If anyone wonders if the IHS and bad TIM are why IB can't be cooled. (hint, yes)

[Ferzerp]

Well, the stock TIM was a weird hard and flaky material too. So it was thick, and also totally not what we traditionally think of as good TIM (but who knows it if is or not.)

I still have more work to do and won't have all my materials in until Wednesday at the earliest to do real tests.

Sadly, this isn't supposed to be scientific, I'm just seeing what I can do with it, and not keeping very good records.

 

[Ferzerp]

I have it down to 72C peak in IBT maximum settings (after 2 runs or around 15 minutes). 63/72/72/72

Does anyone have a 3770k and an H100 that they can set where the CPU-Z read voltage is 1.224 during the run, 44x multiplier, and 1.4625V PLL to have a comparison? I also have all 4 dimms populated and am running 1866 Corsair Vengeance memory

 

[Puppies04]

Quick question.

While people are throwing around the comment "intel uses cheap TIM" has anyone actually proved this.

Last thing I heard from intel was they changed the TIM material to make it more robust as the heat cycling was more intense on IB than SB and they needed a material that would last longer even it it had poorer conductive properties than their usual TIM. Does anyone have proof to the contrary?

 

[Kenmitch]

Quick question.

While people are throwing around the comment "intel uses cheap TIM" has anyone actually proved this.

Last thing I heard from intel was they changed the TIM material to make it more robust as the heat cycling was more intense on IB than SB and they needed a material that would last longer even it it had poorer conductive properties than their usual TIM. Does anyone have proof to the contrary?

Somebody would have to delid then scrape off the TIM and use it on another cpu to test it against a know good TIM to see.

I'd think the gap would influence overall temp greater than TIM. I guess a combo of crappy TIM and gap would be not a good thing at all tho.

 

[Rvenger]

I bet just some of the CPUs probably have a poor application of the TIM. I bet the TIM is fine but the way its being applied is the issue here.

 

[Ferzerp]

Somebody would have to delid then scrape off the TIM and use it on another cpu to test it against a know good TIM to see.

I'd think the gap would influence overall temp greater than TIM. I guess a combo of crappy TIM and gap would be not a good thing at all tho.

The TIM is really flaky and odd. I'm not sure that you could reuse it well. Some of it was goopy, but some was flaky. Not really sure what to make of it.

 

[Puppies04]

Somebody would have to delid then scrape off the TIM and use it on another cpu to test it against a know good TIM to see.

I'd think the gap would influence overall temp greater than TIM. I guess a combo of crappy TIM and gap would be not a good thing at all tho.

See once again your use of the words "crappy TIM" are what I am asking about. Do you mean it has poorer conductive ablility? If so fair enough but if what intel said was correct and they were forced to change the TIM as the existing compound would have degraded prematurly due to the increased heat cycling of IB compared to SB then it really isn't "crappy" at all just more fit for purpose.

 

[Rvenger]

The TIM is really flaky and odd. I'm not sure that you could reuse it well. Some of it was goopy, but some was flaky. Not really sure what to make of it.

Sounds like its not completely cured. Thats how the Zalman super thermal grease is. Some goopy and some flaky.

 

[Ferzerp]

Sounds like its not completely cured. Thats how the Zalman super thermal grease is. Some goopy and some flaky.

Regardless, even if it were good TIM, I don't see how it could work well when the IHS and the die have the gap that they do.

 

[Kenmitch]

See once again your use of the words "crappy TIM" are what I am asking about. Do you mean it has poorer conductive ablility? If so fair enough but if what intel said was correct and they were forced to change the TIM as the existing compound would have degraded prematurly due to the increased heat cycling of IB compared to SB then it really isn't "crappy" at all just more fit for purpose.

I'm not saying Intel is using crappy TIM. Just that to see if the TIM is crappy or not somebody would have to use it and test it against good TIM to see.

I'm not gonna buy a IB and tear off the heatspreader and try it out. I just figured maybe somebody that goes tru the trouble could salvage enough of the TIM to try it and see is all

Regardless, even if it were good TIM, I don't see how it could work well when the IHS and the die have the gap that they do.

I would think the gap plays a greater part also. If a person goes thru the trouble to delid I see no reason why they wouldn't take the xtra time to minimize the gap at the same time. I wouldn't think it would take too much sanding on the heatspreaders mating surface at all.

 

[Ferzerp]

I'm not saying Intel is using crappy TIM. Just that to see if the TIM is crappy or not somebody would have to use it and test it against good TIM to see.

I'm not gonna buy a IB and tear off the heatspreader and try it out. I just figured maybe somebody that goes tru the trouble could salvage enough of the TIM to try it and see is all



I would think the gap plays a greater part also. If a person goes thru the trouble to delid I see no reason why they wouldn't take the xtra time to minimize the gap at the same time. I wouldn't think it would take too much sanding on the heatspreaders mating surface at all.

I'm going to pick up some coarser paper for material removal instead of smoothing and see what I can do.

I am tempted to just sand it down so far that I know it will be mated to the die, but my concern is the pressure that would then be exerted, so I may be more conservative.

 

[Kenmitch]

I'm going to pick up some coarser paper for material removal instead of smoothing and see what I can do.

I am tempted to just sand it down so far that I know it will be mated to the die, but my concern is the pressure that would then be exerted, so I may be more conservative.

