Ivy Bridge's heat problem is indeed caused by Intel's TIM choice

[Ferzerp]

Got my new mount in. IBT max is now 67C, P95 max is 62C@1.24V (per cpu-z) and 4.4Ghz (I had to up my voltage a bit, I was losing GFLOPs in IBT)

I'm using NT-H1 TIM in both spots. I'm about to apply my IX once I study the process for an H100

 

[Lepton87]

Got my new mount in. IBT max is now 67C, P95 max is 62C@1.24V (per cpu-z) and 4.4Ghz (I had to up my voltage a bit, I was losing GFLOPs in IBT)

I'm using NT-H1 TIM in both spots. I'm about to apply my IX once I study the process for an H100

I don't have time to read the whole thread again, so how much of a reduction in temp are we talking about? i.e what was the temperature before?

 

[Ferzerp]

I don't have time to read the whole thread again, so how much of a reduction in temp are we talking about? i.e what was the temperature before?

I don't have good before numbers, but initially I was in the low-mid 80's at those settings. So, conservatively, 10-15C.


Sadly, I broke my H100 mount before starting this, so my before numbers are purely from memory.

 

[RampantAndroid]

I don't have good before numbers, but initially I was in the low-mid 80's at those settings. So, conservatively, 10-15C.


Sadly, I broke my H100 mount before starting this, so my before numbers are purely from memory.

Wait, I'm confused. What did you do to get a 10-15C drop? Is IX indigo xtreme? Can it really drop temps THAT much?!

 

[SickBeast]

Wait, I'm confused. What did you do to get a 10-15C drop? Is IX indigo xtreme? Can it really drop temps THAT much?!

It would actually co-inside with some other results out there.

 

[The|Hunter]

You expect them to comment? No offense meant, but good joke

We will get no more than the thermal density response, given the CPUs work at stock.

well it would be fair



But then again they make a ton of money, i bet they made this lame design on purpose. They said nope, no 5ghz for you xD

 

[Ferzerp]

Wait, I'm confused. What did you do to get a 10-15C drop? Is IX indigo xtreme? Can it really drop temps THAT much?!

I did the following:

Removed the IHS, removed all the adhesive, sanded the IHS to make it fit more snug (removed some space between die and IHS). lapped the IHS on the IHS-Heatsink side. Lapped my H100

Put Noctua NT-H1 between the die and IHS, put IX between the H100 and the IHS.

This let me reduce my voltage from better cooling (trying offset voltage now, and only at 1.208-1.224V, before for true stability I needed 1.24V which compounds these latest numbers in this post)

I just ran IBT at 4.4Ghz and my peak temp was 63C. So far in P95, 62C is my peak, but it's only been running 15 minutes.

edit: to correct voltage max

 

[Rvenger]

I had to go on a thermal paste run. I found some Antec Nano Diamond 7 at staples, it is expensive @ $15 but whatever. Dropped it in my Bulldozer and it cut my temps down 6c idle and about 8c load compared to Ceramique 2. I wonder what this stuff will do under the IHS.

 

[RampantAndroid]

I did the following:

Removed the IHS, removed all the adhesive, sanded the IHS to make it fit more snug (removed some space between die and IHS). lapped the IHS on the IHS-Heatsink side. Lapped my H100

Put Noctua NT-H1 between the die and IHS, put IX between the H100 and the IHS.

This let me reduce my voltage from better cooling (trying offset voltage now, and only at 1.208-1.224V, before for true stability I needed 1.24V which compounds these latest numbers in this post)

I just ran IBT at 4.4Ghz and my peak temp was 63C. So far in P95, 62C is my peak, but it's only been running 15 minutes.

edit: to correct voltage max

Forgive me while I sit here with my mouth hanging open. Jeez! If only I was willing to take the lid off my 3770k...

 

[Ferzerp]

Forgive me while I sit here with my mouth hanging open. Jeez! If only I was willing to take the lid off my 3770k...

I have to admit, I was lax in the lapping. No mirror finishes for me. Just enough to make sure it's smooth and flat.

