Intel Haswell 4670k + 4770K Overclocking guide

[Doktor Haze]

I have to thanks you for that tutorial.

 

The only thing is that XTU 8h or 24h tests were passed and then my computer crashed when i played games, just had to add 0.1 vid more.

 

So, reading another tutorial and using x264 to stress, i found the right stress test for me.

Also, following a user advice, i run prime95 27.9 with fft 1344 - 1344 using 90% of avalaible RAM, setting time to 60min, doing those tests first helped me a lot to find closed to stable settings.

 

 

So, i would recommend your tutorial procedure, but i would recommand prime95 27.9 (above setting) 1 hours tests to find core ratio and voltage stable settings, then validate with x264 (+ XTU, more tests can't be useless), then do the same for ring. I can't say anything about ram, anyway it is not covered here.

Considering the time tests take, fast eliminate unstable setting with p95 tests is extremly usefull imo, please consider it.

 

/!\ test with prime95 requires to set voltages to static, dunno for x264 as i was in static already. (You can set back to adaptive later though it doesn't seems to change anything for saving power options).

 

My final OC with 4670k, msi g45 gaming and CM Hyper 212 EVO are :

VCCCIN : 1.9

Core : x44

VID : 1.24 (vcore 1.256)

ring : x42

VRING : 1.23

 

Gonna receive new case and new cooler soon, i already know i'm stable at x45 and i also gonna try x46.

 

Thanks again.

[ProKoN]

I have to thanks you for that tutorial.

 

The only thing is that XTU 8h or 24h tests were passed and then my computer crashed when i played games, just had to add 0.1 vid more.

 

 

 

So, i would recommend your tutorial procedure, but i would recommand prime95 27.9 (above setting) 1 hours tests to find core ratio and voltage stable settings, then validate with x264 (+ XTU, more tests can't be useless), then do the same for ring. I can't say anything about ram, anyway it is not covered here.

Considering the time tests take, fast eliminate unstable setting with p95 tests is extremly usefull imo, please consider it.

 

/!\ test with prime95 requires to set voltages to static, dunno for x264 as i was in static already. (You can set back to adaptive later though it doesn't seems to change anything for saving power options).

 

 

Thanks for your valuable input.

 

I will most likely do a stress testing revision \ update soon.

 

XTU has worked well  for me but I know its not the toughest test out there and may not work for everyone.

 

Main reason I like it is because you can stress using adaptive voltage, without overvolting.  xtu and the rogbench are the only stress test I have found that works on adaptive voltage.

 

I am fine with setting a manual voltage, for me that's easy and fine, but I figured it could be an issue with some users. One less step makes things more universal..

 

Thank you for participating and sharing your results!

 

1st pass of x264 will spike voltage on adaptive 2nd pass does not.

[Doktor Haze]

Thanks for information, x264 V2 only sends 2nd type passes though.

[Harlin]

x124 is typically caused by unstable ring frequency

 

your ring voltage is fairly low even for x36.

 

i would bump up ring voltage to 1.30-1.40V.

 

and try again

 

keep your ram where its at for now.

 

Righto.

So to update again

 

Currently 'semi-stable' (sounds a little perverted but hey) at:

 

4.4MHz - Multiplier x 44

Uncore x 36

Vcore: 1.35

VCCIN: 2.0

VRING: 1.2

LLC: Extreme

K OC Enabled (Can't remember if I had that on Auto before but I enabled it anyway)

C3 Enabled - All other C States Disabled (other than Thermal)

Intel Turbo: Enabled

 

I say semi stable, as I've run both XTU and AIDA64 for 20m each just to see if I was close. I have tomorrow off so will run some longer tests.

The only cause for concern is HWMonitor is showing VCore max as 1.368v, where as VID  is what I set it to, 1.35v.

 

Temps were okay: AIDA64 84c peak, XTU 83c peak.

[ProKoN]

Righto.

So to update again

 

Currently 'semi-stable' (sounds a little perverted but hey) at:

 

4.4MHz - Multiplier x 44

Uncore x 36

Vcore: 1.35

VCCIN: 2.0

VRING: 1.2

LLC: Extreme

K OC Enabled (Can't remember if I had that on Auto before but I enabled it anyway)

C3 Enabled - All other C States Disabled (other than Thermal)

Intel Turbo: Enabled

 

I say semi stable, as I've run both XTU and AIDA64 for 20m each just to see if I was close. I have tomorrow off so will run some longer tests.

The only cause for concern is HWMonitor is showing VCore max as 1.368v, where as VID  is what I set it to, 1.35v.

