[jessterb]

Hi all!

I posted this on another forum, but thought this might be the perfect place to get some feedback/suggestions!

I overclocked my 3570k to 4.3GHz (with vcore set to auto->1.27v), running solid at what looks to be reasonable temperatures:
Prime95 v28.5 -> maxing at 77c (average is 74c) -> stable
Prime95 v26.6 -> maxing at 71c (average is 67c) -> stable

Now, I'm trying to overclock it to 4.5GHz, but the only way I could do that is by having my vcore at 1.35v... anything lower and it crashes. My temperatures:
Prime95 v28.5 -> maxing at 87c (average in the low 80's) -> crashed
Prime95 v26.6 -> maxing at 82c (average in the high 70's) -> stable

So my question is, is 1.35v okay? Reading up, it seems to me I should be aiming more for around 1.25v and that my hardware should support that (?). Not sure if that's correct thinking on my part, but if anyone can clarify and maybe look at my settings below to make any recommendations... thanks!!!

Specs

CPU: 3570k
Mboard: MSI Z77A-GD65
Cooler: Cooler Master Hyper 212 EVO (with heatsink fan pushing to exhaust case fan)
Memory: G.SKILL Ripjaws X Series DDR3-1600 CL9-9-9-24
Case: Corsair 300r (one push, one exhaust... will be adding a top exhaust soon)

My OC settings for 4.5

 

[Sharchaster]

Okay as long as you can control the temps. But I will NOT use those voltage for 24/7 use, I use it for benchmark only...anything higher than 1.3 volt, I will NOT use it for 24/7 use for Ivy and Haswell / DC.

 

[Aparition]

1.35 is basically the top limit for air cooling for people aiming for 5.0 Ghz benches.
That sucks that you need it for 4.5. Also seems really off.

Are you even stable at full stock settings on auto?

 

[writer21]

Quote:

Originally Posted by Sharchaster

Okay as long as you can control the temps. But I will NOT use those voltage for 24/7 use, I use it for benchmark only...anything higher than 1.3 volt, I will NOT use it for 24/7 use for Ivy and Haswell / DC.

Been using 1.35v for 3570k for 2+ years now with Hyper Evo @ 4.7ghz. Been fine so far. Under load it drops to 1.296v though or 1.304v. I've even ran 4.8ghz @ 1.4v for a while. Up to you and depends on your temps.

 

[Sharchaster]

Quote:

Originally Posted by writer21

Been using 1.35v for 3570k for 2+ years now with Hyper Evo @ 4.7ghz. Been fine so far. Under load it drops to 1.296v though or 1.304v. I've even ran 4.8ghz @ 1.4v for a while. Up to you and depends on your temps.

Good to hear...I'm not comfortable to run 24/7 with voltage beyond 1.35 volt, though. Even the temps is fine....

 

[writer21]

Well I do turn off my pc when not in use. Also I use Intel Speedstep so when I'm just browsing it drops vcore with offset voltage. I also use c1e power saving sometimes.

I'm looking to go 6 or even 8 core Intel soon so I got plenty of good use out of this chip and I think it can go even longer.

 

[jessterb]

Quote:

Originally Posted by writer21

Been using 1.35v for 3570k for 2+ years now with Hyper Evo @ 4.7ghz. Been fine so far. Under load it drops to 1.296v though or 1.304v. I've even ran 4.8ghz @ 1.4v for a while. Up to you and depends on your temps.

Like you, this PC wouldn't be on all the time, only a few hours of game play every day... but I'm a bit iffy about 1.35v based on what I've been reading/hearing. Ideally, I wanted to be close to 1.25v with temps under 80 at full load. If you're getting 4.7 with 1.35v, I'm guessing you'd be getting 4.5 with 1.3v if not less. I'd feel more comfortable with that! But no luck so far.

 

[jessterb]

Quote:

Originally Posted by Sharchaster

Okay as long as you can control the temps. But I will NOT use those voltage for 24/7 use, I use it for benchmark only...anything higher than 1.3 volt, I will NOT use it for 24/7 use for Ivy and Haswell / DC.

