New guy, some advice wanted on 4790K overclocking

Hi, I'm running into some issues (or so I think) with my 4790K.

I've been using my system for FEM analyzes for quite some time and encountered heat-related shutdowns (checked in CPU-Z, it did go 100 degrees Celsius on some occasions). Added a Noctua NH-U14S, which helped to some degree, but still low to high 90's, not desirable...

Now that I no longer need to run FEM analyzes I just want my system to run stable and to start overclocking it a bit maybe
I have replaced my MB with an ASRock Extreme4 and have just recently started it up to see it's stability.
I have not yet tried any oc, just wanted to see what it does in standard form; when running Prime95 it now slowly creeps towards 80 degrees Celsius.

I don't really know if that is acceptable, but I'm just a little bit afraid I might wreck it..

I've been reading up on overclocking (found a bunch of guides, so that will not be an issue I guess) and delidding.
Now on delidding I can't find definitive awnsers or guides on whether or not it really does something for the Devil's Canyon (some say it shaves of 15 degrees, some say it doesn't do sh*t). So, does anyone know how to delid a 4790K and if it is worth the effort?

Also, I've seen some people lapping the IHS and cooler (sanding and polishing essentially). But again, contradicting opinions everywhere. So, is lapping the IHS and my cooler going to help me achieve a more stable oc?


Tl;dr
I want to oc my 4790K:
I'm hitting 80 degrees (Celsius) with Prime95 on standard clock, is this normal/good/bad?
How do I delid a 4790K, is it worth it?
Is lapping my IHS&cooler worth it?

It does seem that you may have some heat problem. 80c is acceptable and P95 is about the hottest workload you can apply.
I have a question about your case set up and fans. Which case how many fans and how are they oriented. A quick test to see if you need better airflow through the case is to run your test with the side panel off. If it runs noticably cooler then you need to work on case flow

I'm using an Antec Two Hundred V2 case with two Noctua Industrial fans (120 and 140 mm). 120mm at the rear, 140mm at the top.
Will run the test with the side panel removed later today

It would also be handy if you can put your PC specs in a sig to follow your posts. I was curious what you vid card is

Jeroen64 said:

Hi, I'm running into some issues (or so I think) with my 4790K.

I've been using my system for FEM analyzes for quite some time and encountered heat-related shutdowns (checked in CPU-Z, it did go 100 degrees Celsius on some occasions). Added a Noctua NH-U14S, which helped to some degree, but still low to high 90's, not desirable...

Now that I no longer need to run FEM analyzes I just want my system to run stable and to start overclocking it a bit maybe
I have replaced my MB with an ASRock Extreme4 and have just recently started it up to see it's stability.
I have not yet tried any oc, just wanted to see what it does in standard form; when running Prime95 it now slowly creeps towards 80 degrees Celsius.

I don't really know if that is acceptable, but I'm just a little bit afraid I might wreck it..

I've been reading up on overclocking (found a bunch of guides, so that will not be an issue I guess) and delidding.
Now on delidding I can't find definitive awnsers or guides on whether or not it really does something for the Devil's Canyon (some say it shaves of 15 degrees, some say it doesn't do sh*t). So, does anyone know how to delid a 4790K and if it is worth the effort?

Also, I've seen some people lapping the IHS and cooler (sanding and polishing essentially). But again, contradicting opinions everywhere. So, is lapping the IHS and my cooler going to help me achieve a more stable oc?


Tl;dr
I want to oc my 4790K:
I'm hitting 80 degrees (Celsius) with Prime95 on standard clock, is this normal/good/bad?
How do I delid a 4790K, is it worth it?
Is lapping my IHS&cooler worth it?

Click to expand...