 

Temps were okay: AIDA64 84c peak, XTU 83c peak.

 

 

I just checked HWmonitor

 

I can confirm the CPU Vcore value is accurate. checked with my trusty multimeter.

 

same for me with MSI board. I set Vcore in the bios for 1.38V. It hits 1.408 under load. hwmonitor shows VID at 1.38.

 

Hwmonitor CPU Vcore value is accurate.

[Harlin]

I just checked HWmonitor

 

I can confirm the CPU Vcore value is accurate. checked with my trusty multimeter.

 

same for me with MSI board. I set Vcore in the bios for 1.38V. It hits 1.408 under load. hwmonitor shows VID at 1.38.

 

Hwmonitor CPU Vcore value is accurate.

 

Well I'm guessing that my Vcore is within 'safe' range based on my readings from here

I'm happy with that at the moment, just need more testing to stabilise, and possibly get the Core down a little if stable. I did seem to have more stability once I enabled only C3, and disabled the others. I did drop the Dram voltage down to rated 1.35 (Samsung 12800 Green) from 1.5, but I doubt that would have an effect on stability (was/is running at 1333)

[aurello]

1- Get a better cooler. Go to a AIO 240mm+ water cooler or custom loop . this costs money but it is well worth it. Haswell is a beast to cool.

 

2- Delid. I dont recommend this because of the inherent risks associated. but it is a cheap solution to bring temps down. I do it myself

 

NH-D14 is a great cooler ( my favorite air cooler) but it can Not cope to well with Haswell heat output past 1.35V

 

Hei Prokon, i just redo my rig.

 

1. I delid the CPU, still using NH-D14.  But looks like air cooler really no good for this cpu. i still reach 100c, so weather bad TIM because to much/less TIM.  (i use NT-H1 for TIM) or i get really bad cpu.

 

2. not happy with number 1 result, i go for EK-Kit L360, this one really worth the price.  my minimum temps around 35c and max up to 80c with same setting.

 

now i redo my OC, i can get stable 4.4Ghz@1.28v uncore 30@1.28v VCCIN 1.9v with max temps below 70.

i can pass CB many times, and pass XTU more than 6 hours (i stop it before because i've to use the PC) BUT i can't finish XTU Benchmark, i get blue screen when 3rd quarter of benchmark running.

 

I've to increase the vcore (also uncore) to get benchmark finish, and it gets to 1.32v and making highest cpu temp 85c on benchmark result.

 

anyone having the same issue? where you can pass the XTU stress test but not the benchmark.  And does it matter?

i may not care but sometimes i want to see my benchmark.  While i could get cpu stable at lower voltage, the issue cannot pass the benchmark somehow annoying.

 

am i doing in wrong direction? any suggestion?

[ProKoN]

Hei Prokon, i just redo my rig.

 

1. I delid the CPU, still using NH-D14.  But looks like air cooler really no good for this cpu. i still reach 100c, so weather bad TIM because to much/less TIM.  (i use NT-H1 for TIM) or i get really bad cpu.

 

2. not happy with number 1 result, i go for EK-Kit L360, this one really worth the price.  my minimum temps around 35c and max up to 80c with same setting.

 

 

that made me lol. the L360 looks like a pretty nice kit. Its a bit overkill but your going to have some great results!

 

i can pass CB many times, and pass XTU more than 6 hours (i stop it before because i've to use the PC) BUT i can't finish XTU Benchmark, i get blue screen when 3rd quarter of benchmark running.

 

I've to increase the vcore (also uncore) to get benchmark finish, and it gets to 1.32v and making highest cpu temp 85c on benchmark result.

 

anyone having the same issue? where you can pass the XTU stress test but not the benchmark.  And does it matter?

i may not care but sometimes i want to see my benchmark.  While i could get cpu stable at lower voltage, the issue cannot pass the benchmark somehow annoying.

 

am i doing in wrong direction? any suggestion?

 

 

the xtu benchmark will overvolt your chip if your vcore mode is set to adaptive. make sure its set for static if benching xtu.

 

I have never experienced this issue. its not likely software related if you can complete the xtu benchmark. at 1.32V

 

I will be updating the guide very soon with some tougher stress tests that people have had good luck with.

 

does it matter hmmm well yes. If you cant pass a benchmark I think your not stable. my opinon.

[aurello]

that made me lol. the L360 looks like a pretty nice kit. Its a bit overkill but your going to have some great results!

 

 

 

the xtu benchmark will overvolt your chip if your vcore mode is set to adaptive. make sure its set for static if benching xtu.

