Help with overclocking a 4790k?

[tolis626]

So I bought my PC a couple of weeks ago, and I may have a dud CPU on my hands, but I'm not ready to give it up just yet. So I thought I'd ask here, in case someone with more knowledge than me has any tips of ideas. Specs are what you see in my sig. I don't have a GPU yet because I have exams coming up and I don't want gaming to screw my studying even more.

Long story short, I've been trying to overclock my CPU with no huge amount of success. Until now I've got it stable at 4.6GHz at 1.275V (Maybe it could go lower, I just didn't try), then 4.65GHz at about 1.3V (Which did overheat to 100C during Prime95 Small FFT, but was ok during OCCT) and 4.7GHz needed about 1.31V to be stable. 4.8GHz hasn't been stable at up to 1.33V, not even for over 20s of OCCT. Over 4.8GHz is obviously out of the question. I also tried with the multiplier at 47 and BCLK at 102.2 to get 4.8GHz, but still the same results.

Before someone mentions it, RAM isn't the culprit here. 2200MHz CL10 is 100% stable at 1.65V (Damn thing doesn't want to do 2400MHz CL10. Unless there's something I'm missing here too, in which case suggest away), but I also tried with RAM set to 1600MHz CL9 and the stock XMP profile of 2133MHz CL11, both at 1.5V.

I also haven't messed with the cache at all.

So you'll say "Accept you have a dud and leave it at 4.6GHz. It's more than fast enough anyway", and you'd be right. Thing is, this is my first high-end rig after a long time, and I was hoping for more, like say a 24/7 usable 4.8GHz. Excitement has got the best of me it seems. In the (I hope not too distant) future I will be doing a custom water loop, so I will be able to push clocks and volts further, not to mention that lower temps might help with stability. For now, 1.3V, maybe even 1.31V is the max I'm willing to throw at it with the H110 and its crappy fans. I'll upgrade the fans to Phanteks' fantastic PH-F140SP fans (Partly for lower noise, but also their static pressure will help with dealing with the dust filters' high restriction), and maybe I will do push-pull on the thing (Maybe I'll even use a pair of PH-F140SPs and the fans Corsair included), but things won't change much I suspect.

Any tips are welcome, obviously. Thanks in advance guys!

PS : Could the fact that I'm using the iGPU affect the CPU's stability at high clocks? I don't think so, but I'm throwing this out there.

 

[tolis626]

I saved my signature after I finished my post, so it doesn't show in the OP. Sorry about that!

 

[Burpo]

It's been stated that the 5th Gen CPU's overclock better with adaptive voltage rather than fixed voltage. Are you using XMP profile? Also, Asus had their overclock settings at 103 X 47 in a youtube demo video about overclocking..

 

[Enigmoid]

4.6 or 4.7 seem to be average. 4.8 is pretty uncommon.

4.6 @ 1.275 is pretty decent.

 

[tolis626]

It's been stated that the 5th Gen CPU's overclock better with adaptive voltage rather than fixed voltage. Are you using XMP profile? Also, Asus had their overclock settings at 103 X 47 in a youtube demo video about overclocking..

Yeah, that's because adaptive goes well over what you set it to once the CPU is heavily loaded. For example, I've set mine to 1.28V adaptive for 4.65GHz and during OCCT/Prime95 it will boost the voltage to 1.3V. 1.28V isn't stable on my CPU at 4.65GHz. 1.3V is.

I'm also not using XMP. I prefer everything manual. Although I did try XMP to see if I messed anything up. It seems I didn't, but I'm full of suprises when it comes to messing up.

4.6 or 4.7 seem to be average. 4.8 is pretty uncommon.

4.6 @ 1.275 is pretty decent.

I think 4.8GHz is pretty uncommon under 1.3V. At 1.33V, I think most could do it, but wouldn't be comfortable with it. I could be wrong, but that's the idea I've been given.

 

[Freddy1765]

I don't understand why you think your chip is a dud. Getting a 4.6 GHz OC at less than 1.3v is pretty decent. Just because the 4790k is set to 4 GHz out of the box doesn't mean it achieves the same clock increase as a 4770k would from a much lower outset.

 

[Enigmoid]

I think 4.8GHz is pretty uncommon under 1.3V. At 1.33V, I think most could do it, but wouldn't be comfortable with it. I could be wrong, but that's the idea I've been given.

Looking at the reviews (AT, toms, techreport, hardocp, pcperspective, etc.) 4.8 ghz is pretty uncommon.

I think your chip is average.

 

[boondocks]

My 4790K does 4.7 @ 1.272 but it takes a jump to 4.8 @1.380.
But I would say your cpu is about average...just like with the 4770K I've seen the DC cpu's all over the place.
You might be able to bump the cache voltage and squeak a little more out.

