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

[BonzaiDuck]

I understand that, but I'm wondering if there is a any "slow degrade" pattern to it if the transfer ability is observed over a long period. I get that once it's reached the "failure" point it will take just moments for the temperatures to spike.

For the diamond stuff, you'd think that a month or two would be a good enough profile, with shutdowns, cooling down and cold starts.

Thermal Cycling/TIM Failure Test

System Info:

3570K Delidded with IC Diamond between the die and IHS
ZT-10D HSF with Arctic Silver Ceramique between the IHS and the HSF
4.5Ghz @ 1.232v (after vdroop)

Delidded and IC Diamond applied 8/29/2012

After one week:
(9/4/2012)

Ambient temp - 25C
3570K max temps after Prime "blend" for 30 minutes - 67C, 76C, 74C, 73C

More results to come...

Was it better than that after you first ran it with the IC diamond? These numbers don't look half-bad, and your choice of a heatsink . . well, I searched for a comparison review:

http://www.frostytech.com/articleview.cfm?articleid=2557&page=5

. I suppose that could be right in a 77F room ambient. You wonder which groupings of two or three are more accurate over a fourth outlier. The sensors are supposed to be accurate within a range, + or -.

Those temperatures are maybe 6 or 7C above my Sandy's at 4.6. But your IB is more heat tolerant, at least relative to the given specs.

Suppose it stabilizes, or remains the same. You might then want to apply some more ICD to the bottom of the IHS to see if it improves to a new stable set. Or so it would be my own inclination, if I were de-lidding an IB like yours. Of course, the cost rises a bit more steeply given the price of the diamond paste.

Eventually we're going to start making sound conclusions about it.

 

[dqniel]

For that voltage and speed, during the small FFT portion of the test, I'm happy with that maximum temperature range. Frostytech uses an unrealistic (albeit repeatable) method to mimic actual CPUs, and the different heat "profile" (for lack of a better word) of real CPUs can make the heatsinks act quite differently. The heat coming out of a 3570k is coming from a much more confined location than, say, their LGA775 IHS-sized copper interface die.

Anyway, these temperatures are similar to what I got a week ago at similar settings. They're a bit higher, but my ambient temperature was also about 2C lower then.

 

[BonzaiDuck]

For that voltage and speed, during the small FFT portion of the test, I'm happy with that maximum temperature range. Frostytech uses an unrealistic (albeit repeatable) method to mimic actual CPUs, and the different heat "profile" (for lack of a better word) of real CPUs can make the heatsinks act quite differently. The heat coming out of a 3570k is coming from a much more confined location than, say, their LGA775 IHS-sized copper interface die.

Anyway, these temperatures are similar to what I got a week ago at similar settings. They're a bit higher, but my ambient temperature was also about 2C lower then.

I was trying to say that the rank-ordering of those tests is consistent with other, just as comprehensive tests as the FrostyTech, so you could expect just a little less with your heatsink than with either my D14 or IDontCare's Corsair H100.

At that point, it may be guesswork. But it doesn't seem "out of line." If it remains stable, you could choose to leave it alone and maybe see what happens with the next two higher multipliers, or you could see what happens by adding more TIM.

I can bet that IDontCare is puttering away with his rig right now to see where this goes. If we as "enthusiast" warranty-breakers can't resolve the problem we initially saw before we [some of us] delidded our IB's, I'll feel worse for those who already "popped their caps."

Here's another thought, and I know it may also sound crazy. Before this Ivy Bridge issue ever emerged, I thought I was examining web-pages for "Indigo-Xtreme" as well as the old IC Diamond, and . . . I THOUGHT . . . . I read something about "use in combination with other TIMs." I don't want to raise any hopes, and I'm a firm believer that the simplest choices lead to the best result. I don't think "TIM combinations" fill the bill there, but maybe someone will look into it. Carefully . . . . before actually executing such an approach . . .

No . . . just checking the Indigo product page. That would be absolutely wrong. Maybe it was something I read at IC . . .