 

I have never experienced this issue. its not likely software related if you can complete the xtu benchmark. at 1.32V

 

I will be updating the guide very soon with some tougher stress tests that people have had good luck with.

 

does it matter hmmm well yes. If you cant pass a benchmark I think your not stable. my opinon.

 

Ok, so i've stable result with core 4.4@1.32 uncore 4.3@1.32 VCCIN 1.95 (which can pass stress test and benchmark)

 

now how do i pass to higher OC?  i've try 4.5 with core and uncore up to 1.4 but still got bsod 124.  should i keep raising my voltage?

my max temps is 88c (i'm not too worried with temps now)

 

any suggestion?

[ProKoN]

Ok, so i've stable result with core 4.4@1.32 uncore 4.3@1.32 VCCIN 1.95 (which can pass stress test and benchmark)

 

now how do i pass to higher OC?  i've try 4.5 with core and uncore up to 1.4 but still got bsod 124.  should i keep raising my voltage?

my max temps is 88c (i'm not too worried with temps now)

 

any suggestion?

 max temp 88c was at 1.32V or 1.4V?

 

leave uncore  at default settings. when looking for core stability.

[aurello]

 max temp 88c was at 1.32V or 1.4V?

 

leave uncore  at default settings. when looking for core stability.

max temp 88c at 1.4.

 

i keep increasing from 1.32 up to 1.41. but still bsod.

 

so, you're saying default setting, is it auto? or my stable setting?

[ProKoN]

max temp 88c at 1.4.

 

i keep increasing from 1.32 up to 1.41. but still bsod.

 

so, you're saying default setting, is it auto? or my stable setting?

 

are you on the latest bios? something seems off.

 

did you really need 1.28 volts to get uncore stable at x30?

 

please update your sig with your hardware so i can reference it easily.

[aurello]

are you on the latest bios? something seems off.

 

did you really need 1.28 volts to get uncore stable at x30?

 

please update your sig with your hardware so i can reference it easily.

 

I'm on latest bios, 1402

 

So, let me state it again before i go further on my OC.

 

i can get stable 4.4Ghz@1.28v; uncore 30@1.28v ; VCCIN 1.9v with max temps below 70.  Perhaps the uncore voltage shouldn't that high because i didn't lower it when i drop my ringbus from 42.

i could pass 8 hours stress test, but somehow i get blue screen when i do XTU benchmark.

 

Next thing i did is increasing the vcore until 1.32 until i can pass the benchmark.  So my initial test would be run the Cinebench, XTU Benchmark, XTU Stress Test (it should pass in that order before changing any setting)

 

With new setting, 4.4G@1.32; uncore 30@1.32; VCCIN 1.9, i can pass until 8 hours stress test.

This picture to confirm:

 

 

Now my latest setting are 4.4G@1.32; uncore 42@1.32, VCCIN 1.95, still running 8 hours stress test (now 4 hours and running)

 

My goal is 4.7Ghz. though i'm happy with 4.5.  i just stretch my goal.

 

let me know your opinion.

[ProKoN]

Now my latest setting are 4.4G@1.32; uncore 42@1.32, VCCIN 1.95, still running 8 hours stress test (now 4 hours and running)

 

My goal is 4.7Ghz. though i'm happy with 4.5.  i just stretch my goal.

 

let me know your opinion.

 

Thank you for updating your sig and posting the screenshot.

 

I think your doing very good at these settings. do you have the L-360 yet or still NH-D14?

 

Thanks for the stress testing recommendations. I like your suggestion with the xtu benchmark

 

update will most  likely include a different stress testing program rather than xtu and HWmonitor to monitor vitals

[Akolyte]

I'm sorry I'm an overclocking noob, but I recently built a pc which you can check the full specs in my profile.  But, I'm wondering whether you think my components could do the mild overclock because I know my motherboard's was not designed for overclocking:

 

mobo: Asus gryphon z87 w/ armor kit

 

cpu: i5 4670k

 

cooler: hyper 212 evo. 

 

I've been pretty blessed with the parts I think, because I've got like a golden SSD with crazy speeds over the maximum rated speeds according to Samsung magician, and so I'm tempted to overclock my cpu and see if I'm decently lucky there too.  

[aurello]

Thank you for updating your sig and posting the screenshot.

 

I think your doing very good at these settings. do you have the L-360 yet or still NH-D14?

