Intel Haswell 4670k + 4770K Overclocking guide

[BigDave]

Since I am super lazy, has anyone reached 5GHz on Haswell 4770k yet?

No lazy people have reached that high :-)

[theawesomejman]

No lazy people have reached that high :-)

So 4790K Devils Canyon it is...

[ProKoN]

Hello,

First of all great guide and a great read through all the helpful posts!

I felt the need to create an account here as i just had a go at overclocking my new 4670k in my newly built rig, and this is where i got the most help to get stable at 4.4ghz.

Im fairly new to overclocking and sadly I got one of those voltage hungry chips that wants to put up a fight. Ive been through so many guides and different bios configs over the last days that im almost burnt out.

But thanks to this guide i was able to reach 4.4ghz on my chip with a gigabyte z87x d3h mobo.

I really had to push the voltage to get there with this chip and its now sitting at 1.285v in the bios. Though HWmonitor show VID at 1.286 and it shows vcore at 1.296v under stress tests for some reason. I also turned off everything that could effect the vcore but no change. Not that it matters because i suspect that the 1.296v vcore is what is needed for stable and i have tuned the vcore with this in mind after discovering it.

My uncore is at 3.5ghz and 1.20v. I'll try to tweak this later on. So far I've established that i cant do 1:1.

My vrin is at 2.020

I've done a 8 hour XTU stress test and also a 10 hour Aida64 stress test as well as several 10 round standard runs in Intelburntest for quick heat and initial stability tests.

Max temp in XTU was at 76c. In Aida64 max temp was 84. And in intelburntest it varies a bit but 87c i think was the highest. My cpu cooler is a Corsair h100i set to exhaust in a Nzxt h440 case.

My initial goal when starting out was an oc at 4.5. But because of the chip being voltage hungry, and with the heat that follows, my cooler isnt fit to the task. Im quite happy with 4.4 with the situation being what it is.

Also i think the sweetspot for my chip is at 4.3ghz because that required way less vcore. I think it passed intelburntest at 1.175v....

So again, thx for the great guide. It really helped me!

 

 

Welcome to the LTT community!

 

Thank you for posting your results!

 

Thank you for you kind comment

 

post your results in meltymoons DB and jumpers cinebench thread!

 

enjoy your stay.

[ProKoN]

So 4790K Devils Canyon it is...

 

keep us posted! can not wait to see some DC results.

 

im thinking 5.0ghz on DC is just a pipe dream, but hey we are allowed to dream right

[theawesomejman]

keep us posted! can not wait to see some DC results.

 

im thinking 5.0ghz on DC is just a pipe dream, but hey we are allowed to dream right

Haha I'm about to pull the trigger on a pre-order... Should I go for it and HOPEFULLY receive it by June 6th at most? (Expected in warehouse the 30th)

[theawesomejman]

keep us posted! can not wait to see some DC results.

 

im thinking 5.0ghz on DC is just a pipe dream, but hey we are allowed to dream right

What cooling would you suggest for the beast, I believe I can reach 5 Ghz if it turbos to 4.4 like they claim.

[ProKoN]

What cooling would you suggest for the beast, I believe I can reach 5 Ghz if it turbos to 4.4 like they claim.

 

i suggest 240mm aio or better

 

h220

h320

 

extreme water 2.0\3.0

 

h105

 

I think its gonna hit 4.1 max turbo, dont think we gonna see 4.4. time will tell.

[paleale]

Hi all, i'm in need some help overclocking an ASRock MOB. I have been following this guide as best i can with the diffrent terminology used between diffrent manufactures. I'm overclocking a 4670K with a new MOB, ASrock Z97 pro4. I have found system stibilty for 24+ hours with the following setting.

4.7GHz @ 1.40V

4600MHz ring bus @ 1.40V

VCCIN 2.0V

DDR3- 1866MHz 1.6V

 But whenever i try to run the adaptive mode for either the core or ring voltage i get BSOD. I even tryed reseting to factory setting with adaptive voltage and couldn't even log into windows. If you could point me in the right diection it would be very much apprecitaed.

