Delidded my i7-3770K, loaded temperatures drop by 20°C at 4.7GHz

[GreenChile]

I'm willing to have my skepticism eliminated via education by numbers, lets see a few more reports of this actually working before we just assume it isn't fake.

Any volunteers?

It looks like there are quite a few who have had luck with this technique. http://www.overclock.net/t/1376206/how-to-delid-your-ivy-bridge-cpu-with-out-a-razor-blade

Seems reasonable to me.

 

[taq8ojh]

IDC, how much of that Noctua paste did you use on the IHS? I just got to play with it yesterday, and man it's thick as hell. I had to use razor to even try to spread it at least half evenly. It seems like I have to use really large amount to cover the whole area.

I actually haven't delidded yet, just messing around stil... a bit afraid lol.

 

[ehume]

IDC, how much of that Noctua paste did you use on the IHS? I just got to play with it yesterday, and man it's thick as hell. I had to use razor to even try to spread it at least half evenly. It seems like I have to use really large amount to cover the whole area.

I actually haven't delidded yet, just messing around stil... a bit afraid lol.

That does not sound like the Noctua paste I've worked with.

 

[Idontcare]

IDC, how much of that Noctua paste did you use on the IHS? I just got to play with it yesterday, and man it's thick as hell. I had to use razor to even try to spread it at least half evenly. It seems like I have to use really large amount to cover the whole area.

I actually haven't delidded yet, just messing around stil... a bit afraid lol.

I've used numerous (at least 4, maybe 5 by now) tubes of NT-H1, all bought separately and at different times, and none of them were remotely like what you are describing.

My tubes of NT-H1 are more like hand lotion in texture and viscosity. It isn't as thick as toothpaste, which is about the consistency of AS5.

Maybe you got a bum tube? I'd be skeptical of the mount in that case, you may be getting poor thermal performance as well.

 

[taq8ojh]

balls
I will squeeze some more out first and try to reapply tomorrow. Curious if I will see any temperature changes (I described some suspicious stuff here)

 

[Smartazz]

Anybody have plans to delid Haswell? Can we assume it will be achievable again?

 

[Idontcare]

Anybody have plans to delid Haswell? Can we assume it will be achievable again?

I was told by someone at Intel (a person who ought to know) that Haswell will be the same deal as IB for CPU TIM and IHS. Nothing changes there between IB and HW. But that is not confirmed as I only know what I've been told which makes it secondhand info, well thirdhand now that I'm repeating it to you.

 

[Unoid]

wtf? If intel is after cooler temps for notebooks why are they gimping themselves with the spreader interface material?

 

[Idontcare]

wtf? If intel is after cooler temps for notebooks why are they gimping themselves with the spreader interface material?

Mobile form factors don't come with the heatspreaders, for example sandy bridge laptop chips don't have the IHS.

The IHS is basically only used for desktop/workstation/server machines.

 

[oceanside]

I was told by someone at Intel (a person who ought to know) that Haswell will be the same deal as IB for CPU TIM and IHS. Nothing changes there between IB and HW. But that is not confirmed as I only know what I've been told which makes it secondhand info, well thirdhand now that I'm repeating it to you.

Hope not... I will skip Haswell if this is the case. Hopefully I can wait until Haswell-E, if not I may have to change the upgrade path.

 

[hawtdawg]

I seriously dont understand this strategy. They start selling OC insurance, and at the same time, start making their CPU's run 20C hotter than they should be. This is either a really clever strategy to sell OC insurance with a reduced risk on their end (using a very crude methode I might add, contact gap caused by glue?), or Intel is losing their edge.

 

[taq8ojh]

I'd much rather see them offer no warranty void if delidded service

 

[Idontcare]

I'd much rather see them offer no warranty void if delidded service

We still don't know why they technically changed the POR (process of record) away from the solder.

If it was due to lifetime reliability degradation brought on by the thermomechanical stresses then there is no chance they will bring the solder back.

However, if was more of a cost-saving effort via reduction in manufacturing complexity then it is possible they might bring the solder back again at some point in the future.

However after seeing how nice a bare-die mounted IvyBridge performs, personally I wish Intel would offered warrantied bare-die desktop socketed enthusiast chips.

They warranty their bare-die chips when they are sold to OEMs for mobile form factor builds like laptops and so on. So they must have some degree of comfort and clarity on the warranty claim exposure that occurs with bare-die applications. It wouldn't be a big risky unknown to them.

 

[TheVrolok]

Legit thread is legit.

 

[Smartazz]

We still don't know why they technically changed the POR (process of record) away from the solder.

If it was due to lifetime reliability degradation brought on by the thermomechanical stresses then there is no chance they will bring the solder back.

However, if was more of a cost-saving effort via reduction in manufacturing complexity then it is possible they might bring the solder back again at some point in the future.

