Lapped my i7-3770K and there was zero improvement in operating temperatures!?

[Puppies04]

The argument about warranty makes more sense. What we're wondering is what risk they encountered in an established process applied to this new die-shrink, that would make them choose such a s***ty and ineffective alternative.

Perhaps the exact risk they reported when all this "Intel is using shitty TIM" arguement started. The increased thermal cycling caused by the smaller denser cores and new transistors meant that the existing solution could have failed over time.

P.S Your use of the word ineffective seems a bit off considering I haven't heard of a single IB chip that runs excessivly hot before it is overclocked.

 

[Puppies04]

Just a quick thought, could there have been a correlation between the shape of the unlapped IHS and the H80 contact plate which is now no longer there because of the lapping or is the H80 plate either already lapped or known to be sufficiently flat? I know it doesn't sound very plausible but I just thought I would throw it out there.

 

[graysky]

What i was suggesting is sanding some material off the bottom of the IHS so that when you reseat it onto the CPU it will make the gap beween the IHS and the CPU die smaller, which should theoretically improve thermal performance.

I think that is a bad idea. My hypothesis is that when the caulk is removed (black dots) the headroom above the die (red arrows) decreases by that same distance. So if you lap the bottom of the IHS, you will decrease further the internal headroom perhaps to the point of crushing the die when locked in place. Just my thought. Others?

 

[BonzaiDuck]

Just a quick thought, could there have been a correlation between the shape of the unlapped IHS and the H80 contact plate which is now no longer there because of the lapping or is the H80 plate either already lapped or known to be sufficiently flat? I know it doesn't sound very plausible but I just thought I would throw it out there.

I posed that possibility earlier in the thread. No -- it was the orientation of the waterblock. The OP needed to reinstall it so that it had been turned 90 degrees relative to the original installation.

ON the "ineffectiveness" issue -- I'll answer with another question: "What's the purpose if releasing a "K" version of a CPU, if you can't change the multiplier much beyond the default "turbo" level?" That is, some IB purchasers feel limited to that degree.

After the years of my worklife and what I've seen, I don't trust that even people of great experience or large established organizations never miss simpler solutions and explanations. But that's why we're here -- we think they may have had better fixes for the problem and simply chose the quick-n-dirty option. We're at least exploring that possibility, and there's already some initial data based on better TIM choices.

You're right that the "problem" isn't that severe, since we can see people OC'ing these chips to Sandy Bridge levels. But I've seen cases of people posting in forums outside Anandtech reporting much improvement with various TIM options.

 

[BonzaiDuck]

I think that is a bad idea. My hypothesis is that when the caulk is removed (black dots) the headroom above the die (red arrows) decreases by that same distance. So if you lap the bottom of the IHS, you will decrease further the internal headroom perhaps to the point of crushing the die when locked in place. Just my thought. Others?

I was thinking the same thing . . . . actually . . . I sort of felt that if you're going to separate the cap from its adhesive, best to replace it the same way. . .

 

[Ferzerp]

I did just that (reduced the gap) when I replaced my TIM. I know IDC had it on his list of things to do. I actually sanded the standoff of the IHS down to reduce it a bit more.

 

[Denithor]

As long as you don't reduce the remaining gap too much you'll be fine. The thinner the layer of TIM the better the heat transfer through it.

Which is also the reason why lapping works. You smooth out the two contact faces, eliminating any large gaps present due to concave/convex surfaces. The result is a much thinner gap for the TIM to fill and better heat transfer through it.

 

[BonzaiDuck]

As long as you don't reduce the remaining gap too much you'll be fine. The thinner the layer of TIM the better the heat transfer through it.

Which is also the reason why lapping works. You smooth out the two contact faces, eliminating any large gaps present due to concave/convex surfaces. The result is a much thinner gap for the TIM to fill and better heat transfer through it.

As I noted on the other thread, some pictures posted by Maximum PC of the die/PCB don't suggest any risk for using a CoolLaboratory "Liquid" product. Some outfit in Japan called PC Watch tried it with Liquid Pro for a 15C temperature reduction, 4.6Ghz @1.2V.