Cheers.

[ProKoN]

Hi all, i'm in need some help overclocking an ASRock MOB. I have been following this guide as best i can with the diffrent terminology used between diffrent manufactures. I'm overclocking a 4670K with a new MOB, ASrock Z97 pro4. I have found system stibilty for 24+ hours with the following setting.

4.7GHz @ 1.40V

4600MHz ring bus @ 1.40V

VCCIN 2.0V

DDR3- 1866MHz 1.6V

 But whenever i try to run the adaptive mode for either the core or ring voltage i get BSOD. I even tryed reseting to factory setting with adaptive voltage and couldn't even log into windows. If you could point me in the right diection it would be very much apprecitaed.

Cheers.

 

 

can you please post some screen shots of your bios. Im not familiar with asrock's parameters, but we will get er figured.

 

thanks for posting.

 

welcome to the LTT community!

[paleale]

Thanks for the quick reply ProKon, and thanks for the guide. Its been very helpful. Here are the screen shots of the bios. I hopefully someone can tell me what im doing wrong :blink:

[ProKoN]

Thanks for the quick reply ProKon, and thanks for the guide. Its been very helpful. Here are the screen shots of the bios. I hopefully someone can tell me what im doing wrong :blink: 20140529_200610.jpg

here is my best guess for 1.4V adaptive voltage.

 

cpu vcore voltage mode - adaptive

 

Vcore adaptive voltage - 0.800V

 

Vcore voltage addition offset - 0.600V

 

This is just a thought you may have to play with the adaptive voltage and offset to get er right.

 

my thought is it will drop down to 0.800V while idle and add an additional 0.600V under load.

 

play with those two values, keep the frequency stock until you figure out what combination of values will give you the desired results. same principle applies for the cache values.

 

fairly similar to an asus board, confusing.

 

msi makes this a breeze - input 1.4V set for adaptive mode...done.

 

let us know how you make out.

[JoeBlough]

Wanted to compliment Mr. Prokon for preparing this.  Looked at many guides, all the same pretty much, this one gave some additional insights.  I started to "do the math" and couldn't belive the amount of time put in with 8 hour tests after each tiny voltage increase.  I have had great success overclocking over the years which I account to two things:

 

1.  Don't "go cheap" when buying components

2.  Been Lucky in the CPU Lottery.

3.  Read a lot

 

When built my son's box with a 2600k was at 4.8 GHz and 25% GFX OC within hours.....same w/ most builds.   When I built this box, I was very lucky.... I set Vcore, put voltage control on Adaptive and everything worked up to 46 Multiplier / 46 cache which were my original goals.   That was w/ BIOS 0804.  After a BIOS upgrade, I never saw those OCs again...... put minimal effort into 11xx and 12xx and finally got around to trying on 1402.  Reading this thread helped a bit tho I'm too impatient .... well better sad, I just don't have the time .... to use small voltage increases and run 8 hour test in between.

 

While not as detailed and unlikely to produce as much useful information as the guide, I took a "Minimlist Approach".  Hopefully Prokon has some tweaking ideas to the following to make my "quick and dirty approach" more useful.  I'd advise using what he wrote if ya can as it's waaaay better and likely to produce more fine tuned results.... but if ya have all those annoyanmces like a wife, kids, business to run, etc .... this is what I did.

