Overclocking i7 4790k

Associate
Joined
24 Nov 2018
Posts
42
Hi all,

I've been playing around with my 4790k, learning the ropes of overclocking and trying to get the most performance from my rig.

I've looked through several guides and threads regarding this cpu, I think I'm in line with what others are getting, but I want to take my cpu further.

My rig and current stable OC is:

Asus z97 pro gamer
BCLK 104
System agent 1v with offset +0.23
IOD offset 0.225
IOA offset 0.22

I7 4790k @ 4.8ghz (4780) 1.275v vcore, 1.95 vcin
4.6ghz (4560) cache @ 1.26v
Cooler - arctic freezer 240 74 CFM using 4 fans in push pull config
Kingston beast 2080mhz 9-11-11-27 @1.7v

The voltage used to run at 4.8ghz seems to be on the good side of average, but my limiting factor is heat.

This current OC hits 85C absolute maximum during prime95 (266), which means it runs around 75C peaks during normal usage, which I'm happy with. However if I try to give it anything over 1.3v vcore, I start hitting temperatures in the 90s during the same prime95 (266) runs.

I'm not sure why my temperatures are so high though, I've seen some people saying they've given theirs 1.4v, and able to keep it low 70s with mid-high end air coolers, yet my cpu starts to burn up with anything over 1.31v on water. Additionally there seems to be a delta of 10C between the hottest (#1) and coolest (#4) cores during stress testing, where core 4 can be in the low 80s whilst core 1 reaches low 90s.

I've been able to 'stabilise' 4.9ghz at 1.32v iirc, and by that I mean it passed hyperpi, cinebench, and real bench, but running prime95 266 would scare the crap out of me with the temp on core 1 hitting 94C.

I want to get that mythic 5ghz people, I believe this chip has the potential, I just need to be able to cool it down somewhat, any suggestion of things I can try would be most welcome!

Thank you,

M
 
Last edited:
Thanks for the quick replies!

I think delidding is probably the only other thing I can try really isn't it... I'm up for trying it, but I am worried that I'll mess it up.

I've read a few articles on delidding, my main concern with using liquid metal is that people often state you must be precise with the amount you use, less you cause a short.

Is that mostly a precautionary statement, or is delidding really quite the challenge?

Thanks again,

M
 
For some more detail: the reason you're seeing uneven temperatures across cores is likely because the integrated heat spreader (IHS) on top of the CPU isn't event mounted from the factory. If you remove it (delid) and remount it properly you should see both improved temperatures, and better consistency core to core. Thermal Grizzly Conductonaut is a liquid metal that's generally the popular choice for this job (under the IHS - use normal thermal paste on top for the cooler). The majority of very high overclocks on Haswell i7s (and newer generations) use this method.
 
Thanks for the quick replies!

I think delidding is probably the only other thing I can try really isn't it... I'm up for trying it, but I am worried that I'll mess it up.

I've read a few articles on delidding, my main concern with using liquid metal is that people often state you must be precise with the amount you use, less you cause a short.

Is that mostly a precautionary statement, or is delidding really quite the challenge?

Thanks again,

M

It's not without risk, but if you take the right precautions and you're careful, you should be ok.

Using a delid tool such as this makes the delidding part much easier:
https://www.overclockers.co.uk/der8auer-delid-die-mate-2-hs-003-dr.html
There are other tools available and some people even use a vice or razor blade, but the latter 2 options are not normally recommended.

Once you've got the IHS off, you'll want to clear off the adhesive using something relatively soft (e.g. plastic, like a credit card). Be careful not to damage any surface mount devices such as capacitors (I don't think there any on the 4790k, but just take care anyway).

You can coat the metal contacts that are near the chip with nail varnish to protect against excess liquid metal causing a short. Gamers Nexus had an article that advised using nitrocellulose based nail vanish. They also said to avoid formulations that have benzene derivates such as toluene.
https://www.gamersnexus.net/guides/3068-how-to-delid-intel-i9-cpu-and-apply-liquid-metal

When applying the liquid metal, apply less than you think you need at first, because a little can go a surprisingly long way. You can always add a little more after spreading it out. Remember to apply the liquid metal to both the chip and the part of the IHS that'll be in contact with the chip (i.e. not the side of the IHS that contacts the CPU cooler). Don't worry if you added too much liquid metal, as you can suck it back up into the original Conductonaut syringe. Isopropyl alcohol also seems to work well if you need to wipe and remove the liquid metal. Definitely watch some videos of people applying liquid metal to see the do's and don'ts. Some big don'ts are never apply liquid metal to aluminium and never use liquid metal if you're doing sub-zero (e.g. liquid nitrogen) cooling.

Lastly you may want to reseal the IHS, but this is optional.