Let's just say that my H100 has kept it cool at every voltage I've been willing to throw at it. (4.7Ghz@1.4V does hit 80C still in IBT as a peak though)

I kind of have a lemon as far as OC'ing goes. It needs that much voltage just for 4.7.

 

[Blades]

I have to admit, I was lax in the lapping. No mirror finishes for me. Just enough to make sure it's smooth and flat.

Let's just say that my H100 has kept it cool at every voltage I've been willing to throw at it. (4.7Ghz@1.4V does hit 80C still in IBT as a peak though)

I kind of have a lemon as far as OC'ing goes. It needs that much voltage just for 4.7.

What grit sandpaper? 2K or 3K?

 

[Ferzerp]

I went up to 2500 on the IHS (but didn't spend enough time to make it mirrored, really should have probably still been on 1500), but only 1500 on the H100.

I got tired of it to be honest.

 

[mdlam]

This is a great way to return a bum CPU, just swap out the IHS haha. criminal.

 

[MustangSVT]

A guy on HardForum has done it properly and posted some photos:

http://hardforum.com/showthread.php?t=1693361

thanks! going there now!

 

[MrTeal]

I had to go on a thermal paste run. I found some Antec Nano Diamond 7 at staples, it is expensive @ $15 but whatever. Dropped it in my Bulldozer and it cut my temps down 6c idle and about 8c load compared to Ceramique 2. I wonder what this stuff will do under the IHS.

No offense, but I have a very hard time believing that difference is due to just the TIM. Perhaps you did a much better job of applying it and mounting the heatsink, or different ambient temperatures or fan speeds?

As a reference, I have a roll of this stuff at home.
http://www.digikey.ca/scripts/DkSearch/dksus.dll?x=0&y=0&lang=en&keywords=3M10321-ND
It's a 0.5mm thick thermal pad with 3.0W/mK thermal resistivity, so pretty average performance. An AM2/AM3 CPU heat spreader is 37.5mm square, or 0.0014m^2. I could put one of those pads on the IHS, and the thermal resistance would be 0.12C/W. If I started with some perfect TIM with no difference in temperature between the IHS and heat sink and then replaced it with a thermal pad, I'd expect about a 12C difference in temperature between the IHS and HS with a 100W load.

 

[ArchAngel777]

No offense, but I have a very hard time believing that difference is due to just the TIM. Perhaps you did a much better job of applying it and mounting the heatsink, or different ambient temperatures or fan speeds?

As a reference, I have a roll of this stuff at home.
http://www.digikey.ca/scripts/DkSearch/dksus.dll?x=0&y=0&lang=en&keywords=3M10321-ND
It's a 0.5mm thick thermal pad with 3.0W/mK thermal resistivity, so pretty average performance. An AM2/AM3 CPU heat spreader is 37.5mm square, or 0.0014m^2. I could put one of those pads on the IHS, and the thermal resistance would be 0.12C/W. If I started with some perfect TIM with no difference in temperature between the IHS and heat sink and then replaced it with a thermal pad, I'd expect about a 12C difference in temperature between the IHS and HS with a 100W load.

I am with you. IDC did some really thorough testing in another thread, I don't recall thermal paste having more than an 4c delta at load between all the different pastes (worst to best), but admittingly, I could be remembering the results incorrectly.

 

[MrTeal]

I am with you. IDC did some really thorough testing in another thread, I don't recall thermal paste having more than an 4c delta at load between all the different pastes (worst to best), but admittingly, I could be remembering the results incorrectly.

The most comprehensive review I remember seeing on the web was Benchmark Review's 80 Way shootout.
http://benchmarkreviews.com/index.php?option=com_content&task=view&id=150&Itemid=62

The whole thing is a great read, but the TL;DR version is that other than a couple outliers, the best to worst performers had a 2.5C difference between them. Included in the excellent performers are things like MG Chemicals 860, of which I have a 150g tube of that cost me $12 or so. If you have two nice flat clean surfaces and good pressure, different TIMs won't make a huge difference.

