profile for a OC on the GPU yes, not using it. Same thing w/e I use.  

My setting are: Everything on Low, 1080p. Does not matter if i lower the resolution etc.

 

**Sorry i'm off-topic here with the BF4 related problem  Just wanted to let you know why i started this and that I have the problem.  Allways good to have people opinion on stuff, might solve it**

[sleepy_hollow]

your mother board is not overheating these are not AMD chips. your cpu is likely drawing  under 100W at full load. most boards can handle this!

 

use msi command center \ HW monitor to monitor core voltage and temps

 

motherboard temp is labeled "system " temp

 

post your system  specs in your sig please.

 

then is labeled as "SYSTIN" in HW Monitor yeah fortunately it below 50C

 

but i still left my ring bus speed at 3800MHz with auto voltage, seeing it's reach 67C with less airflow i didn't dare to bring it up at 1:1 ratio or 100-300 MHz less than CPU clock

 

and there's some strange in my RAM, if i enabled the XMP profile and set it to 2133MHz (my RAM's is PC17000) it'll automatically BSOD after boot even i put it's voltage to 1.55V, but if i disabled it and insert 2133 speed manually (with same voltage) it works fine. it doesn't matter if my cpu in overclocking condition or default condition

so what's wrong with this, is it my ram? or my motherboard? or something else

[ProKoN]

then is labeled as "SYSTIN" in HW Monitor yeah fortunately it below 50C

 

but i still left my ring bus speed at 3800MHz with auto voltage, seeing it's reach 67C with less airflow i didn't dare to bring it up at 1:1 ratio or 100-300 MHz less than CPU clock

 

and there's some strange in my RAM, if i enabled the XMP profile and set it to 2133MHz (my RAM's is PC17000) it'll automatically BSOD after boot even i put it's voltage to 1.55V, but if i disabled it and insert 2133 speed manually (with same voltage) it works fine. it doesn't matter if my cpu in overclocking condition or default condition

so what's wrong with this, is it my ram? or my motherboard? or something else

enabling xmp does much more than adjust the main timings and frequency, xmp adjusts tertiary timings as well

 

xmp is a suggestion, not a stable profile. if your kit says xmp 2133 1.55V there is a good chance you will need 1.6-1.65 Volts to run xmp values.

 

when you input it manually you are most likely not running the same latency values xmp would load, hence your perceived stability @ 1.55V

[Mjolkspam]

Okey, i tried at 4.4Ghz with 1.3V tonight, it must have crashed just before the 8 hour mark.  I woke up after 6 hours, went to sleep again and woke up at 8 hour mark. It had restarted by then.

 

I have lots of minidumps, one for tonight. But what minidump "reader" should i use? I havn't looked into dumpfiles for a long time

[ProKoN]

Okey, i tried at 4.4Ghz with 1.3V tonight, it must have crashed just before the 8 hour mark.  I woke up after 6 hours, went to sleep again and woke up at 8 hour mark. It had restarted by then.

 

I have lots of minidumps, one for tonight. But what minidump "reader" should i use? I havn't looked into dumpfiles for a long time

 

 

been thinking about your last post post with bf4 ..... I would suggest trying a different gfx card if you can or perhaps put your gfx card in the second x16 slot.

 

I dont view windows logs sorry.  i just know if I crashed Im unstable.

 

what kind of max temps are you hitting at 1.3V

 

i just re-read your original post and thinking about where you are now.

 

i will make a suggestion

 

increase your VCCIN to 2.2V and stress test again at 1.3V vcore

 

i would leave it at 2.2V for finding ring stability as well.

 

lower it after your done both

[Mjolkspam]

been thinking about your last post post with bf4 ..... I would suggest trying a different gfx card if you can or perhaps put your gfx card in the second x16 slot.

 

I dont view windows logs sorry.  i just know if I crashed Im unstable.

 

what kind of max temps are you hitting at 1.3V

 

i just re-read your original post and thinking about where you are now.

 

i will make a suggestion

 

increase your VCCIN to 2.2V and stress test again at 1.3V vcore

 

i would leave it at 2.2V for finding ring stability as well.

 

lower it after your done both

I have tested my old HD5870, same thing. I'm going to ask my half brother if i can test he's R9 290x card. And see if it could be something like my card is so much older than the CPU, so it works bad together or w/e.

 

I'll ask my brother then, he knows a program i think. I was at 76 when i was on 1.29V and i didn't have a chance to see the temps late into the stress test. But i would think i was around the same or maybe 80.

VCCIN at 2.2V, i'll try that tonight.  Would it get more stable even if i used VCCIN at 2.2V and lower my vcore again to say 1.28?  Or does it not have that of an big inpact?

 

What i mean is try to start over but with VCCIN higher.

 

**EDIT**    I have now checked my minidumps. The one i got this morning was the following:     Bug Check Code:  0x00000124  -   Caused By Driver: ntoskrnl.exe

Looks like it is what you said, X124?   And every single one of the BSOD is 0x00000124.

[KTM]

Same BSOD as I get.

I pass all stress tests, but with bf4 I get bsod (not all the time).

Wiped all nvidia drivers and installed it again, clean. Still bsod.

I raised VCCIN with 0.05 (1.900 => 1.950). I also manual oc'd ring bus, so no auto mode. Stress test passed.

So far so good, no bsod. But it's too early to know if it was VCCIN and ring bus voltage.

[ProKoN]

I have tested my old HD5870, same thing. I'm going to ask my half brother if i can test he's R9 290x card. And see if it could be something like my card is so much older than the CPU, so it works bad together or w/e.

 

I'll ask my brother then, he knows a program i think. I was at 76 when i was on 1.29V and i didn't have a chance to see the temps late into the stress test. But i would think i was around the same or maybe 80.

VCCIN at 2.2V, i'll try that tonight.  Would it get more stable even if i used VCCIN at 2.2V and lower my vcore again to say 1.28?  Or does it not have that of an big inpact?

 

What i mean is try to start over but with VCCIN higher.

 

**EDIT**    I have now checked my minidumps. The one i got this morning was the following:     Bug Check Code:  0x00000124  -   Caused By Driver: ntoskrnl.exe

Looks like it is what you said, X124?   And every single one of the BSOD is 0x00000124.

 

x124 is typically related to ring bus instability

 

solutions

 

increase vccin

 

increase ring voltage

 

reduce ring multiplier