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

[Borealis7]

EK has just announced a "naked ivy" mechanism for it's water blocks:

http://www.techpowerup.com/182088/E...-Lidded-Ivy-Bridge-CPU-Cooling-Solutions.html

 

[Idontcare]

EK has just announced a "naked ivy" mechanism for it's water blocks:

http://www.techpowerup.com/182088/E...-Lidded-Ivy-Bridge-CPU-Cooling-Solutions.html

Love it. Wish the air-cooler guys like Noctua would get on this as well

 

[tigersty1e]

ended up delidding my processor and slapping prolimatech tim on the die and ihs.

temps dropped a massive 22C.

im gonna watch out for the pump-out effect. I've skimmed various threads and i've only seen as5 and nt-h1 suffer, so I'll have to see.

I'm happy with the temp drop. I'll go with CLU only if the temps start to get worse over time. hopefully it doesn't.

 

[Idontcare]

ended up delidding my processor and slapping prolimatech tim on the die and ihs.

temps dropped a massive 22C.

im gonna watch out for the pump-out effect. I've skimmed various threads and i've only seen as5 and nt-h1 suffer, so I'll have to see.

I'm happy with the temp drop. I'll go with CLU only if the temps start to get worse over time. hopefully it doesn't.

Nice! If you do observe pump-out type issues (rising temperatures in a few months) please report back and alert the community

 

[taq8ojh]

I just found a video showing the most enormous temperature drop I ever heard of: http://www.youtube.com/watch?v=XXs0I5kuoX4

 

[Blades]

So step reduces the gap? In other words, how can I reduce the gap while retaining the stock TIM? I'm going to guess 'black gluey stuff'

 

[tigersty1e]

Nice! If you do observe pump-out type issues (rising temperatures in a few months) please report back and alert the community

sure will. i hope its not viral marketing from the coolaboratory folks to get people to buy their stuff.

but i'll probably go with CLU very soon. even if regular tim lasts 3 years, this system will still be chugging along 5+ years later. i have my old p4 system chugging along 10 years later at my parents business.

 

[Magic Carpet]

Love it. Wish the air-cooler guys like Noctua would get on this as well

I wonder, how big a market for this is. I suppose, that would be interesting for Intel as well. all those people doing it wrong and buying part again. good for business, I guess. Make it a TV show, nation wide, so people could have a new hobby.


Come on, Noctua. Everybody's counting on you

 

[Idontcare]

So step reduces the gap? In other words, how can I reduce the gap while retaining the stock TIM? I'm going to guess 'black gluey stuff'

I tried that, didn't work out. The stock TIM just doesn't like to work well once remated.

 

[PliotronX]

I have never regretted removing a noobshield. I took it a step further and desoldered my Q6600 using a blowtorch and then lapped the dies. Left the lid off and it resulted in higher overclocks at lower voltage. The only mod required was on the TRUE I added a nickel under the clamp to compensate for the missing IHS. A tip for safety, use a paper thin razor like these instead of a utility blade. The utility blades tend to bias towards the PCB and increase chances of nicking. Using these blades I get lids off in seconds with a nice amount of sealant remaining, not even close to damaging the CPU.

 

[tigersty1e]

there's a scenario I think would give better temps.

the mobo clamp is pushing down on the sides of the ihs and now that the glue is gone, the ihs gets pushed to the PCB and this creates a convex bow right over the CPU.

I'm thinking that we should just remove the mobo clamp and use the cooler mounts to secure the cpu to the mobo. the his will sit flush on the die and possibly never touch the PCB ( although with 50+pounds of mount pressure that is questionable)


Posted from Anandtech.com App for Android

 

[tigersty1e]

there's a scenario I think would give better temps.

the mobo clamp is pushing down on the sides of the ihs and now that the glue is gone, the ihs gets pushed to the PCB and this creates a convex bow right over the CPU.

