Intel Haswell 4670k + 4770K Overclocking guide

[BigDay]

It depends on the title or application. 1600Mhz is the standard, but higher memory clocks can yield benefits. I've seen 3000Mhz DDR4 take 5 frames or more over 1600Mhz DDR3, but this could be due to the architecture. It's probably on par with Cache Ratio overclocks. Some applications love it, some couldn't care less, others are amicable. I might try it for fun.

 

 

Nice! Yeah if you do overclock your memory, let me know how it goes!

 

Question:

 

Should I avoid stress testing my CPU with the Uncore Ratio and Vring set to AUTO? I was thinking about doing this so that I could potentially run my 4.6GHz core clock with these settings 24/7, but if these settings are set to AUTO, there will be no way of knowing if everything will be stable during 24/7 use because the CPU may be running all kinds of different Uncore and Vring settings. Am I right?

 

Thanks!

[AngryGoldfish]

Nice! Yeah if you do overclock your memory, let me know how it goes!

 

Question:

 

Should I avoid stress testing my CPU with the Uncore Ratio and Vring set to AUTO? I was thinking about doing this so that I could potentially run my 4.6GHz core clock with these settings 24/7, but if these settings are set to AUTO, there will be no way of knowing if everything will be stable during 24/7 use because the CPU may be running all kinds of different Uncore and Vring settings. Am I right?

 

Thanks!

 

AFAIK, setting the Uncore Ratio and Uncore Voltage to auto will do nothing destructive. That is the stock setting and the motherboard will not draw any more than the CPU can handle. It won't run crazy amounts of Vring or cache frequency unless you're running a stress test, which should not be done unless you are overclocking and have set the voltage to static or fixed. That's the way I read it. As long as you don't run abnormal stress tests—which prime95, IETU, AIDA64, OCCT all are—setting your Uncore and Vring to auto should not do any harm.

 

[BigDay]

I'm trying to complete a full 8 hour stress test with IETU with a core frequency of 4.7 GHz

 

Settings:

 

4.7GHz Core Frequency @ 1.34V Vcore

3.4GHz Cache Frequency @ 1.20V Vring

1.90V VRIN

1333MHz RAM

 

I made it through 6 hours in the previous test before getting a BSOD. My temps in this test were:

 

Maximum: 81 C

Average: 73 C

 

I'm on the 2nd test at the moment:

 

4.7GHz Core Frequency @ 1.35V Vcore

 

Temps so far:

 

Maximum: 81 C

Average 74 C

 

Question: Is it safe to run these long tests at these temperatures?

[arniks8]

I'm trying to complete a full 8 hour stress test with IETU with a core frequency of 4.7 GHz

Settings:

4.7GHz Core Frequency @ 1.34V Vcore

3.4GHz Cache Frequency @ 1.20V Vring

1.90V VRIN

1333MHz RAM

I made it through 6 hours in the previous test before getting a BSOD. My temps in this test were:

Maximum: 81 C

Average: 73 C

I'm on the 2nd test at the moment:

4.7GHz Core Frequency @ 1.35V Vcore

Temps so far:

Maximum: 81 C

Average 74 C

Question: Is it safe to run these long tests at these temperatures?

It's safe, but some people don't like it that high ( talking about maximum, average is perfectly fine ).

[SnakeEyes]

I'm trying to complete a full 8 hour stress test with IETU with a core frequency of 4.7 GHz

 

Settings:

 

4.7GHz Core Frequency @ 1.34V Vcore

3.4GHz Cache Frequency @ 1.20V Vring

1.90V VRIN

1333MHz RAM

 

I made it through 6 hours in the previous test before getting a BSOD. My temps in this test were:

 

Maximum: 81 C

Average: 73 C

 

I'm on the 2nd test at the moment:

 

4.7GHz Core Frequency @ 1.35V Vcore

 

Temps so far:

 

Maximum: 81 C

Average 74 C

 

Question: Is it safe to run these long tests at these temperatures?

 

nice hitting 4.7!

