How To Overclock Your CPU: Intel Haswell & Haswell-Refresh CPU Overclocking Guide (4670K, 4770K, 4690K, 4790K, 5820K, 5930K, 5960X)

[CC268]

Although this isn't Haswell related, I have found AIDA64 and XTU to be VERY easy to pass on Skylake. x264 and RealBench are considered to be the "middle of the road" tests for Skylake. Interestingly, I can pass AIDA64 (FPU only and Full) and XTU at 4.6GHz on 1.28V. However, with RealBench and x264 I fail within minutes. It takes me about 1.32ish V for me to be stable on RealBench and x264 - this is much more inline with what I have seen from others with Skylake as well. 

 

Anyways...I don't know if this applies to Haswell, but you may want to check (KidA39) your OC with RealBench and x264 because 4.8GHz on 1.35V is pretty impressive.

[KidA39]

Interesting, I've been doing some gaming in between stress tests just out of curiosity and I've actually found I've lost about 15 to 20 FPS in Witcher 3. I was gonna wait to bring it up until I had finished the process (like finding my cache ratio and Vring and setting adaptive voltage).   

[v1raloner]

Okay I passed with these settings...

9hrs XTU CPU Stress Test

Vcore 1.37 x49

Vring 1.20 x40

Memory 1333Mhz Auto 1.5volts

100% Vdroop

Adaptive Enabled

Everything else on Auto.

I will post up screen shots in a bit...

How does this look?

What next?

Edit:

Okay I tried to run Realbench x264 and I get BSOD, the "whea" one, what is this?

Raise vcore?

Must vcore be on adaptive?

[CC268]

I am not an expert on OCing Haswell so I would defer to BigDay. It has just been my findings from Skylake and my research (which has been A LOT) that XTU and AIDA are very "light tests". It seems that RealBench and x264 custom (over on overclock.net) is a much better test - more middle of the road. If you are failing RealBench then yes you would raise your Vcore...

[v1raloner]

I am not an expert on OCing Haswell so I would defer to BigDay. It has just been my findings from Skylake and my research (which has been A LOT) that XTU and AIDA are very "light tests". It seems that RealBench and x264 custom (over on overclock.net) is a much better test - more middle of the road. If you are failing RealBench then yes you would raise your Vcore...

I agree..

With my 4.7Ghz overclock I can pass with 1.2 vcore for 8 hours, but I haven't really mess with x264 with the methods in this forum, however. When I was going by the OCN thread which I frequent*aka home forum*, I am able to pass 25 runs of x264 and 4 hours of Realbench stress and THEN pass 10 loops with the bench mark. I have a funny feeling I won't be passing it with 1.2 vcore.

[CC268]

Yea haha...it's crazy how I can do 1.28V and 4.6GHz on AIDA and XTU for 14+ hours each but it won't pass 5 minutes of x264 and Realbench. Now if x264 and Realbench were considered to be some of the hardest tests then I wouldn't care...but they are considered to be of medium "hardness". So I just feel like AIDA and XTU maybe aren't the best indicators of stability.

So far I haven't even passed realbench on 1.315V...I think I should be good on 1.32V though...we will see.

[v1raloner]

I agree with XTU, and it's along the lines of using AIDA64.  Not as stressful.

Guides say to use more than 1 type of stress test and it so happens that I use RealBench and x264 v2.06 and I fail miserably in the latter.

 

I read somewhere there is an average "bump" of vcore when using these "light" tests to be considered stable, I believe it was like .02 or something.

 

I asked BIgday to trying stressing with the aformentioned tests since he put together this guide, don't get me wrong the info is priceless but I question the procedure of the actual test.