If it was me I'd be on the conservative side. Maybe use a tiny dab of TIM without any sanding first to see the flow. Then sand a little bit and check again.

What's Intel spec on heatsink pressure? Around 120lbs or so?

I wouldn't suggest using a 120lb weight for testing the fit tho

 

[Ferzerp]

You can see a little more of the progress I went through with various things I tried and when I decided to start closing the gap between the IHS and the die in this thread: http://forums.anandtech.com/showthread.php?t=2245260 (which is a reference to: http://pc.watch.impress.co.jp/docs/column/sebuncha/20120511_532119.html) The image that they took when they took theirs off suggests better mating than I saw. I wish I'd taken pictures, sadly. But I had no real desire to document this stuff when I started, but people started asking for results before I'd finished (still have to get new TIM which arrives today, and a new bracket, slated for Wednesday)

I think I'll keep posting in this one because I derailed the other one already enough.

 

[Ferzerp]

If it was me I'd be on the conservative side. Maybe use a tiny dab of TIM without any sanding first to see the flow. Then sand a little bit and check again.

What's Intel spec on heatsink pressure? Around 120lbs or so?

I wouldn't suggest using a 120lb weight for testing the fit tho

I'm not sure. The IHS not only takes the heatsink's pressure, but also the socket clamp's pressure as well.

 

[Kenmitch]

The link from Japan has some good photos.

If I was to undertake such a project I'd remove the IHS and measure the center thickness above the core for a ref point. Clean up both mating surfaces. Cut out the center of the IHS. Use a copper shim about 10/1000ths or so thicker than the IHS to eliminate all the variables and most likely get better contact on the die and the heatsink/water block. Without having a IHS off the die I'm not sure how hard it would be to cut or have cut tho.

 

[Avalon]

Got the lid off!!!

I completely agree with you Ferzerp, the TIM under the IHS is horrid. It reminds me of the crap that was on my Radeon 4850 that I removed and added MX2 to. GPU dropped like 15C. Not expecting that sort of miracle here, but any improvement would be cool!

 

[Kenmitch]

Got the lid off!!!

I completely agree with you Ferzerp, the TIM under the IHS is horrid. It reminds me of the crap that was on my Radeon 4850 that I removed and added MX2 to. GPU dropped like 15C. Not expecting that sort of miracle here, but any improvement would be cool!

Got another rig you could test the Intel TIM with? From images it looks there would be more than enough to salvage for a test.

 

[Avalon]

Well, I'm 10 mins into prime95 large FFT running 4.2Ghz @ 1.12V...my prior temps would peak 70-71C. Now my peak is 64-65C. Nothing earth shattering, although I didn't do the greatest job applying my MX2. I'm sure a good application and a better TIM would result in another 3-5C drop. I did NOT sand off the rubber bond residue from the IHS as Ferzerp did, so that also could be contributing. I literally just took off the IHS, cleaned it and the die off with 91% isopropyl, applied a fine line of MX2 to die, sat the IHS back on, put it in the socket, applied a fine line of MX2 to the IHS, and applied the heatsink.

Also I won't be able to salvage the Intel TIM, I kind of scraped it off down the sink, sorry :\

 

[Ferzerp]

Yay, IX and NT-H1 arrived. Now I just need my replacement bracket.

 

[Kenmitch]

Also I won't be able to salvage the Intel TIM, I kind of scraped it off down the sink, sorry

Maybe another brave soul will read this thread before trying it. I'm sure many members would be interested in the results.

 

[Denithor]

Would Intel have made the switch from solder (which should have excellent thermal transfer properties, so long as good contact between core and IHS is maintained) to TIM just for reduced cost?

I mean, we're seeing like an 8-10C drop in temps just from applying decent TIM, not even lapping or anything. Seems like they really screwed the pooch on this one.

Or maybe they want Haswell to look really good by comparison (switch back to solder...).

Anyone planning to pop their IHS and go naked with one of these chips? IDC?

 

[Ferzerp]

Maybe another brave soul will read this thread before trying it. I'm sure many members would be interested in the results.

Sorry Kenmitch The concept of reusing TIM never dawned on me.

 

[bononos]

Would Intel have made the switch from solder (which should have excellent thermal transfer properties, so long as good contact between core and IHS is maintained) to TIM just for reduced cost?

I mean, we're seeing like an 8-10C drop in temps just from applying decent TIM, not even lapping or anything. Seems like they really screwed the pooch on this one.

Or maybe they want Haswell to look really good by comparison (switch back to solder...).

Anyone planning to pop their IHS and go naked with one of these chips? IDC?

I don't think reduced cost is the reason here. If I were suspicious, I would think maybe Intel wants to deter the casual overclockers from reaching 5Ghz that easily to make upgrading to Haswells more compelling.

 

[Ferzerp]

More info, still with the unfinished product.

In the AT article: http://www.anandtech.com/show/5763/undervolting-and-overclocking-on-ivy-bridge

On their graph, with the intel cooler, and povray (benchmark I guess?) My settings (since I'm using 4.4Ghz, 1.224V) would put the temp right around 83-85C. I ran it and my temp peaks at 64C. There is an H100 vs. Intel liquid cooler difference here, but I don't know how much of that ~20C difference it accounts for.

 

[MacGyverSG1]

It sucks that there are so many variables to consider when comparing temps. Even using different motherboards can have an effect.