If the adhesive is rubbery, just replacing the IHS without further tedium and installing a HSF with the spring-loaded pressure would assure the same level of contact as before.

Well, I could swear on the book that I'd go ahead and do this if I were going to spring for an IB at this moment. It just looks like someone else -- probably one of you -- will be my default guinea-pig. But it just looks fairly straightforward -- not for the mainstreamers -- but us "DIY'ers."

Feels funny, though -- the thought that we'd be buying the product just so we could put ourselves and the purchase through this kind of tedium and trouble . . . Perhaps some much better over-clocks can salvage IB's reputation . . . . while we would stand less chance of looking . . . ah . . . stupd . . .

 

[MustangSVT]

any testing done after de-lidding?

 

[BonzaiDuck]

any testing done after de-lidding?

Depends on who has done it. There's a parallel thread, "IHS removal and results" by Graysky. He only used AS5, and got a 5C drop (if I remember).

Here's a summary of results from some outfit named "PC Watch" quoted and summarized by Maximum PC:

http://www.maximumpc.com/article/ne..._ivy_bridge_overclocking_temperatures#slide-0

 

[MustangSVT]

Depends on who has done it. There's a parallel thread, "IHS removal and results" by Graysky. He only used AS5, and got a 5C drop (if I remember).

Here's a summary of results from some outfit named "PC Watch" quoted and summarized by Maximum PC:

http://www.maximumpc.com/article/ne..._ivy_bridge_overclocking_temperatures#slide-0

thanks! should be a fun read!:thumbsup:

edit. ehh.. nevermind. IDC here has more info then that xD.
So, with higher end thermal paste, the temp does go down, but does it improve the oc ability of the chip?

 

[Idontcare]

any testing done after de-lidding?

Tests have been done, a bunch of intermediate checks, and am only just now getting to the interesting tests (NT-H1 under IHS, etc).

Will post them up later...preliminary result is showing a ~8°C decrease in temps at 3.5GHz, will crank up the clockspeed as the tests proceed.

Bit of a teaser here to tide you over - the CPU PCB cleans up nice with a bit of IPA and the help of a dremel D:

^ Image of the black adhesive after using a razor blade to scrape off as much as possible

^ yes that is a dremel with a buffer pad, saturated in IPA

^ ah the shiny preciouses are now ready for some OC'ing fun :twisted:

 

[MustangSVT]

IDC.. can I have your babies?

I CAN'T wait! great job and wtf... do you have a 3rd arm? how did you take those photos! very nice!

 

[r3pshow]

Great Job IDC.! Can't wait for the results !

 

[Piano Man]

Can't wait to see the results, IDC. I really appreciate the detailed documentation and I hope you continue it. If all goes, well, I am definitely thinking about getting the 3770K and following your example.

 

[zebrax2]

A 8°C drop in temp is amazing! Waiting for the other results as well as your analysis on it.

Anand should hire you to do this kind of things so that it could be posted on the main page.

 

[Ferzerp]

8C at stock is about what I would expect for what I saw OC'ed.

 

[Charles Kozierok]

Once you're going to the trouble of delidding, why not skip the lid entirely and mount the cooler right on the die?

A bit riskier, but eliminates the problem of TIM and balancing the spreader etc. entirely.

 

[Denithor]

IDC, are you planning to test anything other than NT-H1 under the hood?

I would be very curious to see how those liquid metal TIMs function, preferably both under IHS and also between IHS and HSF. Seems like those would be the best thing short of solder like Intel used previously.

And that's another topic - would it be possible to use solder somehow under there?

 

[2is]

IDC, are you planning to test anything other than NT-H1 under the hood?

I would be very curious to see how those liquid metal TIMs function, preferably both under IHS and also between IHS and HSF. Seems like those would be the best thing short of solder like Intel used previously.

And that's another topic - would it be possible to use solder somehow under there?

They work quite well. I saw a 6c drop just by using it between the hsf and IHS over using AS5

 

[BonzaiDuck]

They work quite well. I saw a 6c drop just by using it between the hsf and IHS over using AS5

The proof is already out there: the various CoolLaboratory products, Indigo-Xtreme, etc. are better than IC-Diamond (which I've consistently used since 2007).

