[Help] Overclocking a 4790k

[AlfaPrime]

Hey guys, I needed a little help here with my overclock

PC Specs:

CPU: i7 4790k
Cooler: Corsair H115i
Motherboard: Asus Maximus VII Hero
GPU: 2X 970 Gigabyte SLI
RAM: 4x4 16gb 2400mhz Gskill
Housing: NZXT H440
Source: RM750i
SSD: Samsung 850 EVO
HDD: WD 500gb
External Drive: 3TB Seagate Expansion

So the problem is that I think my H115i makes temperatures too high with my 4790k i7, I think I have caught a bad chip. Why? Well now comes a few questions because I'm not really very understood in this but I need to realize

At the moment I am stable with 1.32vcore (high I know) @ 4.6ghz and temperatures easily reach 85ºc to play The Division (pulls almost 100% by the processor)

Another thing I found strange. Initially I started using the stock cooler that came with the i7 and decided to run the Aida64 to see the temps and noticed that the stock did not arrive to cool the i7, ie the processor easily reached 90ºc. Even after resetting the motherboard factory settings several times, I noticed that the vcore was in the 1,299 stock.

Is this the actual stock value? I think something high, but I still can not confirm this information

Ok, I'm editing this initial topic with the information you ask and tests that you want me to run.

I must remember that the Aida64 is out of question because it conflicts with the Corsair Link and my curve of the cooler stops working and the fans do not increase according to the temperature, we have to use another test.

I then await your comments to better understand what the problem is.
Greetings and thanks

 

[FreedomEclipse]

Thats normal with that CPU as intel use crappy thermal paste under the heat shield of that CPU. For best results you need to de-lid that cpu

A guide can be found here

 

[AlfaPrime]

Thats normal with that CPU as intel use crappy thermal paste under the heat shield of that CPU. For best results you need to de-lid that cpu

A guide can be found here

lol, a really dont feel confortable delidding becuse I have never done it before and I dont have money atm to buy another processor after I ruin this one lool

 

[GreiverBlade]

Thats normal with that CPU as intel use crappy thermal paste under the heat shield of that CPU. For best results you need to de-lid that cpu

A guide can be found here

wasn't the Haswell-Devil's Canyon using a better TIM than Haswell? my 4690K never needed a de-lid to get good temps (around 30°C idle 60° average 71° max load )

At the moment I am stable with 1.32vcore (high I know) @ 4.6ghz and temperatures easily reach 85ºc to play The Division (pulls almost 100% by the processor)

is it really necessary .... my 4690K or my 6600K never needed more than 4.4 @ 1.248V (woops missread my reading) to run any games (inclusive the division) thought i ran my 4690K at 4.5 rather than 4.4

the H115i is a fine cooler in balanced mode, rarely see more than 50°C max on my 6600K but ... the voltage regulator was moved out of the CPU unlike the 4XXX series (which explain my idle of 24°C at 4.4 turbo off )

 

[FreedomEclipse]

lol, a really dont feel confortable delidding becuse I have never done it before and I dont have money atm to buy another processor after I ruin this one lool

then lower your voltages and cut back on your OC or just dont OC at all. Unfortunately for you, You have purchased the top tier 1150 i7 so there is nothing to upgrade to, Only thing you can really do is go backwards and go for a lower tier i7 or i5. Failing that, completely switch the board, CPU and ram out for a new system, but you cant afford that.

Pretty much just what i said. Cut back on your OC and just deal with it

wasn't the Haswell-Devil's Canyon using a better TIM than Haswell? my 4690K never needed a de-lid to get good temps (around 30°C idle 60° average 71° max load )

This is true. It did come with better paste but it still wont hurt to delid. Also Check voltages.

 

[GreiverBlade]

This is true. It did come with better paste but it still wont hurt to delid. Also Check voltages.

well the only issue i see here, is indeed to much V and a pointless OC on a CPU that can still handle anything at a slightly lower OC or even at stock

also ... edit

 

[FreedomEclipse]

well the only issue i see here, is indeed to much V and a pointless OC on a CPU that can still handle anything at a slightly lower OC or even at stock

also ... edit

After googling some CPU-Z shots of the 4790 at 4.6Ghz, Voltages seem to be all over the place. One screenie had 1.37v while others varied between that and 1.20v and 1.13v even (i think) so id say the voltage is within margin of error if it really is 1.32v and not anything higher.

 

[Folterknecht]

wasn't the Haswell-Devil's Canyon using a better TIM than Haswell? my 4690K never needed a de-lid to get good temps (around 30°C idle 60° average 71° max load )

There was a very big variance, some had pretty good thermal performance and others just needed delidding to make them work.