 

[Ferzerp]

With an H100 and 1.232V (peak, often 1.224V), running only 4.4Ghz though, my peaks are 66C. Ambient, I am not so sure about, but my thermostat is set for 74F, so it's within a few degrees of that, likely.

 

[BonzaiDuck]

With an H100 and 1.232V (peak, often 1.224V), running only 4.4Ghz though, my peaks are 66C. Ambient, I am not so sure about, but my thermostat is set for 74F, so it's within a few degrees of that, likely.

Last I checked, you were using AS5 and the TIM IDontCare first applied? Is this result for diamond paste? I assume that it is.

 

[Ferzerp]

Last I checked, you were using AS5 and the TIM IDontCare first applied? Is this result for diamond paste? I assume that it is.

This is with NT-H1 all the way through. These are also consistent with my best results.

My last repair was on 8/22/2012, two weeks ago. Temps haven't changed.

However, as is often the case, my usage profile has. This box used to be a 24x7 box, as it housed my TV tuner for all my MCEs in the house. Now it is gaming only, and one of my HTPCs has the tuner. (Powered on time went from 24 hours a day to anywhere from 0-4 hours a day).

 

[BonzaiDuck]

This is with NT-H1 all the way through. These are also consistent with my best results.

My last repair was on 8/22/2012, two weeks ago. Temps haven't changed.

However, as is often the case, my usage profile has. This box used to be a 24x7 box, as it housed my TV tuner for all my MCEs in the house. Now it is gaming only, and one of my HTPCs has the tuner. (Powered on time went from 24 hours a day to anywhere from 0-4 hours a day).

I would think that an "on-again-off-again" situation would be best for this business . . .

 

[iamgenius]

Very very interesting thread. I wonder if I can squeeze some more MHz out of my 2700k by deli-dding it.

 

[Idontcare]

Very very interesting thread. I wonder if I can squeeze some more MHz out of my 2700k by deli-dding it.

Do NOT attempt this mod on your 2700k, its IHS has already been soldered onto the CPU and if you attempt to remove it then you will rip the CPU off the PCB (destroying it forever).

 

[hachre]

Hey IDC,

great great thread Thanks for all of your hard and very precise work !!!

Didn't you want to test several TIMs for comparison of how good they would do as a replacement for the stock Intel TIM (under the IHS)?!

I'm also very curious about a regular TIM test (on top of the IHS) with whatever the winner of the below IHS test will be to find out what the best combination would be overall...

I'm not sure if I misunderstood you wanting to do these tests or if I already missed them somehow ;D

And I'm also really curious about a direct die test without the entire IHS as well if you're up for it after the other tests )

 

[BonzaiDuck]

Do NOT attempt this mod on your 2700k, its IHS has already been soldered onto the CPU and if you attempt to remove it then you will rip the CPU off the PCB (destroying it forever).

Yeah. More and more or here and there, I noticed some new folks kicking in their two-cents-worth, or ready to run out and buy an Ivy Bridge and de-lid without further thought. I'm sure you will agree with me that you (and others) aren't exactly out of the "risk-zone" yet while conducting these experiments. We want to know that a superior PERMANENT or semi-permanent solution awaits the irreversible step of just de-lidding the Ivy Bridgers.

There's enough info in the thread to explain why this is a no-no for the Sandy Bridge processors or anything earlier. We have the problem of instilling in some the need to read more, more thoroughly, and make choices that are deliberate and well-planned.

 

[Idontcare]

Hey IDC,

great great thread Thanks for all of your hard and very precise work !!!

Didn't you want to test several TIMs for comparison of how good they would do as a replacement for the stock Intel TIM (under the IHS)?!

I'm also very curious about a regular TIM test (on top of the IHS) with whatever the winner of the below IHS test will be to find out what the best combination would be overall...

I'm not sure if I misunderstood you wanting to do these tests or if I already missed them somehow ;D

And I'm also really curious about a direct die test without the entire IHS as well if you're up for it after the other tests )

Yeah I've still got a number of tests planned.