 

[KingFatty]

What struck me about seeing the underside of the IHS when people de-lidded, was how *thick* of a layer of TIM was used. You could literally see the outline of the chip formed in the TIM, it was so thick. Is there a good reason why there is such a huge gap that needs to be filled-in with TIM?

 

[Ferzerp]

What struck me about seeing the underside of the IHS when people de-lidded, was how *thick* of a layer of TIM was used. You could literally see the outline of the chip formed in the TIM, it was so thick. Is there a good reason why there is such a huge gap that needs to be filled-in with TIM?

Dunno, but the IHS is copper and pretty easy to sand down until that gap is gone!

 

[Kenmitch]

What struck me about seeing the underside of the IHS when people de-lidded, was how *thick* of a layer of TIM was used. You could literally see the outline of the chip formed in the TIM, it was so thick. Is there a good reason why there is such a huge gap that needs to be filled-in with TIM?

I watched a video last night with the removal of the IHS. It wasn't in English tho. But watching him play around with the TIM it looks like what is on the bottom of the stock intel coolers. He was rolling it into little balls and it wasn't even staining his fingers like most TIM would. If it's the same stuff as the stock coolers I could see it increasing the gap with the amount that looks to be applied in production. Maybe somebody will get around to testing the stock TIM soon. Would be interesting to see how it performs under a controlled test with known good TIM.

 

[Ferzerp]

I watched a video last night with the removal of the IHS. It wasn't in English tho. But watching him play around with the TIM it looks like what is on the bottom of the stock intel coolers. He was rolling it into little balls and it wasn't even staining his fingers like most TIM would. If it's the same stuff as the stock coolers I could see it increasing the gap with the amount that looks to be applied in production. Maybe somebody will get around to testing the stock TIM soon. Would be interesting to see how it performs under a controlled test with known good TIM.

What you describe from that video describes my experience with that material too.

 

[Kenmitch]

What you describe from that video describes my experience with that material too.

Not sure if this video was linked yet....But here it is just in case anybody wants to view it. Pretty much shows what needs to be done for anyody that is debating about it. Doesn't look hard at all and seems like there is more danger of dicing a finger than damaging the chip

http://www.xtremesystems.org/forums...IHS-Removals-CPU-temp-dropped-from-79C-to-71C

 

[mdlam]

Best practices for sticking the lid back on? Silicone sealant work or tape? The problem is that tape has a halfmm thickness. What did you app use?

 

[Ferzerp]

I used absolutely nothing. My IHS is held on only with the adhesiveness of my replacement TIM and the clamping pressure of the socket retainer, and the heatsink pressure.

If you go this route, be sure to keep the IHS in the proper location while clamping the bracket down, as you can do bad things if you let it slide out of place, or only clamp one side of it. (I had to sand a burr off mine when I had it aligned poorly and it got stuck in the bracket)

But hey I can't complain about the results! http://forums.anandtech.com/showthread.php?t=2246497

 

[Kenmitch]

Best practices for sticking the lid back on? Silicone sealant work or tape? The problem is that tape has a halfmm thickness. What did you app use?

In this case I'd think nothing would be better than something. The hold down and the heatsink will keep it in place anyways. You would just want to make sure it's in the correct orientation and position before locking it down. If you wanted to use something a very, very thin layer of silicone might work. Possibly a tiny dap of crazy glue in the corners? Not sure if it would eat into the pcb or not and one would have to use caution when trying it out. Guess maybe somebody that cracks the core could see what crazy glue would do to the pcb.

I used absolutely nothing. My IHS is held on only with the adhesiveness of my replacement TIM and the clamping pressure of the socket retainer, and the heatsink pressure.

If you go this route, be sure to keep the IHS in the proper location while clamping the bracket down, as you can do bad things if you let it slide out of place, or only clamp one side of it. (I had to sand a burr off mine when I had it aligned poorly and it got stuck in the bracket)

But hey I can't complain about the results! http://forums.anandtech.com/showthread.php?t=2246497

Was tying while the above reply was made. But like he stated nothing would be best as it will not increase the gap.