Sharchaster,

This is what I was afraid of.

Quote:

Originally Posted by Aparition

1.35 is basically the top limit for air cooling for people aiming for 5.0 Ghz benches.
That sucks that you need it for 4.5. Also seems really off.

Are you even stable at full stock settings on auto?

Aparition,

Yes, I'm stable at full (3.4) stock settings, if I'm understanding you correctly.
I'm also stable at 4.3 with vcore set to auto (which is typically around 1.27v).
I'm also stable at 4.5 with vcore at 1.35v, so long as I'm running Prime95 26.6 (latest version 28.5 runs me 5 degrees hotter and I end up crashing).

I just don't get why I need 1.35v to get to 4.5GHz, which is why I'm wondering if I'm doing something wrong? This guide (http://tinyurl.com/pom9z4w) mentioned that "1.25V is a good starting value that should be enough to get any i5-3570K to 4.5GHz" and they were able to get theirs to run stably at 1.21V. Forget 1.25v as a starting point in my case, I can't anywhere there at 4.5!

Should I go back to 4.3 and check what I can lower my vcore to, just to test and see?

 

[boot318]

Been using my 3570k 4.5GHz @1.32v for 8+ months now. It is still perfectly fine. Welcome to the poor OCing 3570k club!

 

[jessterb]

Quote:

Originally Posted by boot318

Been using my 3570k 4.5GHz @1.32v for 8+ months now. It is still perfectly fine. Welcome to the poor OCing 3570k club!

Ha, is this a big club? I guess it figures with me luck...

Ran a few more tests, starting with 1.25v @ 4.3Ghz and it booted fine. Here are the rest of quick tests and attempts at lowest vcores:

4.3 + CV 1.24v (1.23v) = stable (short test), hitting 67c (average 63c)
4.4 + CV 1.24v (1.23v) = crash, couldn't boot to windows
4.4 + CV 1.30v (1.28v) = stable (short test), hitting 72c (average 69c)
4.4 + CV 1.30v (1.28v) = stable (short test), hitting 72c (average 69c)
4.4 + CV 1.29v (1.272v) = stable (short test), hitting 72c (average of 68.5c)
4.4 + CV 1.28v (1.265v) = crashed in prime95
4.5 + CV 1.31v… couldn't boot into windows

I tested the 1.29v @ 4.4 in the IntelBurnTest, just a quick 5 runs at High settings and maxed out at 81c (with an average of 63c).

Anyways, I guess this means 1.29v and 4.4Ghz might simply be the best I can get and I simply didn't get a great CPU?

Or is it possible something is awry? Could this vcore issue be related to me not applying thermal paste properly? I ask that because even at full stock, my package temp is 40c with my cores varying greatly (ie. 26c-40c-17c-34c), which I thought strange and thought I should be getting better temperatures with my 212 EVO. That said, I read that:

1) it can be normal for cores to vary greatly
and
2) more or less ignore IDLE temps, what matters are load temps.

Not sure if that's right or not...

 

[Slink3Slyde]

Well your CPU is better then mine if its any consolation, Under Prime 27.9 custom blend 90% RAM my Vcore goes between 1.320-1.336 depending on the test and hits 75c. Folding at home it drops it down to 1.312. I use the lowest level LLC. I can't get 4.5ghz to boot with any voltage i'm comfortable with. So you got luckier then me

Your temperatures are very good for an EVO IMO, yes variation in temperature is completely normal on Ivy even at full load (I get 10c difference hottest to coolest), and you're right idle temps dont really matter as long as theyre not pushing 60C without any load. It wouldnt hurt to reapply the TIM but I dont think you need to looking at your numbers.

 

[Aparition]

Quote:

Originally Posted by jessterb

Sharchaster,

This is what I was afraid of.
Aparition,

Yes, I'm stable at full (3.4) stock settings, if I'm understanding you correctly.
I'm also stable at 4.3 with vcore set to auto (which is typically around 1.27v).
I'm also stable at 4.5 with vcore at 1.35v, so long as I'm running Prime95 26.6 (latest version 28.5 runs me 5 degrees hotter and I end up crashing).