I'm currently hung-up on attempting to get the gap between the CPU and the IHS to be the same as it was before I delidded the CPU.

That way I can generate some apples-to-apples data for comparing the thermal properties of Intel's stock CPU TIM versus that of the other 3rd party TIM's we have at our disposal.

My efforts to buy/use shim stock did not turn out as epically glorious as planned :\ (I could upload some of the fail for people's amusement if folks are interested in that, as you can imagine I wasn't highly motivated to upload my fail )

I've been thinking up substitutes (my current plan is to just use paper shims until I get the right gap) but in the meantime life has gotten in the way of my second favorite hobby (OC'ing) as well as my first favorite hobby (imbibing ). I'm hoping to get back into the swing of things in the coming week.

 

[hachre]

Cool, thanks for the update I wouldn't mind seeing some fail pictures, I'm curious about what happened Unless it takes too much of your time of course

 

[Idontcare]

Cool, thanks for the update I wouldn't mind seeing some fail pictures, I'm curious about what happened Unless it takes too much of your time of course

You had me at "great great thread" Sure I'll load it up, pretty sure the pics are still on the camera so I can get to it later today.

Also, welcome to forums hachre :thumbsup:

 

[hachre]

You had me at "great great thread" Sure I'll load it up, pretty sure the pics are still on the camera so I can get to it later today.

Also, welcome to forums hachre :thumbsup:

Thanks!! Your thread got me in

 

[Andle Riddum]

Guys at Puget Systems gave it a shot. Seems that temperature drop from changing the TIM might not be as drastic as reported on forums. Culprit could be a bad application the stock TIM during manufacturing. And somehow they managed to damage a cpu (even with care). . So in their opinion it might be not worth the risk.

http://www.pugetsystems.com/labs/articles/Ivy-Bridge-CPU-TIM-Paste-Replacement-160

 

[Ferzerp]

Guys at Puget Systems gave it a shot. Seems that temperature drop from changing the TIM might not be as drastic as reported on forums. Culprit could be a bad application the stock TIM during manufacturing. And somehow they managed to damage a cpu (even with care). . So in their opinion it might be not worth the risk.



http://www.pugetsystems.com/labs/articles/Ivy-Bridge-CPU-TIM-Paste-Replacement-160

Or they did a really poor job. Unless you think we're all just lying?

 

[BonzaiDuck]

Guys at Puget Systems gave it a shot. Seems that temperature drop from changing the TIM might not be as drastic as reported on forums. Culprit could be a bad application the stock TIM during manufacturing. And somehow they managed to damage a cpu (even with care). . So in their opinion it might be not worth the risk.

http://www.pugetsystems.com/labs/articles/Ivy-Bridge-CPU-TIM-Paste-Replacement-160

First, I'm not familiar with GELID, although the name sounds familiar. If the formulation featured use of substances that drastically reduce thermal resistance, they would've mentioned it in their promotions.

Second, with a "so-so" thermal grease/paste that probably fits in the same category, IDontCare and some others (different forums, etc.) showed reductions significantly greater than 10C (or less -- if I read the Puget Systems article correctly).

Third, I could be totally wrong about this, but perhaps it's possible that the improvement is more evident in over-clocking examples that push the thermal wattage somewhat higher.

For the latter, I don't rightly know, and could even suspect myself of sloppy thinking.

But the issue is not only that "some" processors may have TIM inadequately applied, but that all the processors use a sub-grade TIM.

Not trying to swing your conclusion or thought in either direction, I'm only dropping by here frequently to "wait and see."

PS PUget Systems sells parts, including CPUs. I didn't find any Ivy Bridge units among the CPUs they sell. But if they sold IB's, it would certainly be in their interest to discourage people from removing the IHS. They also used Permatex to reapply the IHS, suggesting they MAY actually anticipate people with damaged CPUs trying to RMA carefully "repaired" processors -- pawning them off as "returned in original condition."

