FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,535
Intel Kaby Lake i5-7600K CPU De-Lid & Re-Lid Temp Results - We got in our retail purchased Intel Core i5-7600K Kaby Lake architecture processor and of course the first thing we wanted to do was overclock it to 5GHz, and then very quickly remove the Integrated Heat Spreader and see how much better we could make our load temperatures. Once again, we got very good results certainly worth sharing.

Edit: I did add to the review, the per core temps.
 
Last edited:
I don't understand why Intel doesn't go back to soldered IHSes (on non-HEDT k series processors, anyways). They know damned well that the cheapass TIM they're using is effecting thermals, and they know damned well we're on to their little game.
 
I don't understand why Intel doesn't go back to soldered IHSes (on non-HEDT k series processors, anyways). They know damned well that the cheapass TIM they're using is effecting thermals, and they know damned well we're on to their little game.

I think it is because they dont want the processors to be too much faster than their 6/8/10 core parts, but thats just my opinion.
 
I don't understand why Intel doesn't go back to soldered IHSes (on non-HEDT k series processors, anyways). They know damned well that the cheapass TIM they're using is effecting thermals, and they know damned well we're on to their little game.

It affects thermals sure, but it doesn't really affect the performance. The cpus are just as happy to run at 63c as they are 76c, they aren't going to switch back just for the K processors when the rest of their processors run just fine with the cheap stuff.
 
It affects thermals sure, but it doesn't really affect the performance. The cpus are just as happy to run at 63c as they are 76c, they aren't going to switch back just for the K processors when the rest of their processors run just fine with the cheap stuff.

They are trying to same money with a thinner PCB and TIM. Their thinking is you want solder go socket 2011-3 or AMD. With a stagnate AMD this is what you get unfortunately.
 
An easy 5 Ghz after deliding?

It's price gouging time from Newegg and Amazon!!
 
I don't understand why Intel doesn't go back to soldered IHSes (on non-HEDT k series processors, anyways). They know damned well that the cheapass TIM they're using is effecting thermals, and they know damned well we're on to their little game.

Shorter lifespan.

The CPUs run just as fast and give people that good feeling inside. But instead of lasting 10 years now they may only last for 5. (Just using hypothetical numbers to make a point)
 
I really ought to grab some of that RTV, I've just been letting the IHS float and using the tabs of the motherboard CPU retention clip to hold the IHS down onto the CPU. Seems to work well enough...

I want one of those de-lidding tools too but it just seems to be so much money to spend on something that I'm going to use like two or three times. :p
 
These Delid results are fine and all, but with all the extra voltage and such dumped in to these cores what exactly is the benefit you are getting. I know from what I've seen with the delids and 5GHz OC's you only see roughly 1FPS to 3FPS increase over a stock clocked 7700K, so at this point is it really worth it? I mean the stock MHz is 4500MHz which is basically what you need to remove the CPU bottleneck, is there any point after that?

I think [H] needs to do a full delid - relid review of a 7700K with 3600MHz DDR4 at 5GHz vs a stock 7700K with 3600MHz DDR4 at 4.5GHz and see if its worth voiding the warranty.
 
These Delid results are fine and all, but with all the extra voltage and such dumped in to these cores what exactly is the benefit you are getting. I know from what I've seen with the delids and 5GHz OC's you only see roughly 1FPS to 3FPS increase over a stock clocked 7700K, so at this point is it really worth it? I mean the stock MHz is 4500MHz which is basically what you need to remove the CPU bottleneck, is there any point after that?

I think [H] needs to do a full delid - relid review of a 7700K with 3600MHz DDR4 at 5GHz vs a stock 7700K with 3600MHz DDR4 at 4.5GHz and see if its worth voiding the warranty.

There are other stuff than games.
 
I think [H] needs to do a full delid - relid review of a 7700K with 3600MHz DDR4 at 5GHz vs a stock 7700K with 3600MHz DDR4 at 4.5GHz and see if its worth voiding the warranty.

That's the thing. For years and years the spirit of [H] has been in voiding the shit out of warranties to wring every last bit of juice out of our hardware. There are plenty of other sites that review CPUs off the shelf. Only a handful that tear the IHS off for the sake of a 10 degree difference and then STILL try their best to melt the bastards.
 
These Delid results are fine and all, but with all the extra voltage and such dumped in to these cores what exactly is the benefit you are getting. I know from what I've seen with the delids and 5GHz OC's you only see roughly 1FPS to 3FPS increase over a stock clocked 7700K, so at this point is it really worth it? I mean the stock MHz is 4500MHz which is basically what you need to remove the CPU bottleneck, is there any point after that?