[BigDay]

@v1raloner, @CC268

 

stability, when it comes to cpu overclocking, is a loose term and it's relative

 

in general, you want to find stability in your day to day tasks. if you're gaming, you want to make sure that your pc is stable during games. if you render videos all the time, you want to make sure it's stable for that. if your stable during your tasks, just leave the overclock alone

 

all of these benchmarks are just general measures of stability. some of the benchmarking software is just too hardcore. you won't even come close to stressing your cpu that hard in the real world

 

my recommendation is to just use xtu to find a general sense of stability and then test your system with the overclocked settings by just going through the motions of your day to day work whether that be gaming or video editing etc. if it's stable for that stuff, your good to go

 

obviously, if you want to pass some of the more hardcore benchmarks and "stability software", you're going to need to adjust certain settings

 

just because you fail in realbench doesn't mean your pc is not stable. it may be stable in games and general tasks and be just fine for your use

 

edit:

 

i've used xtu for all my cpu overclocking processes and i'm stable during games like gta v and my general day to day tasks

[BigDay]

Okay I passed with these settings...

9hrs XTU CPU Stress Test

Vcore 1.37 x49

Vring 1.20 x40

Memory 1333Mhz Auto 1.5volts

100% Vdroop

Adaptive Enabled

Everything else on Auto.

I will post up screen shots in a bit...

How does this look?

What next?

Edit:

Okay I tried to run Realbench x264 and I get BSOD, the "whea" one, what is this?

Raise vcore?

Must vcore be on adaptive?

 

when you're stress testing, you never want to use adaptive mode for your vcore and vring. always use static. xtu won't overvolt your cpu when you use adaptive, but just to be safe, use static

 

i would forget about realbench and other software, just use xtu. what do you use your pc for anyway?

 

you have a choice:

1. you can either increase your core ratio to 50 and leave the vcore at 1.37v, re test. the goal here is to find the highest core clock
2. leave your core ratio at 49 and decrease the vcore to 1.36v, re test. the goal is to find the most power efficient core clock

again, vcore and vring must be set to static in your bios

 

edit:

 

what were your temps during your most recent stress test? max average temperature and maximum temperature?

[CC268]

Good points BigDay, I certainly don't disagree with what you said. 

[BigDay]

Good points BigDay, I certainly don't disagree with what you said. 

 

have you overclocked your 980 ti yet?

 

what's your current overclock on your 6600k?

[CC268]

I have done some initial OCing on my 980TI, but I am going to do final tuning after I am done with my CPU. I am sticking with stock voltage on my 980TI as I may keep this card for a while. As of now I am on 109% power, +100 Core Clock (4500MHz) and +500 Memory Clock (8000MHz) - fairly conservative, but like I said if I am going to keep the card for a while then I would rather be conservative.

 

I am still testing my 6600K, but I am using RealBench and x264. Although, I may just go by XTU like you said. I have realized that you can spend days and days with this overclocking stuff. I may just go with XTU and then call it good.

 

Right now I am at 1.325V and 4.6GHz (should be stable - I will see when I get home from work). However, with XTU I would be at like 1.28V and be stable.

[CC268]

Yea...I may try your XTU method...I don't even pass x264 on 1.325V.

[CC268]

BigDay,

 

Would you recommend just using my RAM XMP profile or manually entering it in? Some guys over on OCN recommend manually entering it in, but I am not sure I am comfortable doing that (there is a ton of RAM settings in my BIOS).

[v1raloner]

This is my 4.9Ghz/4.0Ghz/1333Mhz Memory, once I get this 10 hours out of the way I am gonna call it stable for now... Then get my memory in check.

THEN get to 4.9Ghz, question for you guys, how does my chip look? Is it average or good?

[BigDay]

BigDay,

 

Would you recommend just using my RAM XMP profile or manually entering it in? Some guys over on OCN recommend manually entering it in, but I am not sure I am comfortable doing that (there is a ton of RAM settings in my BIOS).

 

i don't know. i have no experience in entering any form of ram settings into the bios manually. if you want to learn how to do it manually, go for it. if not, just use the xmp profile. i'm not sure which method is better, but i would guess that the difference is minimal or we'd be seeing tons of threads pop up about it.

[BigDay]

snip

 

it's a good chip for sure

[Timmahnz]

Hey, are the Gigabyte Z87 series motherboards not able to achieve the adaptive mode on the cache frequency and Vring?

[BigDay]

Hey, are the Gigabyte Z87 series motherboards not able to achieve the adaptive mode on the cache frequency and Vring?

 

no idea, sorry. you can try playing around with the settings. do you know which ones i'm talking about?