However after seeing how nice a bare-die mounted IvyBridge performs, personally I wish Intel would offered warrantied bare-die desktop socketed enthusiast chips.

They warranty their bare-die chips when they are sold to OEMs for mobile form factor builds like laptops and so on. So they must have some degree of comfort and clarity on the warranty claim exposure that occurs with bare-die applications. It wouldn't be a big risky unknown to them.

Isn't trying to use a typical cooler like your H100, difficult on a bare die?

 

[taq8ojh]

I've used numerous (at least 4, maybe 5 by now) tubes of NT-H1, all bought separately and at different times, and none of them were remotely like what you are describing.

My tubes of NT-H1 are more like hand lotion in texture and viscosity. It isn't as thick as toothpaste, which is about the consistency of AS5.

Maybe you got a bum tube? I'd be skeptical of the mount in that case, you may be getting poor thermal performance as well.

offtopic continued:
So I took the cooler off and decided to reapply the paste just to see if it looked different further down the tube. Well... I am not exactly sure.

Maybe everything is ok and I just never saw paste like this before.
It's just the weird consistency. The thing creates sort of "hard" layer on the surface it's applied to, and rubbing it again (in the effort of even spread) virtually takes/tears entire spots away.
If I was to compare it to AS5, it is somewhat thicker (not by a lot really, but it's noticeable), and a bit rubbery, especially when applied (the layer...).

I managed to spread it the way I wanted, but I had to use my finger, and it took several attempts (yes I started to feel like an idiot ). It would definitely not spread enough just by placing a drop on IHS and mounting a cooler, like you would expect from a paste.

Is that normal?

 

[A5]

Isn't trying to use a typical cooler like your H100, difficult on a bare die?

Not any harder than getting the lid itself off:

http://forums.anandtech.com/showpost.php?p=34059601&postcount=579

 

[Idontcare]

Isn't trying to use a typical cooler like your H100, difficult on a bare die?

Not any harder than getting the lid itself off:

http://forums.anandtech.com/showpost.php?p=34059601&postcount=579

Yeah it really isn't difficult. The challenge for Intel and the mobo OEMs is they would have to change the socket retention mechanism to account for the difference in step-height to the top of the silicon die, and of course the HSF mounting kit itself would have to be shortened for the same reasons.

offtopic continued:
So I took the cooler off and decided to reapply the paste just to see if it looked different further down the tube. Well... I am not exactly sure.

Maybe everything is ok and I just never saw paste like this before.
It's just the weird consistency. The thing creates sort of "hard" layer on the surface it's applied to, and rubbing it again (in the effort of even spread) virtually takes/tears entire spots away.
If I was to compare it to AS5, it is somewhat thicker (not by a lot really, but it's noticeable), and a bit rubbery, especially when applied (the layer...).

I managed to spread it the way I wanted, but I had to use my finger, and it took several attempts (yes I started to feel like an idiot ). It would definitely not spread enough just by placing a drop on IHS and mounting a cooler, like you would expect from a paste.

Is that normal?

The way you are describing the stuff - "a bit rubbery", creates a "hard" layer, rubbing tears the spots away, etc - is not my experience with NT-H1.

What you are describing is exactly my experience with the putty-like TIMs - TX-1, TX-2, IC Diamond, IC Perihelion - which were rather thick and spread like cream cheese.

NT-H1 for me was like hand lotion. Yes it would hold its shape when you squirted out a drop of it onto the IHS, but it provided absolutely no resistance to be squished and flattened when spread.

 

[taq8ojh]

Hmm. Weird. It works, though - after reapplying and manually spreading the paste across whole IHS surface, temperatures dropped a little bit:

This is with roughly 1.116V vcore, by the way. I would say this is a proof my CPU is crying out to be delidded. There's no way in hell temperatures should be that high with such voltage. I saw temperatures spike up to 93°C on the hottest cores with higher LLC setting or just a tiny bit higher vcore.

 

[Idontcare]

Hmm. Weird. It works, though - after reapplying and manually spreading the paste across whole IHS surface, temperatures dropped a little bit:


This is with roughly 1.116V vcore, by the way. I would say this is a proof my CPU is crying out to be delidded. There's no way in hell temperatures should be that high with such voltage. I saw temperatures spike up to 93°C on the hottest cores with higher LLC setting or just a tiny bit higher vcore.

Yeah I agree with you, the TIM is probably just fine and fussing over it is proving to be a dead-end for you. The real limit for your CPU is that it hasn't been delidded.

 

[taq8ojh]

So I joined the club.
It was MUCH easier than I though. I was also fairly surprised the razor cut into the black stuff like through butter. Except for one corner, I barely needed any force at all.