 

[tolis626]

I don't understand why you think your chip is a dud. Getting a 4.6 GHz OC at less than 1.3v is pretty decent. Just because the 4790k is set to 4 GHz out of the box doesn't mean it achieves the same clock increase as a 4770k would from a much lower outset.

Anticipation is the mother of disappointment, eh? :|

Looking at the reviews (AT, toms, techreport, hardocp, pcperspective, etc.) 4.8 ghz is pretty uncommon.

I think your chip is average.

Hmm...Maybe. It's more the unachievable 4.8GHz that bugs me, that's true, but whatever. I'll get used to it maybe.

Still, I do think that there's more to overclocking the CPU than, you know, bump the vcore and multiplier and that's it. Maybe I won't achieve higher clocks, but I'd also accept a drop in voltage for the same clocks I have now. I'd love 4.7GHz stable at 1.3V at least. Or maybe I want too much. I dunno.

My 4790K does 4.7 @ 1.272 but it takes a jump to 4.8 @1.380.
But I would say your cpu is about average...just like with the 4770K I've seen the DC cpu's all over the place.
You might be able to bump the cache voltage and squeak a little more out.

Holy crap. That's actually a lot of volts. What do you use to cool it? (EDIT : I just noticed you're on custom water, so I guess it's not a problem, just not an option for 24/7 use)

When you say bump the cache voltage... How much should I set it to? I don't even know at what voltage it's running now.

Also, I guess I should also increase the cache clocks, right? 4.4 sounds good.

 

[Burpo]

Try the 4.7 and see.. 102.2 X 46 @ 1.31 = 4701
With changes to the cooling, it may do it 24/7 ..
I would use adaptive tho..

 

[Ketchup]

Unless I am missing some change since release, 4.7 is pretty much a hard limit on these., and takes some real voltage above 4.6 to do that. True, there are always some special chips out there that will go higher, but if your chip stops pretty hard at 4.7, I wouldn't waste time and money trying to make it go higher.

My 2500k is pretty similar, with different ranges of course. It went up to 4.7 pretty easily, going up to 4.9 took a lot of volts (IMO), and nothing I was comfortable with beyond that will take it up to 5.

It has been said that these chips are binned anyway, and they possibly are, but they should be just fine for years to come at default speed.

 

[tolis626]

Try the 4.7 and see.. 102.2 X 46 @ 1.31 = 4701
With changes to the cooling, it may do it 24/7 ..
I would use adaptive tho..

It's now running stable at 4.7GHz 1.315V. And it's doing so with my memory set to 2200MHz CL9, also stable. If problems don't show up, I'm quite happy. Maybe with better cooling I can squeeze a little more.

Anyway, the thing with adaptive is it will add 0.02V during 100% load. So what I do is set my adaptive voltage to -0.02V from what I got stable with manual and bam!, stable as... You know where it goes from here.

Unless I am missing some change since release, 4.7 is pretty much a hard limit on these., and takes some real voltage above 4.6 to do that. True, there are always some special chips out there that will go higher, but if your chip stops pretty hard at 4.7, I wouldn't waste time and money trying to make it go higher.

My 2500k is pretty similar, with different ranges of course. It went up to 4.7 pretty easily, going up to 4.9 took a lot of volts (IMO), and nothing I was comfortable with beyond that will take it up to 5.

It has been said that these chips are binned anyway, and they possibly are, but they should be just fine for years to come at default speed.

Yup, seems so. My settings now are what I posted above. I'm a liiiiittle uncomfortable with 1.315V, but it's not like my PC will work at 100% load all the time and at lower loads it can easily do 4.7GHz at less than 1.3V, in this case 1.295V. So yeah, as long as temps are in check, I'm cool.

In other news, 4.8GHz wasn't stable with up to 1.36V. I'm afraid to go over that until I at least improve my cooling. It did hit 90C pretty quickly during OCCT at 1.36V and it crashed after a minute or two. Perhaps with better cooling it would be stable. 90C is no joke and I suspect it impacts stability quite a lot.

 

[Ketchup]

...90C is no joke and I suspect it impacts stability quite a lot.

Haswell throttles between 90 and 95C. Some programs do not like throttling at all, some don't care.

 

[Termie]

The only thing that was a dud here was Intel's marketing of the 4790K. I've seen no evidence at all that it can clock higher than a 4770K, and for both, 4.7GHz is above average.

The only think that Devil's Canyon has going for it is lower temps, but that doesn't necessarily mean higher overclocks.

 

[tolis626]

Haswell throttles between 90 and 95C. Some programs do not like throttling at all, some don't care.