 

1.  Set VCCIN to 1.9 and CPU / Cache Voltage Vontrol to Adaptive

2.  Started w/ Auto on everything else till I got a failure

3.  Started at 42 Multiplier, 39 cache, 1600 RAM

4.  Increased to 42 cache

5.  Increased RAM to XMP

6.  After failures on tests with cache and RAM at minimum increased VID by 0.025

7.  After failure with Cache increase, boosted that in steps of 0.125 till cught up with VID

8.  After failures with RAM at 2400, increased voltage to 1.7

9.  Went to 43 Multiplier and repeated above but did cache in 2 steps  39-41-43

10.  Repeated for 44 CPU Multipliers (39-42-42)

11.  Repeated for 45 CPU Multiplier .... (39 - 42 - 45)

12.  Repeated with 46 CPU Multiplier but did cache in 5 steps.... 39 - 41 - 43 - 45 - 46

 

A.  I did voltage bumps of 0.025 for VID initially, later at 0.0125 after I hit 1.25 volts, used 0.0125 for VCC until it hit VID.

B.  I ran the RoG Real Bench Benchmark (< 10 minutes) for initial stability purposes.  Since the only OC I'd save for each CPU Multiplier would be the one with the highest cache ratio, all those intermediate steps were just that .... intermediate.  I saw no need to "confirm" thru long stability testing and I simply don't have the time.  For the final set of settings .... i.e. 45 multiplier / 45 cache / 2400 RAM, I ran the 8 hour IETU (only uses 2-4 GB) but I also ran the RoG Real Bench Stress test for 2 hours and told it to use all 16GB.  I found that I passed IETU on several occasions but still got failures under RoG Real Bench Stress test .... usually between at about 90 minutes.  Did a few 8 hour RoG tests but never got a failure after 90 minutes fo stopped doing.

 

I had initial stability at 1.275 volts VID but not with AVX.    At this point, 45/45/2400 is quite stable with AVX at:

 

VCore 1.318 VCC Ring 1.318 VCCIN (Ev) 1.880 DRAM 1.700

 

Here's the Temps and Observed Settings during that test if anyone finds the data interesting ..... note that the max voltage observation was during the RoG Real Bench Open CL test which uses AVX and spikes voltag about 0.010 - 0.013.  Average Temps are averaged over the duration of the RoG Real Bench Benchmark (Start to Finish)

 

Idle / Average / Max Temps

Core 0 26 49 66

Core 1 26 49 66

Core 2 23 46 60

Core 3 26 50 58

 

Setting / BIOS reading / Observed Reading (w/ AVX)

VCore 1.318 1.040 1.424

VCC Ring 1.318 1.318 1.359

VCCIN (Ev) 1.880   1.904

DRAM 1.700 1.674 1.738

 

Got there after 1.30 / 1.30 failed ..... 1.325 / 1.3125 passed and settled on 1.318.   Have a test running at VCCIN at 1.88 right now (1.85 failed at 91 minutes).

 

The some questions for Prokon  and anyone else still following who has more time invested in this than I do .....

 

-Once getting stable and then trying a lower VCCIN with success, is it worthwhile going back and trying to lower VID and VCC Ring ?

 

-With adaptive, our CPUs will bounce back and forth between 35 and 45 at these settings depending on laod and "what's needed", so why not set the cache ratio to operate over a range also (Min = 39 / Max = 45) ?

 

-With regard to previous comments on Adaptive and AVX, nothing I have read indiucates that the voltage bump is "unpredictable".  You can expect a voltage bump of 0.10 volts minimum.  I typically see a bump of 0.10 to 0.13 ... and that brings me to my question.    At my stable but as yet dialed in 46/46/2400 OC, my VID is at 1.375; thru 3 of the 4 RoG Real Bench tests, my peak instantaneous voltage is 1.424 ..... under AVX in the Open CL test, it climbs to 1.488.  The average voltage in the 2 hour 16GB RoG Stress test is 1.375  So when we talk about safe voltages, a topic I have never seen discussed is ..... what's the relative impact of average voltage that I would think is most applicable to "wear and tear" versus peak voltages which are of short duration ?

[ProKoN]

 

The some questions for Prokon  and anyone else still following who has more time invested in this than I do .....

 

-Once getting stable and then trying a lower VCCIN with success, is it worthwhile going back and trying to lower VID and VCC Ring ?