I think [H] needs to do a full delid - relid review of a 7700K with 3600MHz DDR4 at 5GHz vs a stock 7700K with 3600MHz DDR4 at 4.5GHz and see if its worth voiding the warranty.

This is [H]ardOCP. Perhaps you took a wrong turn somewhere? :)

But in all seriousness there are plenty of situations where even a small bump makes quite the difference. For instance, I do a TON of video encoding for my Youtube channel. I appreciate any amount I can shave off encode times.
 
You'd be better off with MOAR CORES! then OCing a 7600K/7700K for workstation use, even then there are xeons out there.

Didn't take a wrong turn anywhere friend. Used to be a member many years ago but my account hasn't been in use for sometime so I re-created it. I am all for OCing. I delided a 4770K a few years back after a client had me put the system together and no longer wanted it ( his loss ). But I call in to question at what point is it practical for everyday use. Hell my 6900K hits 4.5GHz at 1.3v Realbench and Handbrake for 24 hours. But after 4.5GHz what do we really get in terms of performance? Or is it more beneficial to go the MOAR CORES! route?
 
This is [H]ardOCP. Perhaps you took a wrong turn somewhere? :)

But in all seriousness there are plenty of situations where even a small bump makes quite the difference. For instance, I do a TON of video encoding for my Youtube channel. I appreciate any amount I can shave off encode times.

And another big point is the drop in temps and dissipating that extra heat to offset the overclocking efforts. It will also make more of a difference in cpu limited games and sli/xfire as well. Not sure where his disconnect is on this?
 
. For instance, I do a TON of video encoding for my Youtube channel. I appreciate any amount I can shave off encode times.

This right here is why I shoot for 100% stability on OCs. I wanna have Hyperthreading, AVX, VT-d and QuickSync all working before I call an OC successful. No point splashing for bells and whistles if you're gonna kill half of em for an extra 100mhz and 0.005v lower vcore.
 
dumb question: Why do you need to reglue down the IHS? wouldn't removing the RTV all together eliminate it from adding to the overall height of the IHS and any gap between the IHS and the core?
 
There is no disconnect, I am asking a valid question, after voiding the warranty, increasing power draw and lowering temps what does this actually get us over a stock clocked 7600K/7700K?

It would make for an awesome review from [H]. Maybe even throw in a 6900K as well to see how well a 500MHz increase increase competes against cores.
 
dumb question: Why do you need to reglue down the IHS? wouldn't removing the RTV all together eliminate it from adding to the overall height of the IHS and any gap between the IHS and the core?
Well, as pointed out in the video, when I used very little adhesive and did not allow it to cure, I did not get good results.

When I did not use the adhesive "properly" the metal TIM matched the perf of the NH-T1 TIM. When I used the adhesive around the entire IHS and let it cure, I got the delta showed on the graph. So I think there is totally something to doing a good job at reattaching the IHS.
 
There is no disconnect, I am asking a valid question, after voiding the warranty, increasing power draw and lowering temps what does this actually get us over a stock clocked 7600K/7700K?

It would make for an awesome review from [H]. Maybe even throw in a 6900K as well to see how well a 500MHz increase increase competes against cores.
It did nothing. We did it for fun. Please move along now from this thread and let us have our fun. Jeez.

We heard you the first time.
 
Is it safe to cool these completely delidded without the IHS put back on? I'm pretty sure I saw some shims from one of the big watercooling guys, but not sure if that was for older cores. Love to see how cool it can get with direct die cooling.
 
Is it safe to cool these completely delidded without the IHS put back on? I'm pretty sure I saw some shims from one of the big watercooling guys, but not sure if that was for older cores. Love to see how cool it can get with direct die cooling.
That would depend on your definition of "safe." Those IHS are there for a reason, and it is not to just add another layer of thermal insulation. :) But as many of us know, those CPU dies are extremely fragile. I was lucky I did not chip the one in the video when I dropped the IHS on it.

That all said, sure you can run the CPU naked and cool it. They still make shims to keep from breaking the dies as easily IIRC. Next time I have a naked CPU, I will load it in the socket and check Z height on the die.
 
What heatsink or watercooling system are you using, and was it the same as when you tested temps on the 7700k?
 
Since you've actually done all this work, what's your opinion on the stock Intel TIM then? Is the TIM really crap, or is the temp improvement more due to the reduction in gap between the IHS and die when you remount the IHS by hand?