Really? I think I've only seen mine throttle when it hit 100C, but I could be wrong. Hmmm... I'm probably wrong. Anyway, 4.8GHz hasn't happened and probably won't, so I'm not losing any (more) sleep over it.

The only thing that was a dud here was Intel's marketing of the 4790K. I've seen no evidence at all that it can clock higher than a 4770K, and for both, 4.7GHz is above average.

The only think that Devil's Canyon has going for it is lower temps, but that doesn't necessarily mean higher overclocks.

Yeah, I can agree with this. They were trying to lead us to believe that magic happened. Well, it didn't. Some binning did, but it seems that they just got rid of the really bad Haswells that didn't even make it past 4.2-4.3GHz.

 

[buklau]

Guess my 4690k 4.6ghz at 1.3v is below average.

 

[Ed1]

Guess my 4690k 4.6ghz at 1.3v is below average.

you delidded a 4690K ?
curious as how much it lowered temp wise .

 

[buklau]

you delidded a 4690K ?
curious as how much it lowered temp wise .

Roughly about 5C lower on prime95. It will be lower if I use CLU, but currently just MX2.

 

[poohbear]

Hows 4.6 on a haswell a dud??? Thats awesome! My Sandy bridge 2500k only does 4.5 maximum @ 1.3v! 4.7+ was supposed to be "common" w/ sandy bridges, sometimes our expectations are inflated.

 

[XiandreX]

Hows 4.6 on a haswell a dud??? Thats awesome! My Sandy bridge 2500k only does 4.5 maximum @ 1.3v! 4.7+ was supposed to be "common" w/ sandy bridges, sometimes our expectations are inflated.

You said exactly what I was thinking. "Sometimes our expectations are inflated"

 

[StrangerGuy]

4.6GHz at 1.3V? I would rather settle 4.2GHz stock @ 1.05V, why take a huge volt increase and all the associated negatives for an insignificant <10% clock gain? This is already miles better than many non-DC K chips that will never break 4.2GHz regardless of voltage or cooling.

I never understood the appeal of OCing Haswell myself; for a 4670K that 4-core turboes to 3.6GHz a realistic overclock to 4.1GHz only nets a 13% clock gain. Fine if you want the features of a Z board, but if not you are really better off spending the money elsewhere by not OCing.

 

[Ketchup]

I don't see the appeal of overclocking these either, save that first "see what it will do" test.

I attribute it to aging.

 

[tolis626]

Well, I always expect too much to be honest. By nature, it's not my fault. ()

Happy news is I managed to get my system to be fully stable with the following settings:
- Core at 4.7GHz at 1.315V (1.295V adaptive).
- Uncore at 4.4GHz at 1.25V.
- RAM at 2200MHz CL9 at 1.65V.
And all that netted me 957cb in CineBench. That ain't half bad, it seems. I so love this machine... :wub:

PS : I've stopped overclocking for just the performance gains long ago. They are rarely substantial and the benefits outweigh the problems caused by overclocking even more rarely. I overclock because I like tinkering and because I like getting 101% out of what I spent my money on. And even in cases that overclocking isn't a viable solution (Highly clocked smartphone CPUs come to mind) I will undervolt and optimize. I also sometimes do it for no obvious reason out of impulse. That's fine by me too.

With that said, I can respect people who find overclocking dumb or just not worthwhile. Much like driving fast, it's not for everyone. Some prefer the slow but steady approach, that's fine by me.

 

[GLeeM]

The iGPU will make temps a little higher.

Our expectations were unrealistic because:
1. High OC of delidded chips
2. High OC of chips that are only stable on CPU-Z validation, email and browsing without Flash

It is a very good chip that is Prime 95 v28.5 small ftts stable at 4.7 :thumbsup:

 

[tolis626]

The iGPU will make temps a little higher.

Our expectations were unrealistic because:
1. High OC of delidded chips
2. High OC of chips that are only stable on CPU-Z validation, email and browsing without Flash

It is a very good chip that is Prime 95 v28.5 small ftts stable at 4.7 :thumbsup:

Turns out to be so. I have to say, I was very disappointed at first. But things may just be like what you said. I'm not at all comfortable with delidding though, as I'm quite clumsy. Shame...

In other news, Prime95 with small FFTs is a pig to run. My system ran it ok for a little over 20 minutes straight with no problem. BUT at that point, my cooling got overwhelmed (at least that's my diagnosis of the matter) and the CPU started hitting 90+C (highest was 98C) even with the air conditioner throwing cold air at the PC and the restrictive dust filters removed. I'd blame the overclock, as 4.7GHz and 1.315V isn't easy on the CPU, but I had the same happen at a more sane 4.6GHz at 1.275V, albeit a little later during testing, at about half an hour. Damn it. At least it's stable at Prime95, no matter what.