 

-With adaptive, our CPUs will bounce back and forth between 35 and 45 at these settings depending on laod and "what's needed", so why not set the cache ratio to operate over a range also (Min = 39 / Max = 45) ?

 

-With regard to previous comments on Adaptive and AVX, nothing I have read indiucates that the voltage bump is "unpredictable".  You can expect a voltage bump of 0.10 volts minimum.  I typically see a bump of 0.10 to 0.13 ... and that brings me to my question.    At my stable but as yet dialed in 46/46/2400 OC, my VID is at 1.375; thru 3 of the 4 RoG Real Bench tests, my peak instantaneous voltage is 1.424 ..... under AVX in the Open CL test, it climbs to 1.488.  The average voltage in the 2 hour 16GB RoG Stress test is 1.375  So when we talk about safe voltages, a topic I have never seen discussed is ..... what's the relative impact of average voltage that I would think is most applicable to "wear and tear" versus peak voltages which are of short duration ?

 

Hey Joe!

 

Welcome to the LTT community.

 

Thank you for sharing your results and your insight! its refreshing

 

You have alot to talk about!

 

As per you VCCIN question, I endorse dropping all voltages as low as you can whilst maintaining  stability. this includes VID, VCCIN and VCC ring.  a slight .025-0.50 bump never hurts.

 

 

 

-With adaptive, our CPUs will bounce back and forth between 35 and 45 at these settings depending on laod and "what's needed", so why not set the cache ratio to operate over a range also (Min = 39 / Max = 45) ?

 

I think your referring to the way asus bios operates cache. most manufactures dont have this min\max cache option. for asus i suggest Min =8 / Max=45. your cache should drop down to 800mhz while idle if your adaptive voltage parameters are set correct. and turbo to 4.5 under load.

 

 

 

With regard to previous comments on Adaptive and AVX, nothing I have read indiucates that the voltage bump is "unpredictable".  You can expect a voltage bump of 0.10 volts minimum.  I typically see a bump of 0.10 to 0.13 ... and that brings me to my question.    At my stable but as yet dialed in 46/46/2400 OC, my VID is at 1.375; thru 3 of the 4 RoG Real Bench tests, my peak instantaneous voltage is 1.424 ..... under AVX in the Open CL test, it climbs to 1.488.  The average voltage in the 2 hour 16GB RoG Stress test is 1.375  So when we talk about safe voltages, a topic I have never seen discussed is ..... what's the relative impact of average voltage that I would think is most applicable to "wear and tear" versus peak voltages which are of short duration ?

 

it is not recommended to run past 1.45V as a daily setting. i have taken my chip past 1.7V for benchmarking fun. these short bursts of 1.424-1.488 are perfectly fine. i personally feel 1.5V or less is perfectly fine for the average life expectancy of a chip, 4-5year.

 

I dont suggest running any stress test with adaptive voltage. were these results of 1.4V+ on manual or adaptive settings?

[Lays]

keep us posted! can not wait to see some DC results.

 

im thinking 5.0ghz on DC is just a pipe dream, but hey we are allowed to dream right

 

If it's possible, I'll sure as hell try

 

 

Did you see what I was planning on downgrading to?  (Getting a used 290x for ~300$, a waterblock for it, and an m power z97 + 4790k when it's released)  

 

after I buy everything, i'll still have more performance than I need for 1440p, and I'll have an extra $1000 in my pocket

Probably gonna get an r4 + a 360mm rad and a 240mm.

[JoeBlough]

Much Thanks for the welcome

 

<As per you VCCIN question, I endorse dropping all voltages as low as you can whilst maintaining  stability. this includes VID, VCCIN and VCC ring.  a slight .025-0.50 bump never hurts.>

 

I explained poorly..... with 1.9 VCCIN, ..... I failed, I upped VIN .... I failed, I upped VCCring ...... I failed, I upped VIN .... I failed, I upped VCCring ....

