Beginner Overclocking 4.4ghz with an i5-4670K, Coolermaster Hyper 212 EVO, ASROCK z87

raspute

Would appreciate some advice.

Trying to overclock to 4.4ghz with an i5-4670K, Coolermaster Hyper 212 EVO, ASROCK z87 Extreme4.

Thsese are the settings I am trying:

CPU Ratio 44
CPU Cache Ratio 38
CPU Vcore Voltage - Override 1.250
CPU Cache Voltage Auto
CPU Input Voltage 1.900 Fixed

I did a short FPU system stability test in AIDA64 and the temperatures reached the high eighties.

I want someone to have a look at the options in my UEFI and tell me if I should be changing anything else or what I should try.

Here are some pics.

http://imgur.com/a/8c7LZ

Should I turn off Turboboost in the settings?

Also, does it matter that I didn't start lower and increase up slowly, I just started at 4.4ghz. Is this bad?

game4it70

Haswell can be very hit and miss with overclocking.The temps they produce are high seeing as Intel used glue to stick the Ihs on instead of solder thus causing bad thermal transfer.There just hot feckers even with solving the Ihs problem.

First off download Hwinfo to monitor volts and temps.Look for "Vcore" max when running a test to see how much your mobo is putting through the cpu.Things like load line calibration will effect this.

Change the cpu multi to 42 instead and drop Vcore down to 1.2v.
Leave all the settings you have in the screenshots as they are bar set the cache ratio to auto for now.
Start there as with the settings you used with the 44 multi are making your temps to high.Run your tests and check temps/volts.It shouldn't crash but if it does just up the Vcore a little and repeat tests.

Let me know how you get on.

raspute

Hi, I did as you said and did a test with IntelBurnTest on High for a few minutes. here are a few snaps of the log results.

http://imgur.com/a/RU4FL

These are the temps about half an hour after the test: http://i.imgur.com/IrRRtFh.png

Then I did a stability test in AIDA64 Extreme

Test FPU: Reached around 76: http://i.imgur.com/IU3j5eO.png
Test CPU: Very short test, reached low fifties temperature.

Here is a HWinfo screengrab from during the FPU Stress Test: http://i.imgur.com/tSJe4hL.png


Should I go on to do longer stress tests overnight, or can I just leave it as it is? And no need to turn off Turbo or Speedstep or any of those things. Also, should the voltage be left at override or is there any reason to switch it to adaptive?

game4it70

When using hwinfo look for "Vcore" reading as this will tell you the actual vcore not the vid.
Use max setting for IntelBurnTest but keep an eye on temps.
Atm you temps are ok and gives you a little room.With normal usage they wont get near as high as the likes of IntelBurnTest.
Try 4.3ghz with a longer test and keep all the other settings as they are now bar extra Vcore if needed.
Just work on the cpu ratio for now and get that stable before trying to up the cache ratio.

Should I go on to do longer stress tests overnight, or can I just leave it as it is? And no need to turn off Turbo or Speedstep or any of those things. Also, should the voltage be left at override or is there any reason to switch it to adaptive?

Yes longer tests really should be done.Everyone has different ways of testing for stability.It all depends on your own opinion as to how far you test.
When testing use override but with normal day to day usage use adaptive.Using adaptive while stress testing with certain programs can use much higher volts.Its perfectly fine for day to day and will help lower volts.Speedstep and turbo can be left on all the time.

Here a good albeit long guide for haswell.One thing to note is there are some settings named differently by mobo manufacturers.For instance "Uncore" is called cache ratio on your mobo.Google will help if the guide gives you setting your mobo doesn't have.
http://www.overclock.net/t/1411077/haswell-overclocking-thread-with-statistics/0_100

Feel free to ask me any questions and i'll do my best to help.

raspute

Hi, I don't seem to have a reading for vcore in hwinfo, or does it have a different name?

http://imgur.com/a/iBaVT

I am happy with 4.2 but might try 4.3 out of interest.

What should be the temperatures I should not go beyond?

I ran prime95 for around ten hours and the highest temperature was in the low seventies.

game4it70

Thats weird,doesn't have a different name either.

I would try to keep it in the low eighties when stress testing.IBT will be hotter running than prime so keep an eye on it.

You should be able to try 4.3 then 4.4 if you dont need to raise the Vcore to much and are temps ok.Its all trial and error to find your cpu's sweet spot.

Luck100

Yeah, I have the same problem with HWInfo. For me (i3770k) it shows vcore as something like "Vin5" in the motherboard section (not with the cpu readings).

You can try CPUID HWmonitor (free version)
http://www.cpuid.com/

I was only able to figure out which voltage in HWInfo was actually vcore by comparing the numbers with HWMonitor.