I'm thinking that we should just remove the mobo clamp and use the cooler mounts to secure the cpu to the mobo. the his will sit flush on the die and possibly never touch the PCB ( although with 50+pounds of mount pressure that is questionable)


Posted from Anandtech.com App for Android

we don't even have to remove the clamp. we can just not lock down the clamp. the pressure from the cooler mount should be enough for the pads to make correct contact with the pins.


Posted from Anandtech.com App for Android

 

[Idontcare]

there's a scenario I think would give better temps.

the mobo clamp is pushing down on the sides of the ihs and now that the glue is gone, the ihs gets pushed to the PCB and this creates a convex bow right over the CPU.

I'm thinking that we should just remove the mobo clamp and use the cooler mounts to secure the cpu to the mobo. the his will sit flush on the die and possibly never touch the PCB ( although with 50+pounds of mount pressure that is questionable)


Posted from Anandtech.com App for Android

There isn't much thermal performance left on the table after you've delidded and replaced the stock CPU (and reduced the gap).

Going to the extreme of removing the clamp and the IHS entirely, mounting directly to the die, only improves on temps by a few degrees.

 

[tigersty1e]

Quote:

Originally Posted by tigersty1e

there's a scenario I think would give better temps.

the mobo clamp is pushing down on the sides of the ihs and now that the glue is gone, the ihs gets pushed to the PCB and this creates a convex bow right over the CPU.

I'm thinking that we should just remove the mobo clamp and use the cooler mounts to secure the cpu to the mobo. the his will sit flush on the die and possibly never touch the PCB ( although with 50+pounds of mount pressure that is questionable)


Posted from Anandtech.com App for Android

There isn't much thermal performance left on the table after you've delidded and replaced the stock CPU (and reduced the gap).

Going to the extreme of removing the clamp and the IHS entirely, mounting directly to the die, only improves on temps by a few degrees.

IMAGE#1

yeah, but what I'm proposing requires no effort. I have to look at my mobo clamp again. but I think there was a way to keep the clamp on the mobo and not have it lock it in place.

we would get much better Tim seats because the his doesn't slide 1 mm over the die during the clampdown. I'll take 1-2c for free anyday.


Posted from Anandtech.com App for Android

 

[red454]

First time I have seen a reference as to the reason for use of thermal paste on IB...

http://www.tweaktown.com/articles/5...rements-and-overclocking-thoughts/index2.html

"What this means is your Sandy Bridge and Ivy Bridge coolers should work fine with Haswell. Since this is a new product Haswell should have a more efficient IHS when compared to Ivy Bridge. Intel ran into a little issue with the processor core and the IHS. At one time it was reported the issue had to do with using thermal paste instead of soldering the core to the HIS, but later we learned a thermal paste had to be used because the core didn't make good contact with the metal top."

 

[Idontcare]

First time I have seen a reference as to the reason for use of thermal paste on IB...

http://www.tweaktown.com/articles/5...rements-and-overclocking-thoughts/index2.html

"What this means is your Sandy Bridge and Ivy Bridge coolers should work fine with Haswell. Since this is a new product Haswell should have a more efficient IHS when compared to Ivy Bridge. Intel ran into a little issue with the processor core and the IHS. At one time it was reported the issue had to do with using thermal paste instead of soldering the core to the HIS, but later we learned a thermal paste had to be used because the core didn't make good contact with the metal top."

As an engineer I can fathom a number of different thermo-mechanical stress related reasons why Intel might have been compelled to abandon metal solder in favor of the TIM-paste approach, but the one given there by tweaktown just makes no sense.

Making "good contact" is a very simple matter of surface engineering, I can't understate how simplistic of a process that is. It is so simple that Intel had no issue doing it in years past.

To say they couldn't do it with IB is virtually unbelievable. I mean sure it is possible, in the down-right mentally deficient "der, I haz brain-damage and cants git'r'dun" sort of engineering challenge, that is possible and cannot be ruled out, but not in a "all of our collective rocket scientists were totally stumped by this challenge!" manner.