 

Now I have success .... so I try 1.89 VCCIN ... yeay ..... I try 1.88 VCCIN....yeay ..... I try 1.87 VCCIN ..... boo failure

 

So I go back to 1.88 and save my "known good settings".   NOW .... with a new VCCIN at 1.88, do I then go back and try those lower Vcore, VCCring and DRAM voltages that had previously failed at VCCIN at 1.9 ?   In other words being able to lower VCCIN by 0.02, is there any probability that somehow the VIN of 1.315,  the VCCring of 1.312 or the 1.675 DRAM  that failed before now has a shot.

 

<I think your referring to the way asus bios operates cache. most manufactures dont have this min\max cache option. for asus i suggest Min =8 / Max=45. your cache should drop down to 800mhz while idle if your adaptive voltage parameters are set correct. and turbo to 4.5 under load.>

 

Any time I have ever asked that question, everyoine said "must be same"......even on Asus RoG forums ..... Well that chnages things doesn't it

 

I seem to recall having passes at 46/46/43 that didn't work at 46/46/46 (CPU / Cache Max / Cache Min) ..... Given it took a day to dial in 45/45/45/2400 .... I think i will try Cache Min with 46.  This was so ez w. BIOS 0804 .... set adaptive, VIN and left everything else on Auto and it worked.

 

 

<it is not recommended to run past 1.45V as a daily setting. i have taken my chip past 1.7V for benchmarking fun. these short bursts of 1.424-1.488 are perfectly fine. i personally feel 1.5V or less is perfectly fine for the average life expectancy of a chip, 4-5year.

 

I dont suggest running any stress test with adaptive voltage. were these results of 1.4V+ on manual or adaptive settings?>

 

On a daily basis, the box will bever see AVX.  

 

These are the Vcore readings under RoG Real Bench

 

Vcore Readings RoG RB for 1.318 VID    (BTW,, under 0804 I was stable at 1.275 ./ Auto everything    1.325 Auto everything worked on 1102

Image Ed. - 1.328
Encoding - 1.328
Open CL - 1.424 (uses AVX)
Multitask - 1.328 (use all 3 above at same time)

 

I wouldn't think of running any straight synthetic stress test with AVX above 1.375 VID as that would put the bump over 1.50 .... but RoG RB bumps it up only instantaneously and HWiNFO64 is recordimg the max setting.  Under BIOS 0804, I was maxing out at 46/46/2400 at 1.424 volts .... under the newer BIOS I was seeing peaks of up to 1.488 volts.   My thought ..... er hope  .... was that I was fine with the 3 tests topping bout at 1.39 or so (goal was < 1.40) and let those instantaneous peaks run up to 1.50 .... which it will only see when doing stability testing.

 

So basically to saw it with less words..... if I run the 8 Minute RoG Benchmark..... pay attention to the Max column in HWiNFO64 or the Average

 

If ya have any suggestions on how to approack 46 plz throw it out there.... BTW, I have a complete spreadsheet of all my test runs w/ 1402 BIOS from 42/39/39/1600 thru 45/45/45/2400.  If ya would like to have the data, let me kno where to send the file (xls).

 

Does everybody get "pink" between yellow and red on VCCIN (1.99 -  2.07) or just Asus Folksies ?  Was surprised to see red start at 2.08 given most of the guides recommending 2.1 and 2.2 beingjust fine.  I interpreted the color chnages as:

 

White = Your fine

Yellow = Caution, make sure yoy have good temps

Red = Danger Zone

 

Don't know what to make of pink..... the only thing I could come up with is "Pink = If you are seeing this number as pink, it's time to stop fiddling with your damn computer all day all night and pay some attention to your wife, GF, partner, significant other"

 

Interesting Comparison RoG Image Editing Benchmark (4 runs .... 1st one always sucks)

 

46 Multiplier / 39 max cache / 8 min cache

94,310 / 122,001 / 121,612 / 120,880

 