 

something to keep in mind, stress tests are exactly that. they do not mimic everyday PC usage for 99.9% of people. if you use your PC mainly to game you will rarely (if ever) max out the CPU so those temps will not be common. another thing i learned a few years ago OC'g an i7 920, failing a stress test doesn't mean it will fail at what you use it for. i could never pass the traditional "8 hour" test at 4ghz. i could sit and play games for 8 hours without problems though.

[AngryGoldfish]

I agree, stress tests are not the be-all and end-all, and neither are the temperatures they elicit from your CPU. IETU doesn't draw that much heat compared to Prime95, and neither are they reflective of real-world applications. Certain benchmarks can push your CPU, but not in the same way. If your CPU is hitting 85°C after an hour of Prime95, your gaming max temperatures after 5 hours of solid gaming might only be 40°C.

[BigDay]

Thanks for the replies guys! I appreciate it . I've learned a lot from you guys on this thread

 

Will I be doing any damage to my CPU while stress testing at those temps for 8 hours?

 

Just completed a Cinebench R15.0 benchmark with the following score:

 

686

 

Settings:

 

4.7 GHz @ 1.35 Vcore

4.0 GHz @ Auto Vring

1.80 VRIN

1600 MHz RAM

[arniks8]

Thanks for the replies guys! I appreciate it . I've learned a lot from you guys on this thread

Will I be doing any damage to my CPU while stress testing at those temps for 8 hours?

Just completed a Cinebench R15.0 benchmark with the following score:

686

Settings:

4.7 GHz @ 1.35 Vcore

4.0 GHz @ Auto Vring

1.80 VRIN

1600 MHz RAM

8h test won't damage it.

[arniks8]

BTW nice voltage for 4.7Ghz IMO.

[arniks8]

Sorry double post.

[AngryGoldfish]

Thanks for the replies guys! I appreciate it . I've learned a lot from you guys on this thread

 

Will I be doing any damage to my CPU while stress testing at those temps for 8 hours?

 

Just completed a Cinebench R15.0 benchmark with the following score:

 

686

 

Settings:

 

4.7 GHz @ 1.35 Vcore

4.0 GHz @ Auto Vring

1.80 VRIN

1600 MHz RAM

 

High voltages (more than 1.35V I would say) can cause CPU degradation over long periods of time, but as already said running mulitple stress tests should not harm your CPU. It is still only software causing the heat spikes and stressful environment, unlike voltages and even frequency, which are a hardware function as well as software.

 

[BigDay]

High voltages (more than 1.35V I would say) can cause CPU degradation over long periods of time, but as already said running mulitple stress tests should not harm your CPU. It is still only software causing the heat spikes and stressful environment, unlike voltages and even frequency, which are a hardware function as well as software.

 

 

Where did you read that voltages above 1.35V cause CPU degradation?

 

I've enabled the power saving functions on my MOBO with regards to my CPU. My core and cache frequency and voltages are set to adaptive. This seems like it would keep my CPU healthier for longer by reducing the voltages and thus heat

[BigDay]

When I'm stress testing, is it okay to leave Graphics Voltage on Auto? Or, should I set it to static?

[AngryGoldfish]

Where did you read that voltages above 1.35V cause CPU degradation?

 

I've enabled the power saving functions on my MOBO with regards to my CPU. My core and cache frequency and voltages are set to adaptive. This seems like it would keep my CPU healthier for longer by reducing the voltages and thus heat

 

1.35V is not guaranteed to cause degradation. It's a theoretical point I like to make based on the lifespan of chips used by professional overclockers. I only say it as a precaution to those that think that as long as you're below 90°C, it doesn't matter. If you apply 1.45V to your CPU, even with an exotic water loop and large radiator surface area, you are risking voltage leakage over time. The CPU's were never meant to experience those kinds of volts. 1.35V is perfectly fine. If the CPU does degrade, all you have to do is increase the voltage to 1.36V every second year, shall we say. This is not a big deal, and neither is it guaranteed, but it can happen and it can confuse people. GPU and CPU overclocks never stay stable forever. New hardware changes, different motherboard BIOS, different stress test version, updated operating system, these can all change the dynamics of your overclock.

