Newb Overclocking 2500k

[calarahil]

Hey, so a while back I dabbed into overclocking a little bit but the temps were reaching points where I was reaching a high blue screen rate and while I would increase the voltage based on the blue screen message it put my temps over what I felt comfortable so I scaled back to stock.

Fast foward I decided to give it another go because I had significantly improved my cable management and removed all my hard drive cages for more positive air flow from my front 2 case fans. I've been using this guide.

http://www.overclock.net/t/1198504/complete-overclocking-guide-sandy-bridge-ivy-bridge-asrock-edition#

My current Specs

 

Intel i5 2500k

Corsair H100i w/ 2x Noctua NF-F12's

Kingston Fury 16GB DDR3 1866

Gigabyte GTX670oc windforce 3

Corsair AX850 with white invididually braded cables

ASUS Xonar Essense STX

Fractal Design Define R4 w/ 3x AF140's

 

So where I am confused is the offset portion. I liked the offset a bit more because it keeps the boost tech. Where as with fixed you are always at that fixed clock. What I don't get however is this.

 

Offset Voltage: +0.005v ~Also known as Vcore, and Vcc.
~Master Control that controls how much voltage your CPU gets

 

What confuses me about this is I don't know where to go when I need to change my voltage do I go down do I go up what is a good voltage what is not. currently I'm running at 0.005v at 4.2 Ghz with OCCT with Linpack I was getting around 75-83 C tops while idle my CPU hangs around 36 C.

 

Anyway I could just use some direction as to where I should go with this overclock should I scale back? Should I just go back to fixed and try and find a good voltage?

 

 

 

[JokuTyyppi]

To me that looks pretty damn good if it doesn't crash with that little over voltage! You can try lowering that voltage but I believe you are good now if you dont want any hassle. Of course you can try to up that clock speed also if that is not enough for you The right voltage can only be achieved with a lot of testing. Now just prime95 or some other stress test for several hours to see if it is stable, if not up the voltage a bit. And that temp is fine cos in real world use it will never go that high

[Deddie]

When you use offset you offset the automatic voltage that your motherboard gives your CPU and that is usually higher than it needs to be. So you should try to put the offset to - and test how low you can go.

[calarahil]

When you use offset you offset the automatic voltage that your motherboard gives your CPU and that is usually higher than it needs to be. So you should try to put the offset to - and test how low you can go.

 So by lowering the voltage with off set should I go slowly as I move down how would you move down into negative figures?

 

To me that looks pretty damn good if it doesn't crash with that little over voltage! You can try lowering that voltage but I believe you are good now if you dont want any hassle. Of course you can try to up that clock speed also if that is not enough for you The right voltage can only be achieved with a lot of testing. Now just prime95 or some other stress test for several hours to see if it is stable, if not up the voltage a bit. And that temp is fine cos in real world use it will never go that high

 

Yea, I'd toy with the voltage I just dont understand offset at all. Fixed is pretty straight forward 1.2 volts to 1.3 volts play around. I do like prime95. Only reason I've been using OCCT is because TTL uses it Tiny Tim Logan.

[JokuTyyppi]

 

Yea, I'd toy with the voltage I just dont understand offset at all. Fixed is pretty straight forward 1.2 volts to 1.3 volts play around. I do like prime95. Only reason I've been using OCCT is because TTL uses it Tiny Tim Logan.

 

 

Yea i had hard time understanding offset too Then I found adaptive voltage and i love it! But i believe your mobo doesn't have that feature yet. I used fixed for a little time but then i had to change it cos my pc is mainly on idle all the time so it would be stupid to have high voltage when it now runs at 0.7v idle and around 1,03v under load. In AVX enabled stress tests it goes 1,13v but i have no program that uses avx That 0,1v up in voltage comes from that adaptive voltage.

[calarahil]

Yea i had hard time understanding offset too Then I found adaptive voltage and i love it! But i believe your mobo doesn't have that feature yet. I used fixed for a little time but then i had to change it cos my pc is mainly on idle all the time so it would be stupid to have high voltage when it now runs at 0.7v idle and around 1,03v under load. In AVX enabled stress tests it goes 1,13v but i have no program that uses avx That 0,1v up in voltage comes from that adaptive voltage.

 

Adaptive sounds nice. I feel like my CPU while strong is still showing its age a little coming up within the next year on broadwell and then skylake (the series I'll more then likely upgrade to).

 

So I found a article about off-sets was curious what you guys thought. Was thinking moving from 0.05 offset to -0.10. From what I gather you move the meter up .05 if you need more volts or down 0.05 to see if you can get lower volts and therefore lower temps. I like how offset is because it still uses turbo and isn't running the cpu at full cpu speeds all the time but having a set voltage that someone recommends without finding the best one for my chip left me a little uneasy and hopefully I'm on the right track.

