i7-4770K OC question

[ChuckFx]

I got my 4770k to 4,4 with 1,29 Vcore. It runs at 60C with my H100i. I a new to OC so I am wondering, should I keep pushing or stop here and be happy with what I got?

 

[wabbitslayer]

I'd say it depends on whether that voltage is manually ser or adaptive (under load), and what load you are getting that 60C temp at. I think 4.4 is the sweet spot for most non-delidded Haswells, after that point the ratio of extra volts needed to get higher speeds goes to hell.

 

[ChuckFx]

It's a set a fixed voltage, and those temps are from a stress test at 99-100% load so it keeps fairly cool for an Haswell.

 

[rtsurfer]

It's a set a fixed voltage, and those temps are from a stress test at 99-100% load so it keeps fairly cool for an Haswell.

What stress test.?

 

[ChronoReverse]

See if it survives Prime95 (updated version with AVX support) smallFFTs

 

[stuff_me_good]

Sorry to hijack this thread but does anyone know what is the default voltage for uncore? I know the default ratio is 35, but my mobo gives pretty high voltage to the uncore, even more than to the core itself.

 

[ChuckFx]

I use intel Extreme Utility Tuning to test the CPU.

 

[rtsurfer]

I use intel Extreme Utility Tuning to test the CPU.

Intel XTU isn't the most reliable way to test the stability of an OC.

Try Prime95 Small FFT with AVX or Aida64.
There is a thread pinned at the top of this CPU & OCING section by Idontcare.

While its a bit dated, the info there is still good. It has proper guides to check OC stability.
Check it out.

Also, Caution if you decide to run Prime95 or Intel Linpack, don't try it at anything higher than 1.25V.

I have the same CPU & cooler & my core temps reached 100C while running Linpack with the H100i fans at full speed at room temperature, the volts were at 1.285 I believe.

 

[Tristor]

I use intel Extreme Utility Tuning to test the CPU.

XTU is not sufficient to test stability. It's more of a benchmark than a stability test. Prime95 is also not sufficient. You need to run Intel Optimized Linpack, which will hit every single part of the math pipeline which allows you to truly stress the system under AVX2 loads and ensure full stability and verifiably correct results. Intel Burn Test uses an older build of LinX/Linpack and is useful as a tool for verifying stability/heat-load as well.

 

[ChuckFx]

Intel XTU isn't the most reliable way to test the stability of an OC.

Try Prime95 Small FFT with AVX or Aida64.
There is a thread pinned at the top of this CPU & OCING section by Idontcare.

While its a bit dated, the info there is still good. It has proper guides to check OC stability.
Check it out.

Also, Caution if you decide to run Prime95 or Intel Linpack, don't try it at anything higher than 1.25V.

I have the same CPU & cooler & my core temps reached 100C while running Linpack with the H100i fans at full speed at room temperature, the volts were at 1.285 I believe.

Ok thanks for the infos! At the light of what you say, I'll keep my CPU at 4,4 with the set voltage. Temperature is well under controll and from what I know of Haswell, it's a respectable score. I don't want to bench for the fun of it and XTU showed that the CPU is stable after hours of 100% load so it did what I needed.

Every time I stress tested with XTU, if the CPU passed it remained stable on applications afterward without a hiccup so for me, it seems reliable.

 

[rtsurfer]

Ok thanks for the infos! At the light of what you say, I'll keep my CPU at 4,4 with the set voltage. Temperature is well under controll and from what I know of Haswell, it's a respectable score. I don't want to bench for the fun of it and XTU showed that the CPU is stable after hours of 100% load so it did what I needed.

I don't know what MOBO you have.
But if you have an Asus MOBO, switch to Adaptive voltage from Manual.
I don't know what is the equivalent of Adaptive on other manufacturers, but I think everyone has one.

If you leave the voltage at Manual, your CPU will be fed the same voltage regardless of the clocks.
It will be the same 1.29V for 800Mhz & 4400Mhz.
This is not ideal in neither in the term of power consumption nor temperatures.

Adaptive allows the voltage to scale with clocks.
Manual is only ideal for stress testing.
Adaptive means stocks volts plus a bit more offset to account for the higher clocks.

 

[ChuckFx]

Ok I'll put it to adaptive. If the CPU passed stress test at 4,4 with fixed 1,29, will it handle addaptive the same way or there is a chance that it fail?

 

[rtsurfer]

Ok I'll put it to adaptive. If the CPU passed stress test at 4,4 with fixed 1,29, will it handle addaptive the same way or there is a chance that it fail?

No need to test adaptive.
You use manual to find the Volts you need for a particular clock, 1.29V for 4.4Ghz in your case.

After that juat set the offset in adaptive in such a way that stock volt + offset that you put in equals 1.29V.

It will be stable & needs no testing as long as you configure it correctly.

 

[Bubbleawsome]

If you have the option (my MSI board does override) run either adaptive or override. AVX loads can add up to .1v to the CPU (resulting in an unpleasant 1.39v for you.) Different manufacturers have different names, but there should be a mode where it is always at a constant voltage (fixed for me), a mode where it will drop the volts but not add the extra (seems to be override for me), and a mode where it will drop the volts and add the extra (seems to be offset for me) These names are a giant mess, so good luck.

 

[coffeejunkee]

XTU is not sufficient to test stability. It's more of a benchmark than a stability test. Prime95 is also not sufficient. You need to run Intel Optimized Linpack, which will hit every single part of the math pipeline which allows you to truly stress the system under AVX2 loads and ensure full stability and verifiably correct results. Intel Burn Test uses an older build of LinX/Linpack and is useful as a tool for verifying stability/heat-load as well.

Even that is not enough. My cpu will run Linx all day and crash in Arena chess simulator in minutes. Other people have used Dolphin emulator as a stresstest with good effect. My ultimate test is a 2 hour 8 player game of Supcom FA, but it's not very practical. And even then I won't really know for sure if it wouldn't crash after 3 hours.

So I'd say if it's stable enough for you it's stable enough. As for overclocking further, if you can be bothered, why not. You will find a point where you have to add too much voltage for too little gain soon enough.