Good luck!
 
Thanks again for the quick replies and detailed posts!

It's definitely less intimidating the more I watch videos on the process, I saw somewhere that the 4790k has a line of components along the edges of the die, which can be risky if there's too much LM, so I guess that's where the varnish would come in.

If I got the de-lid kit, the LM, and a fresh batch of thermal paste for contact with cooler, I'm looking at around £45, with the de-lid kit being the biggest expense here, how difficult is a de-lid without the kit?
P.s, I'm not looking to ignore advice here, just trying to save cost really - if the other methods are too risky then I'm basically just gambling the small £30 outlay for the potential cost of a new CPU...

I do find it odd that de-lidding has such an impact, you would have thought Intel would have adopted the use of LM or some other material and learned from the fact that this process works so well. It boggles my mind that they sell even their extreme range CPUs without maximising the potential for LM to reduce temps further... Do AMD chips see the same/similar benefit from de-lidding?

M
 
In terms of DIY: I've done a chip myself in a vice. It went ok but I have a lot of metalwork and woodwork experience (and a good vice). I also bought a cheap delid tool from China for about £8. Worked ok, just had to wait longer for postage!

So it comes down to your confidence level. Some people also borrow, or buy then sell on their delid tool.

Intel used to solder the IHS to their CPUs until 3rd generation Core i series (Ivy Bridge). LM is too risky to be shipping and producing en masse. Whereas the normal TIM and silicone glue method is repeatable and cheap compared to LM and solder. The thermal performance of chips is good enough that Intel stuck with it. Remember that a delid is mainly to allow better overclocks, which intel isn't responsible for guaranteeing performance of... :)
 
Very good points!
I guess for run-of-the-mill CPUs it's redundant anyway, but you'd have thought that for their top of the line XE models you'd get some kind of upgraded TIM or some such, although as you say I suppose that's left for should anyone want to push their chips even further.

In terms of confidence levels, I'm comfortable putting together PCs and working inside a case etc. but I don't have any experience in modifying chips/components. The closest I've gotten to that is installing the CPU AIO, which is to say not close at all to a de-lid :D

Unfortunately, I'm not a dab hand at DIY either so I don't have any vices or other tools that might help me there, so I suppose I should spring for a kit really, otherwise, it'd be a razor blade and that sure doesn't sound too safe!

Also, I've seen quite a range of recommended operating temperatures for CPUs in general, and specifically hot chips such as the 4790k - with my current OC peaking at 85C during prime, my usual operating temperature is peaking at around 75C, is that still too high?
Cooler CPUs are always preferable ofc, and I would love to hit 5Ghz, but it'd be nice to know how this current 4.8Ghz OC holds up if I end up chickening out...

M
 
For your question on whether AMD benefits from delidding, most Ryzen 1000 and 2000 series CPUs are soldered and therefore are extremely difficult (or next to impossible) to delid. However, the Ryzen 2400G and 2200G APUs use TIM and do benefit slightly from a delid (but only really recommended for people using small coolers and want to reduce fan noise).

For max operating temps, I personally don't like my CPUs to exceed 75C max and try to keep them under 70C for most of the time.
 
I had to delid my old 4790k because the temps were crazy. I borrowed a Rockit delid tool from another forum member and once the IHS was off I found that the Intel paste had set like concrete. Once I cleaned it up and applied Coolaboratory Liquid Ultra I relidded it it using some liquid gasket on the inside of the IHS (not the edge that mates with the pcb) and left it in the relidding attachment that came with the tool overnight. Put it in the pc the next day and got a whopping 25 degrees C reduction so it was well worth doing.

Previously I had done my old 4670k. This was before any special tools were available and I used the vice and hammer method which made for a very twitchy couple of minutes. Hitting your cpu with a hammer just isn't a natural thing to do but it went ok and once I had applied liquid metal relidded with liquid gasket (same way as above) stuck it in the motherboard socket and clamped it down with my waterblock. I got 15-18 degrees C off that one.

If you buy a proper delidding tool such as the Derbauer Delid Die Mate 2 it is practically risk free and takes only seconds to remove the IHS. It also comes with a relidding kit and after you have finished using it should be able to sell it on for near enough what you paid for it.
 
This is one of the better vice delid videos, in the fact that that he doesn't use a hammer.
and he makes you aware of the SMD components (resistors/capacitors) mounted on the CPU PCB.

So if you look at delidded images of the CPU, you'll see that the SMD components (these are the bits you want to cover in nail varnish if you use liquid metal or similar) are mounted to the left of the GPU chip, when you view the writing on the IHS (integrated heat sink) the right way up.
This means that it's safer to push off the IHS from right to left. The black sealant that you see has very little give in it, so it only takes 1mm or so of sideways movement to break the seal.
Whereabouts on the planet are you, as I'm sure someone local will have a vice or delid tool.
 