46 Multiplier / 39 max cache / 39 min cache

96,455 / 123,834 / 122,419 / 122,594

 

 

EDIT:   Looks like 46/46/2400 under BIOS 1408 will be a challenge

 

45 CPU / 45 max / 45 min / 2400 passed at 1.318 VIN / 1.318 cache / 1.88 VCCIN / 1.7 DRAM

 

I figured most of the results I have seen were 0.06 or so for the jump in VIN from 45 to 46 .... but I tried to do better .... didn't work.... So boosted VCCIn to 2.07 (end of pink range) and did no betetr at 1.325 1.35 1.362

 

So OK, 1.318 + 0.6 = 1.378

 

With 46 CPU / 39 max / 39 min / 1600 .... used 1.375 VIN / 1.375 cache / 2.07 VCCIN / Auto DRAM and passed .... failed w/ 8 minutes let at 45 / 41 / 41 / 1600

 

Currently at 46 CPU / 41 max / 41 min / 1600 with 1.38 VIN / 1.38 cache / 2.07 VCCIN / Auto DRAM

 

Likely be my last effort of a long unproductive day ....any pointers for me to save me some blind alleys ?

 

BTW, just noticed that temps are 64 - 72 w/ 1.38 VID whereas on previously failure at 1.375 worse was 77 ?     ....seen that before and it's just too danged puzzling

 

 

EDIT:  Last run of night

 

46/41/41/Auto

CPU-Adapt.    / Cache-Adapt.  / 1600
Ambient = 19.7

Coolant (Idle) = 22.2

Idle / Average / Max
Core 0 / 25 / 55.40 / 71
Core 1 / 24 / 56.20 / 70
Core 2 / 20 / 53.40 / 64
Core 3 / 23 / 50.10 / 59

Setting / BIOS / Observed
VCore / 1.800 / 1.040/ 1.488
VCC Ring / 1.380 / 1.403 / 1.419
VCCIN (Ev) / 2.070 /   / 2.096
DRAM / Auto / 1.674 / 1.684

 

Likely be my last effort of a long unproductive day ....any pointers for me to save me some blind alleys ? Like increasing VCCIN at this level raise / lowers Vring / Vcore ?

 

EDIT 2:  Breakfast Run

 

46/43/43/Auto

CPU-Adapt.    / Cache-Adapt.  / 1600
Ambient = 19.2

Coolant (Idle) = 21.8

Idle / Average / Max
Core 0 / 23 / 56.7 / 70
Core 1 / 22 / 56.6 / 69
Core 2 / 20 / 53.5 / 66
Core 3 / 21 / 51.10 / 60

Setting / BIOS / Observed
VCore / 1.380 / 1.040/ 1.488
VCC Ring / 1.380 / 1.403 / 1.422
VCCIN (Ev) / 2.070 /   / 2.093
DRAM / Auto / 1.674 / 1.684

 

EDIT 3.  Lunch Run

 

After 2 hours if stress testing w/ RoG Real Bench .... completed 8 minute RoG Benchmark and system spontaneously rebooted anout 4 seconds after completion as I moved windows around to read HWiNFO64 max voltages ..... these are the things I hate about OCing.

[PocketSmiley]

Hello everyone,

 

Could you guys confirm if your setup fails the Prime95-Haswell Version 28.1 "In-place Large FFTs (maximum power consumption)" test?
 
I'm proud to report that I was successful in overclocking my 4670k using this B85 chipset motherboard. However, I have a problem with the Prime95-Haswell Version 28.1, specifically the "In-place Large FFTs (maximum power consumption)." Right now, mine reliably shows "Fatal Error: rounding" after ten minutes in that test. I used to get Windows 8 blue screens for Clock_Watchdog_Timeout/WHEA_UNCORRECTABLE_ERROR when undervolted (these disappear when I increase vcore), but I solved that by increasing the uncore multiplier to x38 without increasing the vcore. I have no problem whatsoever in IETU nor in Prime95's "Small FFTs" test, both run for more than 10 hours.
 