 

http://forum.overclock3d.net/showthread.php?t=38867

[SpecTre12345]

You have a H100i? Why not push to at least 4.4GHz or more?

 

And your temps are way too high for a 4.2GHz overclock, definitely not mounted or installed correctly. I was getting those temps with a stock H100 @ 4.8GHz @ 1.45v with my 2500k. 

[HazMatt]

I agree on the temps being too high for 4.2GHz. Check what your voltage is at under load. It might be best to go with a fixed voltage rather than leave it on auto because at 4.2 you should not have to pump too much voltage for temps to be bad, even while at idle. For reference my 2500k only needs 1.185V in bios to reach 4.0Ghz which is less than the factory stock voltage, though after 4.0 it starts to climb.

Edit : For 4.2Ghz my CPU needs 1.245V

[calarahil]

I agree on the temps being too high for 4.2GHz. Check what your voltage is at under load. It might be best to go with a fixed voltage rather than leave it on auto because at 4.2 you should not have to pump too much voltage for temps to be bad, even while at idle. For reference my 2500k only needs 1.185V in bios to reach 4.0Ghz which is less than the factory stock voltage, though after 4.0 it starts to climb.

Edit : For 4.2Ghz my CPU needs 1.245V

I'll go back and try fixed I believe I have a saved overclock within my bios I can just go in and try that voltage and see what happens. I don't personally feel that I need to reach a super high overclock but a little overclock to give me a little more performance wouldnt hurt.

 

You have a H100i? Why not push to at least 4.4GHz or more?

 

And your temps are way too high for a 4.2GHz overclock, definitely not mounted or installed correctly. I was getting those temps with a stock H100 @ 4.8GHz @ 1.45v with my 2500k. 

 

No I quite agree. However, I want to say its more or less linpack as it reallllly stress out the CPU. I was asking about offset because that was part of the guide. I'm sure once I work on my volts the temp will go down. It's installed correctly. Has IC Diamond 24 on it and I was careful with the amount I used. Also to note its hard to mess up the mounting for a H100i.

[SpecTre12345]

No I quite agree. However, I want to say its more or less linpack as it reallllly stress out the CPU. I was asking about offset because that was part of the guide. I'm sure once I work on my volts the temp will go down. It's installed correctly. Has IC Diamond 24 on it and I was careful with the amount I used. Also to note its hard to mess up the mounting for a H100i.

 

It's not LinPack I was getting 75-80C load on Intel Burn Test, which raises temps much higher than LinPack. Mind you again, that was at 4.8GHz @ 1.45v with my 2500k on a stock H100. You have an H100i and you are only at 4.2GHz. Temps should not be nearly that high. You probably put on way too much TIM. With better TIM, less is better. You need less than a pea sized amount placed in the center of the IHS. Then you just tighten down the waterblock/pump in a diamond pattern. The pressure from the block will spread the paste evenly throughout. 

[calarahil]

It's not LinPack I was getting 75-80C load on Intel Burn Test, which raises temps much higher than LinPack. Mind you again, that was at 4.8GHz @ 1.45v with my 2500k on a stock H100. You have an H100i and you are only at 4.2GHz. Temps should not be nearly that high. You probably put on way too much TIM. With better TIM, less is better. You need less than a pea sized amount placed in the center of the IHS. Then you just tighten down the waterblock/pump in a diamond pattern. The pressure from the block will spread the paste evenly throughout. 

 

I would imagine that it would vary between test to test. I don't get as high as temps with intel burn test as I do with OCCT. I used that method for my CPU pea size/grain that isn't the issue. Also, I know better then to assume that all CPU's are created equally. Silicon lottery. Either way I'll play around with fixed and update with my temps. I came here for help didn't come here to have someone make claims of being careless.

[SpecTre12345]

I would imagine that it would vary between test to test. I don't get as high as temps with intel burn test as I do with OCCT. I used that method for my CPU pea size/grain that isn't the issue. Also, I know better then to assume that all CPU's are created equally. Silicon lottery. Either way I'll play around with fixed and update with my temps. I came here for help didn't come here to have someone make claims of being careless.

 

I am helping you (the only one who pointed this out mind you), your temps aren't normal. Take the advice or leave it. Your overclock is at 4.2GHz, and you are hitting high 70's to low 80's. Mind you with Sandy Bridge, which doesn't get that hot to begin with. You ignoring my advice is a bad move. Your temperatures aren't right. It has nothing to do with OCCT or LinPack. Nobody is making claims of you being careless, until now, since you are thinking that 80C is acceptable temperatures for a 4.2GHz overclock. That's careless. You should be at mid to high 60's tops. 

[Faa]

Offset just works like this, stock vcore + <offset> = new vcore

Also if your vcore goes up under avx, figure your load line calibration out. You should have a higher vcore at idle and let it drop a bit under load by a small amount.