 

 

edit: And yes, unless you render videos on a daily basis as well as gain for multiple hours, using adaptive voltage settings will greatly extend the theoretical lifespan of your CPU. You should be absolutely fine.

[BigDay]

1.35V is not guaranteed to cause degradation. It's a theoretical point I like to make based on the lifespan of chips used by professional overclockers. I only say it as a precaution to those that think that as long as you're below 90°C, it doesn't matter. If you apply 1.45V to your CPU, even with an exotic water loop and large radiator surface area, you are risking voltage leakage over time. The CPU's were never meant to experience those kinds of volts. 1.35V is perfectly fine. If the CPU does degrade, all you have to do is increase the voltage to 1.36V every second year, shall we say. This is not a big deal, and neither is it guaranteed, but it can happen and it can confuse people. GPU and CPU overclocks never stay stable forever. New hardware changes, different motherboard BIOS, different stress test version, updated operating system, these can all change the dynamics of your overclock.

 

 

edit: And yes, unless you render videos on a daily basis as well as gain for multiple hours, using adaptive voltage settings will greatly extend the theoretical lifespan of your CPU. You should be absolutely fine.

 

Thanks for the great info dude. Very good to know. I appreciate it.

[AngryGoldfish]

No worries, dude!

[Ryoku]

@ProKoN

 

So finally after about a year and a half of owning this i5 4670k I've decided to get around to overclocking it. Your guide has been super helpful so far, would have been lost without it that's for sure.  Starting with the basics I did your initial 4.4ghz settings and that passed an 8 hour IETU stress test for 8 hours hitting an average temp of 51c.  Figuring those were pretty solid numbers I bumped it up to 4.6ghz which Posted, booted into windows and was able to start the stress test, this however failed shortly after starting.

 

So current settings are:

• Multiplier 45
• CPU vcore Voltage 1.25v
• VCCIN Voltage 1.9v
• Memory 1333mhz @ 1.5v

So these settings passed a 12 hour ISTU Stress test, with a max temp of 64c and an average of 56c.  At this point I think I'm going to try and get 4.6ghz stable and then work on the ring bus ratio and memory,  What do you guys think?

 

[ProKoN]

@ProKoN

 

So finally after about a year and a half of owning this i5 4670k I've decided to get around to overclocking it. Your guide has been super helpful so far, would have been lost without it that's for sure.  Starting with the basics I did your initial 4.4ghz settings and that passed an 8 hour IETU stress test for 8 hours hitting an average temp of 51c.  Figuring those were pretty solid numbers I bumped it up to 4.6ghz which Posted, booted into windows and was able to start the stress test, this however failed shortly after starting.

 

So current settings are:

• Multiplier 45
• CPU vcore Voltage 1.25v
• VCCIN Voltage 1.9v
• Memory 1333mhz @ 1.5v

So these settings passed a 12 hour ISTU Stress test, with a max temp of 64c and an average of 56c.  At this point I think I'm going to try and get 4.6ghz stable and then work on the ring bus ratio and memory,  What do you guys think?

 

 

 

looks good. you have thermal headroom to increase Vcore and push for 4.6 or 4.7 if you want....its up to you

[Ryoku]

looks good. you have thermal headroom to increase Vcore and push for 4.6 or 4.7 if you want....its up to you

 Right, I'll give 4.6ghz a whirl and if it give me to much trouble I'll fall back to 4.5 or 4.4, been using the 3.8ghz boost till this point so not to worried about it =P

[BigDay]

@ProKoN

 

So finally after about a year and a half of owning this i5 4670k I've decided to get around to overclocking it. Your guide has been super helpful so far, would have been lost without it that's for sure.  Starting with the basics I did your initial 4.4ghz settings and that passed an 8 hour IETU stress test for 8 hours hitting an average temp of 51c.  Figuring those were pretty solid numbers I bumped it up to 4.6ghz which Posted, booted into windows and was able to start the stress test, this however failed shortly after starting.