I'm asking because everyone just says "crap TIM", and I can't imagine Intel would use something so terrible just to save a couple cents per CPU. I CAN imagine they have a larger gap there due to their manufacturing tolerances.
Did you see how much stock TIM was left on the mating surfaces of that IHS when I pulled it? There was a LOT...at least what I would call a lot. After seeing my "failure" with the mount for the metal TIM, I am leaning more towards Z height being more of a factor than the actual TIM material. That said, I have no way to test that since I cannot return the IHS back to stock height with any kind of accuracy at all.

Now, I think...and this is subjective....that the Noctua is probably worth a degree or two compared to the stock TIM, and the metal TIM is much better than stock. Z height is bigger part of the equation than I think people are giving it credit for however.
 
I always heard that Intel uses Shin-Etsu, and from what I've seen it certainly looks like that's what they use. If that's the case then they could certainly do worse as far as TIM quality is concerned.

I think the reduction in temps comes mostly because often times the original TIM application is REALLY sloppy / uneven / too generous and also that glue they use to attach the IHS to the substrate can cause things to be uneven.

I don't regret de-lidding my 6700k in the slightest. It's been awhile but IIRC the difference in load temps was something like 13 degrees C.... didn't really help my max OC though. :D
 
Any chance to see a similar test done with air cooling? I'm mostly curious to compare stock CPU and without IHS. No need to relid with different paste.

If those results are interesting then I guess shims will sell like hot cakes.
 
Any chance to see a similar test done with air cooling? I'm mostly curious to compare stock CPU and without IHS. No need to relid with different paste.

If those results are interesting then I guess shims will sell like hot cakes.
Then I will just get beat up for not using the right cooler or fan or something else...
 
Thumbnail looks like a pop tart on a processor. Now I am hungry.
 
Nice results! Is it worth using that metal TIM on top of the IHS too or is it just better for under IHS use (on top of the die)?
 
Anyone know what the maximum heat these cpus can be exposed to before dying an evil death? The reason I ask is that I am considering soldering a cooler directly to the cpu. No idea if I could get it to stick to the cpu surface, but worth looking into.
 
Nice results! Is it worth using that metal TIM on top of the IHS too or is it just better for under IHS use (on top of the die)?
I would assume so. That is some electrically conductive messy shit though and I am sure if you used it there, you would want to be VERY careful and certainly made sure you cleaned up any that was pushed out during mating.
 
Anyone know what the maximum heat these cpus can be exposed to before dying an evil death? The reason I ask is that I am considering soldering a cooler directly to the cpu. No idea if I could get it to stick to the cpu surface, but worth looking into.
If you were going to do that, I would suggest delidding, then affixing the heatsink, then remounting the IHS. Also, I don't think soldering is going to do the job. In fact, I do not think that would work at all. You would be better off just going naked with the die and direct mounting.
 
Irrc that liquid metal stuff is a gallium mixture which makes it have a tendency to be quite a bit more corrosive then most would probably like to their IHS and heatsink especially with all the copper ones. Not sure i'd ever use it for long term use as i'd probably have to replace and lapp the surfaces every six months or something to ensure good contact and transfer stays.

Still interesting stuff on the remounting of the IHS i wouldn't have thought much about re-sealing it to get better distance between the die and IHS, surely i'd just let it float and try to crank up the pressure by mounting the heatsink hard onto the IHS to push both down.
 
Since I may upgrade it at after Ryzen comes out (and I got it cheap) I'm using the liquid metal TIM on a delidded 4770K as well as liquid metal between the IHS and the Noctua cooler I have on there.

The was a noticeable reduction in temps when I switched from a "normal" TIM between the IHS and cooler. Had to make sure there was no aluminum in there though. Have you tried it that way before Kyle? I know there's a very large risk with conductive material...

I can only get 4.3ghz stable on my 4770K despite the low temperatures though. I may need to play with voltages, which I haven't done much yet.
 
One of big reason for decrease of temps is because that stock RTV raises the IHS off the dies something like 0.020". It seems Intel want a good cushion in there. There post over in Anandtech where they measured everything and many tested.

The downside to deliding is the compound between the IHS and core never seems to last very long and if you use liquid metal types, it does last longer but when you do have to redo it, its real pain to separate cleanly.

That and it never seemed to give more speed, well maybe if you have real bad temps it might but for avg I don't think its worth it.

Adding back RTV is best way IMO, at least you limit any shifting.
 
I would be fairly curious to see what a direct die cooling temp would look like as well, or is that considered too risky these days?
 
Back
Top