My preferred setting, set in the Gigabyte BIOS:
Core Mult: x43 (4.3GHz)
Turbo Boost enabled, all cores x43
Vcore: 1.155V
Uncore Mult: x38 (VERY IMPORTANT FOR ME!)
Everything else set to auto.
Cooler: Zalman CNPS10X (Cooler Master Hyper 212 EVO/PLUS clone)
Temperature range: fluctuates between 55-65°C in IETU, 60-70°C in Prime95 "Small FFTs (max temp)."
 
Highest reached:
Core Mult: x47 (4.7GHz)
Turbo Boost enabled, all cores x47
Vcore: 1.4V
Uncore Mult: x38
Everything else set to auto.
Temperature range: 70-85°C in IETU, 80-90°C in Prime95. (Too hot.)
 
I didn't take pictures while testing, but here are current CPU-Z screenshots:

[JoeBlough]

I didn't understand the "cache = 38 being very important to you".  Just to clarify .... did you mean ....

 

-You need it therefrom an application performance standpoint as a minimum?

-You need it there to stay stable ?

-You need it there cause oitherwise you need more cache voltage and then things get too hot.

 

Like most peeps who use their PC's for work and play, I am using Adaptive Control on COU and Cache .... that pretty much nixes using Prime 95 as the AVX instructions will add 0.10 to 0.13 volts to ya Vcore.  So not having the experience, I'll have to bow out of iffering any advice there.

[PocketSmiley]

I didn't understand the "cache = 38 being very important to you".  Just to clarify .... did you mean ....

 

-You need it therefrom an application performance standpoint as a minimum?

-You need it there to stay stable ?

-You need it there cause oitherwise you need more cache voltage and then things get too hot.

 

Like most peeps who use their PC's for work and play, I am using Adaptive Control on COU and Cache .... that pretty much nixes using Prime 95 as the AVX instructions will add 0.10 to 0.13 volts to ya Vcore.  So not having the experience, I'll have to bow out of iffering any advice there.

 

Heya,

 

As I said in my post, "I used to get Windows 8 blue screens for Clock_Watchdog_Timeout/WHEA_UNCORRECTABLE_ERROR when undervolted (these disappear when I increase vcore), but I solved that by increasing the uncore multiplier to x38 without increasing the vcore." So reasons 2 and 3 above. I don't have "adaptive control" on my BIOS, the G.1 Sniper B5 (B85 chipset; I got this because it was $70 before, can overclock like a Z87, and best of all, it has an upgradeable op-amp.)

 

The whole point of my post was to confirm if Prime95 IS the problem, as I've read so many times everywhere. Because its "In-place large FFTs" test always fails my setup unless I increase the vcore to disproportionate levels (4.3GHz @1.3V no way). I want to know if I truly need to pass that test, when IETU stress and "small FFTs" tests are stable.

[JoeBlough]

On my board, default cache (aka uncore) is 39 .... it's the increasing to 38 (from 39) part that stumps me.

 

Asa far as P95 goes, I don't use it ..... ever.   At least not since P68 days.   To my mind, it has no applicability to system performance under real world applications.\ back in the day I could see it's applicability as OC was primarily limited by temps.   But now voltage is now the enemy and we have so many things on the CPU die that P95 does not test.  Plus I have to winder, if we worried about AVX stability, the associated 0.013 voltage bump is a concern and I'm thinking all the stress testing will put more AVX load on the CPU that a CPU is likely see see in it's lifetime unless using specialized apps.

 

I use RoG Real Bench .  If you can pass the Multitasking portion of the Benchmark you have a 2 minute indicator that you are very close to stability.   I have never passed a 2 hour 16GB stress test and Real bench and failed IETU.