 

Thanks for the stress testing recommendations. I like your suggestion with the xtu benchmark

 

update will most  likely include a different stress testing program rather than xtu and HWmonitor to monitor vitals

 

L360 installed. messy installation in the first place

 

 

I'm still fine tuning my vcore, so far 4.4@1.315; uncore 39@1.3; VCCIN 1.9  max temp below 80 and stable.

 

Question is, when i change my clock to 4.5Ghz, i've to move up to 1.42v and still getting 101.  is there other setting that i should tuning to make it stable without jumping more that 1.4v ??

[ProKoN]

I'm sorry I'm an overclocking noob, but I recently built a pc which you can check the full specs in my profile.  But, I'm wondering whether you think my components could do the mild overclock because I know my motherboard's was not designed for overclocking:

 

mobo: Asus gryphon z87 w/ armor kit

 

cpu: i5 4670k

 

cooler: hyper 212 evo. 

 

I've been pretty blessed with the parts I think, because I've got like a golden SSD with crazy speeds over the maximum rated speeds according to Samsung magician, and so I'm tempted to overclock my cpu and see if I'm decently lucky there too.  

 

 

well maybe luck is on your side. if your lucky you can use the guide and hit 4.4GHz at 1.25V or less.

 

or you can push for a mild OC and try to get stable at 4.0GHz on all cores with  stock voltage or just a small voltage bump. (1.150V)

[ProKoN]

L360 installed. messy installation in the first place

 

 

I'm still fine tuning my vcore, so far 4.4@1.315; uncore 39@1.3; VCCIN 1.9  max temp below 80 and stable.

 

Question is, when i change my clock to 4.5Ghz, i've to move up to 1.42v and still getting 101.  is there other setting that i should tuning to make it stable without jumping more that 1.4v ??

 

 

you can try to bump up vccin to 2.0-2.2V see if that helps.

 

x101 means you need more vcore typically.

 

Glad to see the L360 install! thanks for the pic

 

you must be pushing max temp around 1.4V

[Akolyte]

well maybe luck is on your side. if your lucky you can use the guide and hit 4.4GHz at 1.25V or less.

 

or you can push for a mild OC and try to get stable at 4.0GHz on all cores with  stock voltage or just a small voltage bump. (1.150V)

Thanks, I'll try go for 4.2 as I've heard thats a common mild overclock.  Or I might just go for 4GHZ, as I want to start small and take it one step at a time. 

[ProKoN]

Thanks, I'll try go for 4.2 as I've heard thats a common mild overclock.  Or I might just go for 4GHZ, as I want to start small and take it one step at a time. 

 

right on bud.

 

lets us know how you make out.

[Akolyte]

right on bud.

 

lets us know how you make out.

I'm really nervous, uhh. Is there some specific thing I need to do before I overclock like turn off certain features? 

[ProKoN]

I'm really nervous, uhh. Is there some specific thing I need to do before I overclock like turn off certain features? 

 

 

 

 

watch the videos and read the guide.. dont be nervous, we were all rookies at some point. take your time, be patient and learn. over time you will gain confidence and be able to help others.

 

your not gonna be pushing to hard at all.

[Jeffee]

Finally delidded my 4770k today. Gonna have to assume that the thermal compound intel had was barely touching my cpu or my h100i wasn't sitting properly even though im sure it was as idles were normal but a 4.6Ghz OC went to 97c when I was doing my oc's a few months ago. Its a lot warmer now since its spring and the 4.6Ghz temps were a 35c drop when I was only expecting a 10c drop. trying your 4.7 oc and i've capped at 76c running intel extreme tuning utility same test as before that was hitting 97.

[ProKoN]

Finally delidded my 4770k today. Gonna have to assume that the thermal compound intel had was barely touching my cpu or my h100i wasn't sitting properly even though im sure it was as idles were normal but a 4.6Ghz OC went to 97c when I was doing my oc's a few months ago. Its a lot warmer now since its spring and the 4.6Ghz temps were a 35c drop when I was only expecting a 10c drop. trying your 4.7 oc and i've capped at 76c running intel extreme tuning utility same test as before that was hitting 97.

 

 

wow nice bro!

 

 

what tim are you using?

 

your results will vary. you may need less voltage than me! my chip is a voltage hungry dog.

 

who knows maybe 4.8 is in the books.

[Jeffee]

wow nice bro!

 

 

what tim are you using?

 

your results will vary. you may need less voltage than me! my chip is a voltage hungry dog.

 

who knows maybe 4.8 is in the books.

 

I got Liquid ultra for under the IHS and shin-etsu g751 for the h100i as it was the brand corsair recommended.