It may be that IDontCare wants to test various pastes. The PC Watch data, though, is already known. This NT-H1 is so far a mystery to me only for Noctua omitting mention of its components.

And of course, somewhere, there is a comprehensive comparison review of all TIMs, or maybe TIMs that we could classify into different categories.

The liquid metal products and metal pads use indium, gallium and possibly silver in combination. They form an amalgam or bond with metal surfaces like copper or nickel, although I've heard it's not very troublesome to break the bond when removing a heatsink.

I don't think they would make any sort of bond with a silicon substrate like the top of the IB die. But the bonding with the other surfaces is probably the closest one would get to a solder made of indium/silver.

CharlesKozierok: Various complications make this sort of thing troublesome. I'd suggested -- to maintain the stability and protection for the CPU die -- grinding off the top of a heatspreader leaving a square-shaped ring that has overall height of IHS minus the thickness of the top. And you would have to use a cooler that had a flat surface -- as opposed to the "direct-touch" coolers, or you would be inclined to.

The difference of less than a millimeter would not likely be a problem for the spring-loaded pressure of the HSF assembly. I don't think, with that approach, there would be any difficulty with clearing the CPU motherboard clip assembly.

It's just a lot of extra work, and would require a certain precision.

 

[Puppies04]

I posed that possibility earlier in the thread. No -- it was the orientation of the waterblock. The OP needed to reinstall it so that it had been turned 90 degrees relative to the original installation.

ON the "ineffectiveness" issue -- I'll answer with another question: "What's the purpose if releasing a "K" version of a CPU, if you can't change the multiplier much beyond the default "turbo" level?" That is, some IB purchasers feel limited to that degree.

After the years of my worklife and what I've seen, I don't trust that even people of great experience or large established organizations never miss simpler solutions and explanations. But that's why we're here -- we think they may have had better fixes for the problem and simply chose the quick-n-dirty option. We're at least exploring that possibility, and there's already some initial data based on better TIM choices.

You're right that the "problem" isn't that severe, since we can see people OC'ing these chips to Sandy Bridge levels. But I've seen cases of people posting in forums outside Anandtech reporting much improvement with various TIM options.

Will these various aftermarket TIM option last the lifetime of the chip without drying out/cracking though? Intel don't have the luxury of using some exotic TIM that will last 3 or 4 years before it needs cleaning off (let alone the 1-2 years which is the max I would use TIM for before removing/replacing it) and reapplying, they need whatever they use to last 6 or 7 years and I am not aware of an aftermarket TIM that I would be happy relying on this for, let alone an aftermarket TIM company that advertises it's TIM will last this long.

 

[Denithor]

Will these various aftermarket TIM option last the lifetime of the chip without drying out/cracking though? Intel don't have the luxury of using some exotic TIM that will last 3 or 4 years before it needs cleaning off (let alone the 1-2 years which is the max I would use TIM for before removing/replacing it) and reapplying, they need whatever they use to last 6 or 7 years and I am not aware of an aftermarket TIM that I would be happy relying on this for, let alone an aftermarket TIM company that advertises it's TIM will last this long.

Well, the liquid metal ones should meet this requirement - if they're 100% metal they should not 'set up' as there's nothing there to dry out.

 

[PlasmaBomb]

IDC.. can I have your babies?

I CAN'T wait! great job and wtf... do you have a 3rd arm? how did you take those photos! very nice!

He just blinked then uploaded...

 

[PlasmaBomb]

Will these various aftermarket TIM option last the lifetime of the chip without drying out/cracking though? Intel don't have the luxury of using some exotic TIM that will last 3 or 4 years before it needs cleaning off (let alone the 1-2 years which is the max I would use TIM for before removing/replacing it) and reapplying, they need whatever they use to last 6 or 7 years and I am not aware of an aftermarket TIM that I would be happy relying on this for, let alone an aftermarket TIM company that advertises it's TIM will last this long.

Intel probably don't have much of a problem with their TIM since the heatspreader was originally sealed down, which should dramatically reduce TIM dry out...