Last edited:
Thank you once again for the quick replies!

I must admit, this encouragement is making me feel more confident in giving it a go! Especially in that video where the guy used a vice to de-lid in a few minutes... so if I take it slow, I may well be able to do it lol

I'm in the SE of UK, Maidstone/Medway Towns area to narrow it down somewhat, that would be awesome if anyone is able/willing to do that :D

Oh quick question; I've seen a lot of people saying about re-lidding, which I think I'd like to do should I go ahead with this - which glue/material is best used for sealing the thing back up? Also I saw about putting nail varnish on over the capacitors, the Mrs. has quite the collection, is there a certain type I should look for? Other than it being clear ofc!

Thank you once again,

M
 
Thank you once again for the quick replies!

I must admit, this encouragement is making me feel more confident in giving it a go! Especially in that video where the guy used a vice to de-lid in a few minutes... so if I take it slow, I may well be able to do it lol

I'm in the SE of UK, Maidstone/Medway Towns area to narrow it down somewhat, that would be awesome if anyone is able/willing to do that :D

Oh quick question; I've seen a lot of people saying about re-lidding, which I think I'd like to do should I go ahead with this - which glue/material is best used for sealing the thing back up? Also I saw about putting nail varnish on over the capacitors, the Mrs. has quite the collection, is there a certain type I should look for? Other than it being clear ofc!

Thank you once again,

M
Personally I'd use coloured nail polish as you can then see that there's good coverage over conductors. Sort of handy! I don't think there's conclusive evidence but people have recommended nitrocellulose... If nail polish even specifies this kind of thing. I like to let it dry at least overnight because solvents continue to evaporate for a long time during drying.

For gluing the IHS back on, I've used a dot of epoxy on each corner. This holds well but makes it easy to remove later as it's not applied all around. Some people recommend using black silicone RTV (not acid-curing) sealant. That's quite close to what Intel have used.
 
ahhhh so I can bling up the insides whilst I'm at it with a nice metallic purple of some sort! :p

I did want to ask about the nail polish idea... would sealing the other components with varnish/other material not cause them to overheat at all? I'm guessing not as I've seen this often as advice on how to avoid short circuits, but wanted to ask all the same!

The dot of epoxy in the corners does sound like the better idea, keeps it together whilst easier to open again if needed. I did also see people saying that the layer of adhesive can cause changes in temperatures just in the sense of it spacing the die further from the IHS, so it seems to me like using the least adhesive possible when resealing is the way forward. I've looked up some epoxy at some stores near me, is this the sort of adhesive we're talking about? (please forgive my DIY noobness lol) - https://www.screwfix.com/p/araldite-2-part-epoxy-adhesive-tubes-opaque-2-x-15ml/2457h

Thank you again for the help folks!

M
 
I've not heard of instances where the nail varnish caused the SMDs to overheat. Don't get it on the CPU chip though!

If you use the Mrs' nail varnish, then you'll miss out on all the fun of walking into a makeup shop and asking for nitrocellulose based (and free from benzene derivatives such as toluene) nail varnish, then receiving blank stares back from the salesperson!

Sorry, I don't have any experience with that araldite epoxy, but I've generally seen the following recommended:
- RTV silicone (aka RTV silicone gasket maker)
- UHU 46735 High-Temperature Silicone Sealant
 
Yep the Araldite is just what you want. Main difference between types is simply how long it takes to set. I don't like the 5 minute stuff as I get panicky around glue so I have 30 minute stuff.
 
You guys are awesome!

If I end up de-lidding this thing, I'll make sure to let you know how I get on, and I'll likely at the very least take some pictures as if it all goes well I won't be opening another for some time!

So just to make sure I've got this right:
Get a kit like De8auer's - https://www.overclockers.co.uk/der8auer-delid-die-mate-2-hs-003-dr.html
Remove the IHS using the kit, and scrape away the black adhesive Intel used (using a credit card or similar), clean off the TIM using some paper towel dipped in cleaning solution/similar.
Apply nail varnish to the SMDs & let it dry for a while.
Apply a dab of LM to the die & spread with tool in kit, repeat for the underside of the IHS, apply a dab of this epoxy stuff on the corners, place the IHS down and pray... let set for an hour?
Hopefully profit from reduced temps!

Sounds simple enough, I don't believe there are any places near me that offer a de-lid service so if I want it, I gotta step up :D

Regards,

M
 
To clean the old TIM off, I'd recommend using isopropyl alcohol for the cleaning solution.

The rest looks good to me.

All the best!
 
Last edited:
Back
Top Bottom