The silicon backside of IB is the same as SB. The 22nm stuff, including the finfets, is all on the other side of the chip, opposite to the surface that is being mated to the IHS. If they had a process that enabled soldering the chip to the IHS with 32nm then that same process would work for 22nm, the solder and the IHS don't know 32nm from 22nm.

 

[red454]

As an engineer I can fathom a number of different thermo-mechanical stress related reasons why Intel might have been compelled to abandon metal solder in favor of the TIM-paste approach, but the one given there by tweaktown just makes no sense.

Making "good contact" is a very simple matter of surface engineering, I can't understate how simplistic of a process that is. It is so simple that Intel had no issue doing it in years past.

To say they couldn't do it with IB is virtually unbelievable. I mean sure it is possible, in the down-right mentally deficient "der, I haz brain-damage and cants git'r'dun" sort of engineering challenge, that is possible and cannot be ruled out, but not in a "all of our collective rocket scientists were totally stumped by this challenge!" manner.

The silicon backside of IB is the same as SB. The 22nm stuff, including the finfets, is all on the other side of the chip, opposite to the surface that is being mated to the IHS. If they had a process that enabled soldering the chip to the IHS with 32nm then that same process would work for 22nm, the solder and the IHS don't know 32nm from 22nm.

Yep - I'm and engineer too (mechanical) and you have to wonder what changed. If it has worked for so many years, why the shift? Did the bean counters get involved?

 

[Face2Face]

Here is a great solution for anyone who owns a EK Supremacy water block and would like to Delid and not use the IHS. You may be able to get them to work with other configurations?

http://www.eteknix.com/ek-announce-...ce-water-blocks-for-de-lidded-ivy-bridge-cpus

 

[Revolution 11]

I am not a engineer, but isn't making "good contact" something the ancient Greeks and Romans could do. Certainly not on a level of precision required for CPUs but the principle is the same.

 

[tigersty1e]

Quote:

Originally Posted by Idontcare

As an engineer I can fathom a number of different thermo-mechanical stress related reasons why Intel might have been compelled to abandon metal solder in favor of the TIM-paste approach, but the one given there by tweaktown just makes no sense.

Making "good contact" is a very simple matter of surface engineering, I can't understate how simplistic of a process that is. It is so simple that Intel had no issue doing it in years past.

To say they couldn't do it with IB is virtually unbelievable. I mean sure it is possible, in the down-right mentally deficient "der, I haz brain-damage and cants git'r'dun" sort of engineering challenge, that is possible and cannot be ruled out, but not in a "all of our collective rocket scientists were totally stumped by this challenge!" manner.

The silicon backside of IB is the same as SB. The 22nm stuff, including the finfets, is all on the other side of the chip, opposite to the surface that is being mated to the IHS. If they had a process that enabled soldering the chip to the IHS with 32nm then that same process would work for 22nm, the solder and the IHS don't know 32nm from 22nm.

Yep - I'm and engineer too (mechanical) and you have to wonder what changed. If it has worked for so many years, why the shift? Did the bean counters get involved?

I think a more likely scenario is that they purposely made the heat transfer more inefficient to avoid the pump out issues on top of the ihs.

on a factory ivy bridge chip, you don't run into pump out issues (the pump out may show itself in years rather than weeks as some have seen on delidded dies) on the ihs because the very hot cores don't get concentrated into the same spots on the ihs. the ihs and Intel tim transfers a more muted uniform heat to the heat sink, so the Tim there can be a more liquid goop and doesnt have to do tough job of transferring. if they had used solder, the tim on the ihs would be more susceptible to pump out because the heat transfer would be so great and Intel would then have more heat failures from people using cheap Tim.

that is also the reason they made the Tim so thick. you want thick Tim that has more solids.

as far as the gap. there could be a dozen reasons. but the more likely reason is that you can't be for certain how thick the black glue can get. you'd rather be safe and have that gap than to get the gap to be zero and have some cpus come out with glue that doesn't stick.