 

[GreiverBlade]

After googling some CPU-Z shots of the 4790 at 4.6Ghz, Voltages seem to be all over the place. One screenie had 1.37v while others varied between that and 1.20v and 1.13v even (i think) so id say the voltage is within margin of error if it really is 1.32v and not anything higher.

well yep 1.32 if it's the max then it's withing margin of error, you're right... the OP also miss something ... 85°C is still acceptable as a max ... but for a h115i i find that a tad high .... unless being "silence obsessed" and running it in "quiet" profile

There was a very big variance, some had pretty good thermal performance and others just needed delidding to make them work.

i never did read that about Haswell-DC ... about Haswell classic yes... but i did read the opposite about DC ... weird ....

 

[AlfaPrime]

well yep 1.32 if it's the max then it's withing margin of error, you're right... the OP also miss something ... 85°C is still acceptable as a max ... but for a h115i i find that a tad high .... unless being "silence obsessed" and running it in "quiet" profile


i never did read that about Haswell-DC ... about Haswell classic yes... but i did read the opposite about DC ... weird ....

guys, do you want me to run some tests to have a better depth of the problem? Just tell me what programs to run and ill run it and post the results

 

[FreedomEclipse]

Prime95 while running and having Real temp and CPU-Z up so we can see your temps and voltages..

Start it all up. then run prime95 in a blend test for 10mins then take a screenshot and share it with us

 

[FYFI13]

At first my 4790k was acting like that and i had to reseat Corsair H60 water block several times. Now i have 53-56C while gaming and up to 65C on stress tests. Your H110i should do a lot better than that. Make sure there are no gaps between CPU and cooler.

 

[GreiverBlade]

At first my 4790k was acting like that and i had to reseat Corsair H60 water block several times. Now i have 53-56C while gaming and up to 65C on stress tests. Your H110i should do a lot better than that. Make sure there are no gaps between CPU and cooler.

ahhhh thanks ... i forgot about re seating issue ... indeed a H115i should be better than 85° ...

well i already wrote it ... thought i forgot to ask if he did re-seat his block.

85°C is still acceptable as a max ... but for a h115i i find that a tad high .... unless being "silence obsessed" and running it in "quiet" profile

though i don't know if the OP is running in quiet mode xD (as i run mine in balanced most of the time or full throttle when i feel masochistic.)

 

[Folterknecht]

i never did read that about Haswell-DC ... about Haswell classic yes... but i did read the opposite about DC ... weird ....

Intel claimed they had that improved with Devils Canyon, but in reality you still had widely spread results when delidding. Some barely "gained" 10 K ofters improved by 20 K. It's just a mess since Ivybridge regardless of generation and they are charging a premium for these K-CPUs.

 

[Kursah]

Both my 4790K's ran at better temps stock than my 4770K's did. That said, I just delidded my "good" 4790K which was able to achieve 4.8GHz on air cooling at 85C or cooler under OCCT loads (meaning 75C or cooler for gaming, usually below that even).

I just performed the delid on my "good" 4790K, and now I'm seeing OCCT loads closer to 65-68C. I used CLU between the die and IHS, a thin layer of permatex RTV to glue the IHS back down and Noctua NT-H1 between the IHS and CPU cooler.

Now my OC hasn't changed yet, I haven't really tried...but my temps went down 10-15C depending on task. That is pretty huge for me, especially because this chip seemed better than average in the slew of 4790K's and I was pretty hesitant to delid it. My bad chip, which is my server chip...runs hot (80C @ stock speeds under 212EVO) and will receive the process next weekend hopefully. I am expecting a more drastic change here.

The H115i should be pretty potent but depending on a few things like airflow, it could be right in-line. I do run a little warmer than I could simply because I run all the fans at the low setting in my 600C, though I can increase speed with the flick of a switch if needed...this time of year it isn't necessary (-8F outside, 65F inside where the PC is).

YMMV but Devils Canyon still can greatly benefit from the delid process. Especially if you have more than sufficient cooling and still are reaching higher temps at a lower OC. That said 1.32v for 4.6 seems pretty aggressive to me...that's a good bit of extra voltage for what is 200MHz past peak stock turbo profiles. That said if you can keep the chip cooler you might be able to drop the voltage a little bit.

 

[AlfaPrime]

Around the 4min mark it started hitting 100, till then it was really nice, around 78-84ºc

 

[FreedomEclipse]

1.87vCore on corsair link

 

[GreiverBlade]

1.87vCore on corsair link

no joke it's overheating if that's true ... actually Corsair LINK is quite accurate when it come to reading ... it always report result closer to HW64Info than CPU-Z does ... and if 2 software on 3 report the same result while the third is slightly above or bellow, then the 3rd probably not accurate enough.