Getting the stuff off took a little longer, but the razor did the job pretty well. I manned up and used it on the PCB as well Rest was taken care of with some cloth rubbing and nail scraping.
Applying the Liquid Ultra was pretty interesting, the stuff is pretty solid in consistence, and yet spreads like water. Funny.
Putting the IHS back took a few attemps, but I guess I managed to seat it back in place fairly well. It was freely rotating on the die btw, and I could clearly see significant gap between it and the PCB. The IHS is also somewhat concave it seems, which might be responsible for uneven temperatures even after the delid process.
Well, long story short:

YAY! I am very happy about the result, even though I am not particular fond of one of the cores being significantly hotter than others - but there's no way in hell I'd take the cooler off and go through the hell of reapplying NT-H1 again.

Oh btw,these temperatures are with 1.140V vcore. Not exactly awesome, but the heatsink is a little bit warmer than before.
I guess I will eventually buy different cooler (and lap the IHS too), because as much as I like the Hyper 212 EVO, the retention bracket sucks donkey balls. There's probably not enough force to hold the heatsink in place, and the pressure comes from basically just one small central spot, which I don't believe helps either. Bigger surface for the bracket please!!
Needless to say, the cooler is absolutely awesome for its cost, though.

 

[Idontcare]

Very nicely done Octopuss :thumbsup:

So I joined the club.
The IHS is also somewhat concave it seems, which might be responsible for uneven temperatures even after the delid process.
Well, long story short:

YAY! I am very happy about the result, even though I am not particular fond of one of the cores being significantly hotter than others

The cores will always be different in temperature like that because the die itself is not symmetric, so the lateral heat flow will always be more efficient in the outer-lying cores (making them cooler relative to the inner-lying cores).

 

[Ed1]

So I joined the club.
It was MUCH easier than I though. I was also fairly surprised the razor cut into the black stuff like through butter. Except for one corner, I barely needed any force at all.

Getting the stuff off took a little longer, but the razor did the job pretty well. I manned up and used it on the PCB as well Rest was taken care of with some cloth rubbing and nail scraping.
Applying the Liquid Ultra was pretty interesting, the stuff is pretty solid in consistence, and yet spreads like water. Funny.
Putting the IHS back took a few attemps, but I guess I managed to seat it back in place fairly well. It was freely rotating on the die btw, and I could clearly see significant gap between it and the PCB. The IHS is also somewhat concave it seems, which might be responsible for uneven temperatures even after the delid process.
Well, long story short:

YAY! I am very happy about the result, even though I am not particular fond of one of the cores being significantly hotter than others - but there's no way in hell I'd take the cooler off and go through the hell of reapplying NT-H1 again.

Oh btw,these temperatures are with 1.140V vcore. Not exactly awesome, but the heatsink is a little bit warmer than before.
I guess I will eventually buy different cooler (and lap the IHS too), because as much as I like the Hyper 212 EVO, the retention bracket sucks donkey balls. There's probably not enough force to hold the heatsink in place, and the pressure comes from basically just one small central spot, which I don't believe helps either. Bigger surface for the bracket please!!
Needless to say, the cooler is absolutely awesome for its cost, though.

nice results .
I have to disagree on center mounting pressure point of the evo , Yes the springs and bracket could be stronger (less flex) but having a center point makes sure pressure is even across the IHS .
Say it had just cross type 4 way brace , then depending on how much pressure you tightened each corner would determine pressure across core .

You could try shimming the springs to see if that would give little more pressure, it might help .

 

[taq8ojh]

Oh, right. So you think I actually did the best I could?

Somewhere in a corner of my mind I somehow hope lower temperatures might help stability, even though common sense says no.

@Ed: well, I don't mind it's "central", it's just that it gives me a bit unstable feeling. I was pretty happy with the way my old Thermalright HR-01 mounting worked.

Guess you get the idea there. It was rock-stable when screwed down. I liked how the pressure was done across a surface rather than a "pin". You could also tweak it even further by inserting whatever you felt appropriate under the bracket.

 

[Ed1]

Oh, right. So you think I actually did the best I could?

Somewhere in a corner of my mind I somehow hope lower temperatures might help stability, even though common sense says no.

@Ed: well, I don't mind it's "central", it's just that it gives me a bit unstable feeling. I was pretty happy with the way my old Thermalright HR-01 mounting worked.

Guess you get the idea there. It was rock-stable when screwed down. I liked how the pressure was done across a surface rather than a "pin". You could also tweak it even further by inserting whatever you felt appropriate under the bracket.

What clock speed are you at .
I have a 3570k @4.0-3.8 (200mhz OC across cores) and my temps are max of 61 (thats highest core) in prime95 with 72-74F room temps .

I use the default CM PWM fan but custom profile in Asus bios .
I see you changed to single speed ,maybe it is not doing as good as CM one .

Your delid results look very good to me (15-21c delta ) , good reduction amount .