[SkilledRebuilds]

I use a "Fixed" voltage for my 2600K

(What I know it can do for 4.8Ghz), however enabling C1E power saving it still drops to 1.8Ghz and with the lower intended voltage that requires, which is the main thing not keeping higher volts on idle.

 

But... all motherboards vary so... yeah.

[jmaster299]

I bumped my 2600K up to 4.7GHz on air cooling and get better temps.

http://linustechtips.com/main/topic/179069-having-fun-with-sandy-bridge/

[calarahil]

I agree on the temps being too high for 4.2GHz. Check what your voltage is at under load. It might be best to go with a fixed voltage rather than leave it on auto because at 4.2 you should not have to pump too much voltage for temps to be bad, even while at idle. For reference my 2500k only needs 1.185V in bios to reach 4.0Ghz which is less than the factory stock voltage, though after 4.0 it starts to climb.

Edit : For 4.2Ghz my CPU needs 1.245V

Hey I wanted to follow up with a test I did with prime 95. I'm also currently doing a test as well with intel burn in test. I'll be able to finish that once I get the 20 tests to complete I placed it on maximum so it should be running pretty crazy. Now that I have this 4.2 ghz stable overclock at 1.25v how far would you go before you went to 1.3? Based on the temps I'm seeing thus far with intel burn test my max is 72. For w.e reason OCCT is the most brutal and hardest test I have done out of the 3 main ones. 

 

So I did a prime95 with real temp in the background showing temps I took a picture. Temps were good. I'll write down the settings I had in case I did something wrong. I have a feeling perhaps I had done something wrong I went back and changed something things that I had from what I had previously done. 

 

 

CPU Ratio: All Core

All Core: 42

Spread Spectrum: Disabled

Intel Speedstep tech: enabled. 

Intel Turbo Boost tech: enabled. 

additional turbo voltage: lowest I can set it to +0.004

Internal PLL overvoltage: disabled

 

These were suggested to be set to max however they noted to put 10,000 to get max but it would only reach Max when I did only reached 500. I left these alone on auto

 

Core Current Limit:

Long Duration Limit:

Long Duration maintained: 

Short Duation Power limit: 

Secondary Plane current limit: 

 

CPU Core Voltage: Fixed 1.25 (my motherboard goes from 1.20 to 1.25. As you can see based on the temps based on prime 95 maybe I could push the multiplier up if I decided to.

CPU load line calibration: I set this to 2 

 

 

Core Current Limit: Auto 

 

 

 

When I turned the test off temps went down at low as 29c and my current max is 67c. This is just with Prime95 other tests may be different I'll move onto intel burn test. 

[HazMatt]

@calarahil Looks pretty good. You should be fine leaving all of the limiters on auto. I would disable speedstep to keep the CPU at 4.2GHz all of the time as you wont see too much difference with idle temps. If you do go to 1.3V you may be able to push 4.4Ghz or 4.5Ghz though I would try to keep all the core temps below the mid 70's. You could sacrifice silence and turn up the fans if you wanted to, you could also take the middle drive cage out of your case if you aren't using it to give you more airflow.

[calarahil]

@calarahil Looks pretty good. You should be fine leaving all of the limiters on auto. I would disable speedstep to keep the CPU at 4.2GHz all of the time as you wont see too much difference with idle temps. If you do go to 1.3V you may be able to push 4.4Ghz or 4.5Ghz though I would try to keep all the core temps below the mid 70's. You could sacrifice silence and turn up the fans if you wanted to, you could also take the middle drive cage out of your case if you aren't using it to give you more airflow.

 

Thanks! Yea I know I did everything right just thought perhaps it was a setting thing. I like my computer to be kept pretty quiet hence why I got the Fractal R4 and at 4.2 its reasonably quiet. Not sure what to expect though if I were to turn up the clock a bit more maybe try 1.3v and turn up the multiplier then down volt. But when I do raise my volts won't my temp go up or am I not taking into account the fans and cooling making up the difference and it won't be as big of a jump as I'd think?

 

[HazMatt]

Thanks! Yea I know I did everything right just thought perhaps it was a setting thing. I like my computer to be kept pretty quiet hence why I got the Fractal R4 and at 4.2 its reasonably quiet. Not sure what to expect though if I were to turn up the clock a bit more maybe try 1.3v and turn up the multiplier then down volt. But when I do raise my volts won't my temp go up or am I not taking into account the fans and cooling making up the difference and it won't be as big of a jump as I'd think?

If you raise the voltage you would have to raise the fan speed to keep the same temps. You could try to bump your core up to 4.3 or 4.4Ghz just to see if it is stable at those speeds with 1.25 volts and if it isn't stable leave it at 4.2 or bump the voltage up. Unless you really need the performance I would stay at the highest stable clock you can get with 1.25V given it would be quiet and still be good temp wise.