Posted from Anandtech.com App for Android

 

[Charles Kozierok]

To say they couldn't do it with IB is virtually unbelievable. I mean sure it is possible, in the down-right mentally deficient "der, I haz brain-damage and cants git'r'dun" sort of engineering challenge, that is possible and cannot be ruled out, but not in a "all of our collective rocket scientists were totally stumped by this challenge!" manner.

I agree. That sounds like pure BS to me.

 

[tigersty1e]

i have clu coming soon. i'll be using it directly on the die.

are there pump-out problems with using regular tim on top of the ihs?

should i also go ahead and put the clu on my ihs?

 

[Idontcare]

i have clu coming soon. i'll be using it directly on the die.

are there pump-out problems with using regular tim on top of the ihs?

should i also go ahead and put the clu on my ihs?

According to IC Diamond, traditional paste-TIMs are susceptible to pump-out over the course of 1-2yrs of use.

If you go to their website they have a study published on it in which they show (naturally ) that their IC Diamond paste has been formulated to not have pump-out issues.

Anecdotally people have reported here in the forums that they too have noticed that they need to replace their TIMs (ceramique, AS5, NT-H1, etc) every year or two as their temperatures creep up over time.

So the question is how much effort do you want to go through to avoid pump-out?

I use NT-H1 on my IHS but CLU on the silicon die. I don't mind cleaning and replacing the IHS TIM.

If you put CLU on your IHS then just be aware that should you ever want to remove your HSF then you are going to have to lap the HSF and IHS surfaces to fully remove the CLU. It is not a problem if you intend on leaving them together basically until you replace them.

 

[Blades]

I thought my CPU was a goner when I accidentally exposed some copper bling from the PCB. This is prior to cleanup of PCB.. Once I came to my senses, got the correct exacto blade, and began to use caution... I tried a couple things to easily remove the black glue... Mass air flow cleaner.. IPA on roids... Works wonders, just don't even think of flames/sparks.

I haphazardly put everything back together, expecting a no-boot or fire... To my surprise, it worked perfectly. Without lapping the IHS or the h80 contact plate.. I even dropped (and subsequently dinged the IHS).. Peak temps dropped by 10-12c.. Idle by about 4c... This is at 4.66ghz... LLC set aggressively on stock voltage (so under load voltage gets high).

The most important improvement (for me) is the elimination of what could be described as temperature creep (under continuous load, be it partial or all cores).. Before this, the temperature ceiling would slowly rise from 70 to high 70s.. Now, 70 is the max.

Again, this setup is somewhat botched - yet somehow greatly improved over stock.. So there you have it, the worst case scenario.

 

[Idontcare]

I thought my CPU was a goner when I accidentally exposed some copper bling from the PCB. This is prior to cleanup of PCB.. Once I came to my senses, got the correct exacto blade, and began to use caution... I tried a couple things to easily remove the black glue... Mass air flow cleaner.. IPA on roids... Works wonders, just don't even think of flames/sparks.

I haphazardly put everything back together, expecting a no-boot or fire... To my surprise, it worked perfectly. Without lapping the IHS or the h80 contact plate.. I even dropped (and subsequently dinged the IHS).. Peak temps dropped by 10-12c.. Idle by about 4c... This is at 4.66ghz... LLC set aggressively on stock voltage (so under load voltage gets high).

The most important improvement (for me) is the elimination of what could be described as temperature creep (under continuous load, be it partial or all cores).. Before this, the temperature ceiling would slowly rise from 70 to high 70s.. Now, 70 is the max.

Again, this setup is somewhat botched - yet somehow greatly improved over stock.. So there you have it, the worst case scenario.

Every once in a while, not often but often enough, a member posts the type of post that just makes their username all the more apropo.

< Blades >