OP lower your OC .. you don't need 4.6 ... even at stock a 4790K still blast high (or lower it to 4.4 and check Vcore behavior)

 

[Kursah]

1.87vCore on corsair link

That's the FIVR voltage... the Haswell CPU's had the integrated CPU voltage regulator. I have mine adjusted down to 1.65v with LLC on Auto (runs at 8, overvolts to 1.680v under heavy loads). Mine runs 1.8v stock iirc...and many do.

Frankly you should be able to hit 4.6 pretty easily, getting beyond that is the challenge. At least in my experience, but maybe your chip is just a bad clocker and hot runner... I'd delid it with those kinds of results.

Is that the version of Prime that runs AVX? I don't and haven't used Prime95 in a long time so I'm kinda out of the loop with it.

If you are using AVX instructions, that's why you're getting so hot... AVX instructions in tests will cause Haswell and newer CPU's to overvolt... mine goes from 1.26v to 1.31-1.32v. I DO NOT recommend stress testing with AVX tests...

I also prefer OCCT for testing, use CPU:Linpack, default settings except for using all logical processors on systems with HyperThreading, (default meaning DO NOT CHECK the AVX option), run on Automatic for a 1 hour test. It's free, try it and let us know how those results are too.

 

[GreiverBlade]

That's the FIVR voltage... the Haswell CPU's had the integrated CPU voltage regulator. I have mine adjusted down to 1.65v with LLC on Auto (runs at 8, overvolts to 1.680v under heavy loads). Mine runs 1.8v stock iirc...and many do.

Frankly you should be able to hit 4.6 pretty easily, getting beyond that is the challenge. At least in my experience, but maybe your chip is just a bad clocker and hot runner... I'd delid it with those kinds of results.

Is that the version of Prime that runs AVX? I don't and haven't used Prime95 in a long time so I'm kinda out of the loop with it.

If you are using AVX instructions, that's why you're getting so hot... AVX instructions in tests will cause Haswell and newer CPU's to overvolt... mine goes from 1.26v to 1.31-1.32v. I DO NOT recommend stress testing with AVX tests...

I also prefer OCCT for testing, use CPU:Linpack, default settings except for using all logical processors on systems with HyperThreading, (default meaning DO NOT CHECK the AVX option), run on Automatic for a 1 hour test. It's free, try it and let us know how those results are too.

he manage to get 4.6 easily ... that's just the temps who are off ... not the OC (well the V are still wonky ... )

 

[Kursah]

he manage to get 4.6 easily ... that's just the temps who are off ... not the OC (well the V are still wonky ... )

Shouldn't need 1.32v for 4.6Ghz IMHO...between a possible bad die-to-IHS contact, possible bad IHS-to-cooler contact, too much voltage for the OC (hopefully), and too much FIVR voltage (hopefully) he's running too damn hot for what amounts to nothing more than negligible differences in performance that won't be noticed outside of benchmarks.

When I say get to 4.6 easily, I mean without this much work and heat.

Though I say that, and I bet my crappy 4790K (in my server), might run the same course as the OP's chip...though I don't even think it was that bad honestly...but now I'm interested in finding out! I'll surely have to try it out down the road...when I get real server hardware and can reclaim my other 4790K/Z87 combo! That chip tho...just last night, I ran an OCCT test as I directed the OP above... 26C idle, 82C load between coolest and hottest cores under a 212EVO stock fan setup, running stock, no undervoltage ATM, all defaults, just sync'd turbo so all cores run at 4.4 under turbo load. Might as well take advantage of the free OC.

 

[AlfaPrime]

Guys, a friend of mine told me to try overcloking with the rams at default 1600mhz instead of the 2400mhz, what do you guys think?

 

[Kursah]

Yep, OC the CPU first. Find out what it can do... then add RAM. The memory controller is built-into the CPU core...so going over 1866-2133 on many 4790K's can be stressful, causing increased voltages needed.

BUT with that said, you have a heat issue here...and that really needs resolved before anything. Did you try OCCT without AVX? I'm curious of your results.

 

[AlfaPrime]

Yep, OC the CPU first. Find out what it can do... then add RAM. The memory controller is built-into the CPU core...so going over 1866-2133 on many 4790K's can be stressful, causing increased voltages needed.

BUT with that said, you have a heat issue here...and that really needs resolved before anything. Did you try OCCT without AVX? I'm curious of your results.

Will post the results in 1h, helping a friend with an issue. Will 10 mins of OCCT suffice? To start of course, at the end ill do stability tests

 

[Kursah]

I would do no less than 10 minutes, but sure that should give you an idea of its test suite.

If you just want to stress the CPU, reduce the RAM usage to 10MB (the minimum allowed), the default test stresses RAM, the IMC between CPU and RAM and the CPU. It does a good job of stress testing and building heat without AVX.