[Perpatratur]

Hey,

 

Great guide, brought me up to speed after a long time away from intel.

 

Have I got the worst chip in the world?

 

1.465v to get stable at 4.4ghz (mem at 1333 and vin at 2.0v)

 

i5 4670k

Gigabyte Z87X-UD3H

Patriot Black Mamba 2400Mhz DDR3

Antec Kuhler 920 h2o

 

 

I haven't even tried a full stress test. Amazed it lasted the hour it did. I had so much trouble just posting or getting to the desktop.

 

[PocketSmiley]

Hey,

 

Great guide, brought me up to speed after a long time away from intel.

 

Have I got the worst chip in the world?

 

1.465v to get stable at 4.4ghz (mem at 1333 and vin at 2.0v)

 

i5 4670k

Gigabyte Z87X-UD3H

Patriot Black Mamba 2400Mhz DDR3

Antec Kuhler 920 h2o

 

 

I haven't even tried a full stress test. Amazed it lasted the hour it did. I had so much trouble just posting or getting to the desktop.

 

Hi. I have the same proc. and a B85 Gigabyte board with similar BIOS options as yours:

 

Mine:

 

Yours:

 

The only things I changed were:

 

CPU Clock Ratio: x43 (4.3GHz)

Turbo Boost enabled, all 4 cores set to x43

Vcore: 1.155V

Uncore Ratio: x38 (above this I get random freezes and blue screens)

Everything else set to auto.

 

I was able to reach 4.7 GHz but it was getting too hot (see my other post above).

 

I still haven't passed Prime95, and as Joe (and originally, Prokon) pointed out, overclocked Haswell chips aren't for programs compiled for AVX (one popular example being Handbrake).

[Perpatratur]

Hi. I have the same proc. and a B85 Gigabyte board with similar BIOS options as yours:

 

Mine:

 

Yours:

 

The only things I changed were:

 

CPU Clock Ratio: x43 (4.3GHz)

Turbo Boost enabled, all 4 cores set to x43

Vcore: 1.155V

Uncore Ratio: x38 (above this I get random freezes and blue screens)

Everything else set to auto.

 

I was able to reach 4.7 GHz but it was getting too hot (see my other post above).

 

I still haven't passed Prime95, and as Joe (and originally, Prokon) pointed out, overclocked Haswell chips aren't for programs compiled for AVX (one popular example being Handbrake).

 

Yep, that is mine. I will give it a whirl tonight. I highly doubt it though. Just to get it stable for gaming and general usage (it was fine over the weekend) I am at x40 1.3v (left uncore at default, x34 and 1.015v I think but xmp memory.) Any stress test whether it is prime, intel extreme tool or aida just blue screens it within 10 minutes

[PocketSmiley]

Yep, that is mine. I will give it a whirl tonight. I highly doubt it though. Just to get it stable for gaming and general usage (it was fine over the weekend) I am at x40 1.3v (left uncore at default, x34 and 1.015v I think but xmp memory.) Any stress test whether it is prime, intel extreme tool or aida just blue screens it within 10 minutes

 

I don't know if this is Gigabyte specific, but leaving uncore ratio at default x34 gave me similar problems to yours. I had to raise vcore up to silly just to get stable. By raising uncore ratio to x38, I can lower vcore to acceptable.

[Socratic]

Seems a great thread. I'll be sure to have a read through it when I decide to start tweaking again. I'm at 4.6 with 1.27v. Can't remember what my uncore is, probably 1:1, but when or if I plan on going for 4.8 I'll have a read through.

[Perpatratur]

I don't know if this is Gigabyte specific, but leaving uncore ratio at default x34 gave me similar problems to yours. I had to raise vcore up to silly just to get stable. By raising uncore ratio to x38, I can lower vcore to acceptable.

 

I gave it a whirl. No joy at 38 for me. I had better luck with 36. Stable at x40 and 1.25v. 19vin. Terrible but better than nothing I suppose.