I think I can help you with some of these questions on the Devil's Canyon as I have had some experience with delidding and lapping them. The reason you are getting data all over the map about the effectiveness of delidding these CPUs is that the TIM application job from the factory varies in how well it is done and also because different people use different replacment TIM products. This means that some delidders will get dramatic temp improvements and some will not. You can expect somewhere between about 6c and 16c improvment if you use Liquid Ultra. If you use conventional TIM you will see more like 4-6c improvement. I strongly recommend using Liquid Ultra under the lid for the best results but you must be careful in the application process as it is conductive. I have some recommendations for how to make the application of LU easier if you get to that point. Use Liquid Ultra under the lid but conventional paste on the top side between the CPU and heatsink. If you use Liquid Ultra on the topside you will knock of a couple more degrees but it creates a dangerous scenario if you ever remove the heatsink because it tends to flake off and drop onto the motherboard traces upon separation.

As far as lapping, don't waste your time. I've done it more than once and it makes little or no difference

Thanks for the reply, definitely worth it if I can shave of 16C.

trents said:

As far as lapping, don't waste your time. I've done it more than once and it makes little or no difference

Click to expand...

I've been reading on and found a post of someone who delidded a 3770K after lapping, indeed, didn't improve anything.

Johan45 said:

It would also be handy if you can put your PC specs in a sig to follow your posts. I was curious what you vid card is

Click to expand...

Indeed handy, added it
I'm running a GTX 1050Ti, budget GPU next to a i7 , still a massive improvement over my old 550Ti

Okay, update time.

Ran Prime95 again, removed one side panel and ran it for 15 minutes (not long enough/too long?) And it reached a max of 81.
Then I heard the fans kick up a couple of rpm's, so shut it down and changed the fan behavior in the UEFI/BIOS. Ran Prime95 for another 10 minutes (was running out of time, people were trying to sleep ) still 80 degrees, but consistently lower (80% of the time under 78).
Attached two images of the results in CPU-Z and RealTemp.


Now for the next "issue", I've been reading the Overclockers.net 3 Step guide to overclock Intel Haswell. (http://www.overclockers.com/3step-guide-to-overclock-intel-haswell/) And more specifically, I'm interested in the following bit:

Now, I’ll share a secret imparted by the folks at ASUS who gave several reviewers some tips on overclocking the retail stepping Haswell chips: Set Vcore to 1.20 V. Set all cores to 46x (which would be a 4.6 GHz overclock), save & reboot. If the system boots past the UEFI and either begins to load or, ideally, makes it into the OS and is stable, you have a 50th percentile or greater chip on the Haswell overclocking-ability bell curve. If it won’t at least boot there and make it into the UEFI, you probably have less than a 50th percentile chip. You can expect chips in the lower 50th percentile to top out in the 4.4-4.5 GHz range at 1.25 V.

Click to expand...

Now I couldn't find a way to lock my Vcore (again, was running out of time), so I just figured I'd set my multiplier to 46 and see what Vcore comes from it.
So, booted it and it went straight to Windows. No issues whatsoever. I opened CPU-Z and found that it was pulling a "Core Voltage" of 0.952V (now that should be Vcore right?).


So I'm a bit confused now, is it easier for a cpu to reach a certain clock on lower voltages or am I reading the wrong "Vcore"?

Also reviewed the core voltages on the Prime95 runs. 0.86ish, also really low, idk I'm really confused right now


UPDATE:
Found another thread on a 4790K and someone posted a pic of their UEFI/BIOS (also an ASRock Extreme). So I now know how to lock/override Vcore, will try that later today
Still confused about the voltages though

No that V_Core reading is not correct. Do you have Windows power plan set to balanced? If so that's why you see such a low voltage.

Put a load on the cpu and watch the voltage jump up.

I have now put Windows Power Plan on performance.
I have set All Core on 46 and Vcore on override at 1.20V. And CPU-Z still claims it's pulling 0.952V at idle.
Running Prime95 kicks it up to 0.96something, with temperatures quickly going to 95.

I can't seem to get the UEFI screenshot to work (blank notification on screenshot info with external HDD plugged in)...

I'm still really confused

I will look into other stress tests like AIDA 64.
Any recommendations on other stress tests and how to get the voltages to correlate with what I set up in the UEFI?

UPDATE:
Not ideal, but I grabbed some pics with my phone.