[Timmahnz]

no idea, sorry. you can try playing around with the settings. do you know which ones i'm talking about?

 

To be honest.. I do not have a clue I haven't overclocked anything since my old Q6600.

 

Today's aim was to try get to 4.2Ghz using the lowest voltage.

[BigDay]

To be honest.. I do not have a clue I haven't overclocked anything since my old Q6600.

 

Today's aim was to try get to 4.2Ghz using the lowest voltage.

 

i managed to overclock my 4670k to 4.6 ghz on the core @ 1.32v (vcore)

[BigDay]

Update: October 28, 2015 @ 12:34 PM

 

Added a new section that provides Gigabyte Z97 and X99 users the ability to set their Cache Ratio and Vring to adaptive mode through the use of Intel Extreme Tuning Utility (or XTU). This section has also been added to the Google Doc:

 

In order to set the Vcore and Vring parameters to adaptive mode on a Gigabyte Z97 or X99 motherboard, you have to use Intel Extreme Tuning Utility (XTU). Follow these steps:

1. Restart your PC and boot into the UEFI/BIOS
2. Set the UEFI/BIOS to its optimized defaults (stock settings)
3. If your overclocked settings require a different VCCIN or VRIN than the stock setting, set this parameter now to the setting you found stability with during your stress tests
4. If you have found stability with the XMP profile enabled on your memory, enable this setting now; adjust the memory voltage if needed (you should have found a stable setting for the memory voltage during your stress tests)
5. Boot into Windows and load XTU in administrator mode
6. Set your Core Ratio, Vcore, Cache Ratio (or Uncore Ratio), and Vring to their overclocked values in XTU (do NOT click “Apply” yet!). The values that are being adjusted should be highlighted in yellow
7. Now that you have set all of the overclocked values in step #6, you may now click the “Apply” button. Do NOT click “Apply” until you have set each value to their overclocked state
8. Save these settings as a profile by clicking “Save”

We use XTU because some Gigabyte boards do not allow you to set adaptive mode for the Cache Ratio (or Uncore Ratio) and the Vring.

Sometimes when you restart your PC and boot into Windows, your overclocked settings will revert back to their stock values. If this happens, go back into XTU and select your desired overclocked profile and click “Apply”. Make sure the profile you wish to enable on a consistent basis is #1 in the saved list in XTU.

If you adjust one setting at a time and click “Apply” in XTU, it may result in a crash, especially if you tweak the voltage settings. This is why I recommend you adjust all of the settings first and then click “Apply” at the end.

 

[v1raloner]

when you're stress testing, you never want to use adaptive mode for your vcore and vring. always use static. xtu won't overvolt your cpu when you use adaptive, but just to be safe, use static

 

i would forget about realbench and other software, just use xtu. what do you use your pc for anyway?

 

you have a choice:

• you can either increase your core ratio to 50 and leave the vcore at 1.37v, re test. the goal here is to find the highest core clock
• leave your core ratio at 49 and decrease the vcore to 1.36v, re test. the goal is to find the most power efficient core clock

again, vcore and vring must be set to static in your bios

 

edit:

 

what were your temps during your most recent stress test? max average temperature and maximum temperature?

Max temps in XTU are 78-81℃ 9hour"stable" average Max temps are like 60-77℃. Check screen shot of my run.

[v1raloner]

With the above settings when I enable power savings, would I "still" be stable?

I've been tightening my 4.7Ghz and 4.8Ghz overclock*paused on doing 4.9Ghz testing*.

4.7Ghz at 1.205 VID

4.8Ghz at 1.285 VID

Both with 48 runs of X264 8 hours XTU and 10 Realbench benchmarks

[BigDay]

With the above settings when I enable power savings, would I "still" be stable?

I've been tightening my 4.7Ghz and 4.8Ghz overclock*paused on doing 4.9Ghz testing*.

4.7Ghz at 1.205 VID

4.8Ghz at 1.285 VID

Both with 48 runs of X264 8 hours XTU and 10 Realbench benchmarks

1. yes, you should still be stable even while enabling the power saving features inherit to haswell and your motherboard
2. 4.7 and 4.8 is relatively easy in terms of finding stability. 4.9 is the more challenging number to hit and find stable