I just don't get why I need 1.35v to get to 4.5GHz, which is why I'm wondering if I'm doing something wrong? This guide (http://tinyurl.com/pom9z4w) mentioned that "1.25V is a good starting value that should be enough to get any i5-3570K to 4.5GHz" and they were able to get theirs to run stably at 1.21V. Forget 1.25v as a starting point in my case, I can't anywhere there at 4.5!

Should I go back to 4.3 and check what I can lower my vcore to, just to test and see?

Quote:

Originally Posted by jessterb

Ha, is this a big club? I guess it figures with me luck...

Ran a few more tests, starting with 1.25v @ 4.3Ghz and it booted fine. Here are the rest of quick tests and attempts at lowest vcores:

4.3 + CV 1.24v (1.23v) = stable (short test), hitting 67c (average 63c)
4.4 + CV 1.24v (1.23v) = crash, couldn't boot to windows
4.4 + CV 1.30v (1.28v) = stable (short test), hitting 72c (average 69c)
4.4 + CV 1.30v (1.28v) = stable (short test), hitting 72c (average 69c)
4.4 + CV 1.29v (1.272v) = stable (short test), hitting 72c (average of 68.5c)
4.4 + CV 1.28v (1.265v) = crashed in prime95
4.5 + CV 1.31v… couldn't boot into windows

I tested the 1.29v @ 4.4 in the IntelBurnTest, just a quick 5 runs at High settings and maxed out at 81c (with an average of 63c).

Anyways, I guess this means 1.29v and 4.4Ghz might simply be the best I can get and I simply didn't get a great CPU?

Or is it possible something is awry? Could this vcore issue be related to me not applying thermal paste properly? I ask that because even at full stock, my package temp is 40c with my cores varying greatly (ie. 26c-40c-17c-34c), which I thought strange and thought I should be getting better temperatures with my 212 EVO. That said, I read that:

1) it can be normal for cores to vary greatly
and
2) more or less ignore IDLE temps, what matters are load temps.

Not sure if that's right or not...

Most likely that is just the best your CPU can do, sadly. Not to worry too much about it though. 4.4 on a 3570k is no slouch. I myself only run 4.5 on my 3770k and it has no problems powering through everything I throw at it. Realistically the difference between 4.4, 4.5, and say 4.6 are still marginally small while all offer great performance over stock.

Ivy is funny with the way Intel used regular TIM between the IHS and the Die. Some chips are great and some run hotter. Delidding is an option to try to cool the chip better, most just replace the TIM between the IHS and Die. My overclocking attempts started at reaching 4.7 and the voltage kept creeping up so I settled with 4.5.
Quote:

my package temp is 40c with my cores varying greatly (ie. 26c-40c-17c-34c)

This looks like the TIM between the IHS and Die is not great. You could try re-seating the Cooler and apply a bit more pressure if you can. It could be an air bubble from your TIM job but more likely it is one of the cores not in great contact with the IHS.

But again 4.4 on a 3570k is good. I'm not even sure a better CPU cooler would benefit you either as it is high voltage that is the issue here and not temperature itself. If you were running 1.26v and in the 80'c then a better seat or cooler would probably be good, but since you are already at ~1.3v it sounds like you are already at your overclocking limit without additional modification.

 

[neuroxia]

You're always going to have one core running at a higher temperature because of the uneven load, but those differences seem way too much. As it has been told, try re-seating your cpu cooler and use a very little amount of thermal compound, people always tend to use more than necessary.

About the voltage, even if it's within the recommended limits, i think for 4.5ghz it seems really high as i'm running my 3570k @ 4.7ghz with 2.24 peak, 2.08/2.018 after vdrop and i think i could get it even lower if i fiddle around with the settings and i do more testing (but i don't have much time lately). Have you disabled C3 and C6 states? I have looked at your bios screens but seen nothing regarding this... i may be blind though as i'm at work right now and in a hurry)

Is there really such a big difference between different batches of the same generation?

 

[Jpope]

What the others above have said, Been running it at these exact overclocks since the day this chip was released.