To capture a bios screenshot try a usb flash drive instead of an external hard drive. Usually it's the F12 key that takes the picture.

Figured that much, going after an USB drive today (don't have one anymore )

Will list my UEFI settings as well later today (maybe tomorrow).

I'm planning on keeping the clockspeed at stock for the time being, and start upping the Vcore until I'm comfortable with the temps I'm getting when running Prime95 (<92 after 15 mins). After that I'll up the speed until it becomes unstable, then tune it down untill it becomes stable again (if anyone thinks this is a bad idea, please tell me )
If I'm not pleased with the results (I've heard tales of Devil's Canyons reaching 5Ghz on air cooling), then I'll start (to consider) de-lidding

I am having the feeling that the airflow is not optimal, so I might look for another case and some additional fans. Short Prime 95 run on 4.6Ghz @1.2V (set in UEFI, not read in CPU-Z) already saw temps shooting up to 95 within minutes

Depending on what version on p95 I find it makes Haswell to hot(unrealistic) I like aida64, realbench,and occt.
If you really are at 4.6 @1.2v then your chip is. Much better than mine. What are you using for a hsf.
For my chip 4.2. @1.2 v on stock hsf
4.5. @ 1.3v
4.7 @ 1.4
4.8 @ 1.46 ish
4.9 @ 1.5 (bench stable)
I have never hit 5 GHz the best I have done is 49*101

please download hardware monitor and leave it open during a 20 minute p95 run so we can see peak temps and voltages.
I run one of my 4790K neked and the other is delided and retimed.worth it, yea and no, depends on the chip and you.
if he's at [email protected] I might want his chip and if he can make 4.8 at around 1.30 or less there is no reason to delid.

I might already have installed Hardware Monitor...I'll look it up later today and do as you requested on some different settings

I'll also look into those other stress tests. My HSF (heat sink fan, right?) is the standard Noctua NF-A15 140mm PWM fan that came with my Noctua NH-U14S cpu cooler.

In my UEFI I set CPU Vcore Voltage Mode to "Override Mode" and put Vcore on 1.200V, yet when I check with CPU-Z it says it's just 0.952V and spikes at 0.964V... so yeah I'm not that sure about those values

As said, I'll look into Hardware Monitor and run P95 for 20mins, results to follow...

I've been doing a lot of experimenting lately with stress testinbg and have come to the conclusion that passing 4 hr. of the Realbench stress test gives me reasonable assurance that my overclock is stable as does passing 3 hr. of OCCT large data set. If you are using Prime95 as the stress tester, get one of the older versions like 27.7 that has AVX1 but not AVX2 to keep the heat down and run it for two hours.

If your cooling won't handle the intensity of any of the above stress testers, then use the Intel Extreme Tuning Utility (XTU) stress test over night and then add .02 to the minimum voltage it takes to pass that. Or use AIDA64 Extreme overnight. These two are gentler on temps but don't give the same level of assurance with regard to stability of an overclock so I recommend supplementing the voltage after finding the minimum needed to pass them.

Realbench, OCCT and XTU all have the advantage of being able to set the length of the test and also have built in temp monitors. AIDA has a temp monitor but no timer. Prime95 has neither.

Very usefull information, thank you!

Will look into it over the weekend if I can find some spare time (moving, takes more prep than expected)

I've also read that an even older version of Prime95 (26.5) can be used?

For now I have plenty of information to work through. I'll check back with you guys when I've been able to do some testing, thanks again

Prime95 27.7 is the first version of Prime95 to include AVX testing (AVX I). Later versions added AVX II and those are the real heaters. I would not go earlier than 27.7 as you will not get AVX testing at all and it might not accomplish the purpose very well. All Intel CPUs since Sandybridge in 2012 possess AVX instructions sets. If your cooling won't handle that then go another less intense product like XTU or AIDA64. Although AVX is not used in common applications I would not dispense with engaging it altogether in my stress testing. OCCT has a Linpack mode with an AVX option as well but it heats things up pretty good, similar to the InelBurnTest which is older stress testing product that uses the Linpack library.