 

So current settings are:

• Multiplier 45
• CPU vcore Voltage 1.25v
• VCCIN Voltage 1.9v
• Memory 1333mhz @ 1.5v

So these settings passed a 12 hour ISTU Stress test, with a max temp of 64c and an average of 56c.  At this point I think I'm going to try and get 4.6ghz stable and then work on the ring bus ratio and memory,  What do you guys think?

 

 

ProKoN recommends that you stay under 78-80 degrees celcius in terms of an average temperature. You've got lots of room to go. Push that baby passed 4.6 GHz on the core. Up the Vcore if necessary.

 

Right now, I'm running the following settings which are stable:

 

4.6 GHz Core @ 1.29V

4.0 GHz Cache @ Auto Vring (goes up to a max of 1.36V)

1600 MHz RAM @ 1.5V, XMP profile

 

I'm still working on the custom cache frequency and voltage

[Ryoku]

ProKoN recommends that you stay under 78-80 degrees celcius in terms of an average temperature. You've got lots of room to go. Push that baby passed 4.6 GHz on the core. Up the Vcore if necessary.

 

Right now, I'm running the following settings which are stable:

 

4.6 GHz Core @ 1.29V

4.0 GHz Cache @ Auto Vring (goes up to a max of 1.36V)

1600 MHz RAM @ 1.5V, XMP profile

 

I'm still working on the custom cache frequency and voltage

 

I'm also looking at upgrading my cooling to a Swiftech H240x and adding in another Rad + GPU Block, might be able to get those temp down more.  Either way 4.5-4.6ghz I'll be happy with either.

[BigDay]

I'm also looking at upgrading my cooling to a Swiftech H240x and adding in another Rad + GPU Block, might be able to get those temp down more.  Either way 4.5-4.6ghz I'll be happy with either.

 

You should try and push your core frequency up to 4.7GHz. You'll have to increase the Vcore, but see how high you have to push it. I'd be curious to see what your temps would be with a higher Vcore to support 4.7GHz. If your average temps stay below 78-80, why not run with it? After that, you can always try and bring up the cache frequency and voltage.

 

What kind of motherboard do you have?

 

I have a Gigabyte. I wouldn't recommend using one in terms of OC a CPU because it doesn't offer a simple way to implement adaptive and static voltage for the core and cache. ProKoN recommends using an MSI mobo.

 

Gigabyte only offers a VRING voltage offset, which is bullshit because even if you set it properly, your uncore frequency won't budge in terms of downclocking to save power. I've tried many times without luck. I could be doing it wrong, but I doubt it...

 

Fucking Gigabyte...

[AngryGoldfish]

Yeah, I'm not a massive fan of Gigabyte's BIOS either. Same goes for ASRock. ASUS and MSI offer a good BIOS experience. I haven't tried eVGA's yet.

[BigDay]

The next board I buy will either be an ASUS or an MSI board. MSI boards seem expensive for what you get. Gigabyte boards offer a lot of features for less. At least that's what I think... am I right?

 

Anyone else think that dragon logo is hideous? Man...

 

Anyways...

 

I've been using AUTO settings for my cache frequency and cache voltage because it runs on adaptive automatically at these settings. The cache ratio bumps up to 40 on auto and the voltage adjusts to compensate. I would try and tweak the cache frequency and voltage to a higher setting, but I don't want to have them set to static all the time. I think the only way for me to set them to adaptive would be through IETU.

 

AngryGoldfish, what kind of board do you have?

[AngryGoldfish]

Gigabyte boards are usually a little bit better value for money for the same features and warranty, yeah. It's just I personally feel their BIOS is still slightly lacking.

 

I have an MSI Gaming 7 Z97 board. I had an ASRock Z87 Extreme4 before that that broke two weeks after the 1-year warranty was up. I did not like that BIOS, but it had much cleaner and clearer audio. I also quite liked its subtle looks. Why more boards can't be plain black is beyond me.

 

BigDay, are you running stress tests when your voltages are set to auto/adaptive?