You should definitely be looking at your vcore readings when overclocking, as it can be quite different from VID or what you've selected in bios.

raspute

Hi, I tried HwMonitor:

http://i.imgur.com/YEr1Vm5.png

I am currently on 4.4ghz and stress tests get temperatures in the seventies or low eighties usually

Luck100

Wow, your vcore is 0.96 during stress testing? That seems really low for a stable overclock. Temps are ok for stress tests.

game4it70

raspute wrote: »

Hi, I tried HwMonitor:

http://i.imgur.com/YEr1Vm5.png

I am currently on 4.4ghz and stress tests get temperatures in the seventies or low eighties usually

Hwmonitor is showing idle vcore so try cpuz.I've found cpuz to be pretty close to the right reading.It will give you a better idea of what you load volts are.
Also the more load line calibration you add will lower the load volts will be.

Temps are still ok.

raspute

That earlier screengrab was during an idle period but I put a very high test in IBT going and opened hwmonitor and it still seems to be 0.96 here. Perhaps it's not reliable. All this voltage stuff is beyond me, I'm afraid. I wouldn't know a good reading from a bad one.


** Just tried a test in Aida64, it might me be of more use to you. The highest the temp got was 89:

http://imgur.com/a/Ukoui


One thing I was wondering about was this GFlops business. On other 4670k clocks people were reporting GFlops readings in the low hundreds, whereas IBT always gives me around the 60-63 mark. A bit curious about what might be wrong here (if anything). Of course it's amusing because I have no idea what a Gflop means or does.

game4it70

Do you have your vcore set to adaptive or overide?
If you use IBT (maybe Aida too) it runs linpack which will overvolt the vcore if its set to adaptive.That would explain the high vcore showing in the screenshot.
So when your run a stresstest make sure you set to overide not adaptive and then switch back to adaptive for day to day use.

If you where running the stress with overide change load line calibration up to say 5 and test.See what the vcore now shows in cpuz.It should be a fair bit lower.

Gflops can be effected by ram amount,problem size and overclock.
There not to low and more importantly there consistent as if they weren't that can mean overclock instability.

raspute

I may have confused things further, as when tweaking I changed voltages slightly to try improive results, just to be clear, my vcore is now set to override 1.280. I was trying out some steps from an Intel guy on some forum, You may still be looking at my first post. I'm not sure whether 1.280 is high in your opinion, or whether it was just high when you though I had still set it to 1.250. I may put it back down, if that's what you suggest. I was simply trying to see the different results.

raspute

Found a slightly newer version of IBT and ran on high. On the 3rd run it was hitting 99 degrees so I stopped it. I'd say I am at the limit and should dial something back. Would that be the voltage?

raspute

Aida64 shows slightly better temperatures, around the 90 mark

game4it70

Ah ok that makes more sense

I would try lower that voltage for 4.4ghz.This is where the patience comes into play.Ideally you find a clockspeed you want and keep lowering the volts until it becomes unstable or start with a low vcore and work your way up.

Just to give you an idea of the way it works.
My 4670k needs 1.17v for 4.3ghz and 1.36v for 4.6ghz.I had saved profiles for 4.4,4.5ghz in the bios but i updated it so i cant remember what volts i needed

All cpus are different so yours may need more vcore or less vcore than mine to be stable.My cpu is probably average but i've seen alot worse than mine.
With your cooling i wouldn't go much past where you are now,say 1.3v max.

I still haven't seen the official safe max voltage for haswell before it leads to the cpu degrading but i've used 1.36v for 7 months and still fine.

game4it70

raspute wrote: »

Found a slightly newer version of IBT and ran on high. On the 3rd run it was hitting 99 degrees so I stopped it. I'd say I am at the limit and should dial something back. Would that be the voltage?

There is a version of linpack that absolutely stresses the feck out of you cpu and will cause high volts and temps.I ran it once and i hit 80c even thought i delided my cpu and have watercooling.:eek:
Needless to say dont use that one again.
Stick with the older version,aida64 or prime95.

raspute

What's your opinion:
on 4.4ghz

CPU RATIO x44
CPU CACHE RATIO x38

CPU INPUT V 1.9
CPU VCORE V 1.25
CPU CACHE VOLTAGE 1.25

Ran Prime95 for 5 hours, 0 errors or warnings, highest temp 80. Four cores, Max 80, 80, 77, 73. Mostly around the 75 or 76 mark as I was watching it though.

Would you be happy with that or is there anything else? Like any reason to put cache ratio up or down? I tried 1.2 voltage but got a blue screen, WHEA error. I just went to 1.25 again.

game4it70

Put the cache ratio up to 40 and drop the cache voltage down to around 1.18v-1.20v.
You see very little improvement with a higher cache ratio and adds more heat when higher.1.25v is high for cache voltage so deffo drop it down.

Cpu input can be dropped a little too.Try 1.85v.Also you could try raising the cpu input when trying 1.2v on the vcore.Its safe enough up to 2.0v but i wouldn't go higher.

As for vcore if it fails prime at 1.2v only up it to say 1.21v and try again.Even though your temps are fine no need to run at 1.25v if it doesn't need it.

Its all trial and error with patience as i said.