There, I'm speculating, but theirs is just one of several demonstration projects we've seen. It also seems to be the one with the worst results.

 

[ehume]

Gelid GC Extreme is one of the best TIM's for heatsinks: see this. It has low viscosity, which makes it ideal for heatsinks but maybe not for this application.

 

[hachre]

That's why I'm very curious about IDCs TIM-above-and-under-IHS tests

 

[BonzaiDuck]

Gelid GC Extreme is one of the best TIM's for heatsinks: see this. It has low viscosity, which makes it ideal for heatsinks but maybe not for this application.

I think what we're looking for is a paste that dries into something as rubbery as bathroom tile-caulk.

I don't want to elaborate on the TIM comparisons much further, since I've already thrown in my own two-cents-worth in earlier posts. Usually, I'm confident about Benchmark Reviews.

Here, they seem to contradict what I know from Innovative Cooling (as I recall . . .), saying that it contains Aluminum Oxide in addition to carbon (diamond). My understanding was that ICD7 is something like 90% micronized diamond with certain "oils" or a base that assures a paste instead of powder. Then there's their comparison with AS5. It's completely reversed from my own tests and those of other comparison reviews.

Of course, I haven't mentioned GELID, but to the point -- if other data in the table are suspect, then . . . well . . . . gee . . . . I can't trust the table per everything else . . .

 

[ehume]

I think what we're looking for is a paste that dries into something as rubbery as bathroom tile-caulk.

I don't want to elaborate on the TIM comparisons much further, since I've already thrown in my own two-cents-worth in earlier posts. Usually, I'm confident about Benchmark Reviews.

Here, they seem to contradict what I know from Innovative Cooling (as I recall . . .), saying that it contains Aluminum Oxide in addition to carbon (diamond). My understanding was that ICD7 is something like 90% micronized diamond with certain "oils" or a base that assures a paste instead of powder. Then there's their comparison with AS5. It's completely reversed from my own tests and those of other comparison reviews.

Of course, I haven't mentioned GELID, but to the point -- if other data in the table are suspect, then . . . well . . . . gee . . . . I can't trust the table per everything else . . .

I just pulled my Megahalems after 10-month run. The Gelid GC Extreme was still soft and gloppy -- just as it had been when I put it on. It's excellent stuff, and after trying several of the top-rated TIM's I settled on just using the GC Extreme.

I only mention this because either there is something about low viscosity TIM's that is unsuited for direct heatsink-to-cpu contact, or there is something else about their testing that makes their results differ from everyone else's.

IMO the TIM is a blind alley.

 

[BonzaiDuck]

IMO the TIM is a blind alley.

Just MHO, but there are already too many forum "project" threads elsewhere, and IDC's, FerZerp's, and others here which suggest otherwise.

It "may be" that there is variation in Intel's own TIM application, but there are much better TIMs than Intel's.

 

[pwoz]

3 weeks in temps are back to pre-mod level with IC diamond. Hopefully they stop there and I don't have to reapply like Ferzerp, but it's not looking good.

4th week in, temps are same as week 3. That is to say, the same as before the mod. Hoping it is stabilized here. I'll check again next week.

 

[BonzaiDuck]

4th week in, temps are same as week 3. That is to say, the same as before the mod. Hoping it is stabilized here. I'll check again next week.

I won't tilt you toward action either way. If they were low after the ICD application, then something happened to degrade the performance, but it wasn't a degradation of the TIM itself, but degraded contact. Diamond paste -- as a substance -- doesn't deteriorate.

Let's see what some of the others say. I may have said -- a more liberal re-application may show a certain "permanence" and reliability, but . . . well . . . let's see . . .

For me, the plan to spend extra money to convert a Wolfdale box to an IB Z77 build may soon get the best of me -- to spend the approximate $670 on the parts. I just want to see "on paper" that my tenant on the east coast has renewed the lease and that the air-conditioner repair will be the promised $600 -- not the worst case of $3K. After that, I can see letting go of the money just to have a new toy (as if I don't have enough . . . )