Removed the IHS from an Ivy Bridge i5 3570k

it would be nice if they made higher quality aftermarket cpu covers.
 
Last edited:
Just looks like to me that Intel did not use a great TIM between the naked core and the IHS. 10C temp drop by using Arctic Silver is huge.
 
Artic Silver 5 Thermal Conductivity: 8.7 W/(m·K)

50In50Sn Thermal Conductivity 74.1W/(m-K)

Soldered wins hand down by a a factor of 8.5x
 
M8R-mx8r56 said:
I registered just to let everyone know about the following thread. :D

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

Check out the last page in that thread. It seems the guy managed to get direct die liquid cooling to work. And he managed to overclock his i7 3770k to 5ghz.
Click to expand...



With ambient cooling that chip is probably pushing its max clock speed. 10C drop on temps is nice but its not nearly enough drop to let the IB chip crank up on the speed.

On the other hand IB can run 4.5-4.8GHZ depending on chip quality no problem with low voltage and much better temps.

Delidding doesn't hardly seem worth it just to get another 100-200mhz in speed.


Intel needs to improve stepping for better over clocking results.
 
DigitalGriffin said:
Artic Silver 5 Thermal Conductivity: 8.7 W/(m·K)

50In50Sn Thermal Conductivity 74.1W/(m-K)

Soldered wins hand down by a a factor of 8.5x
Click to expand...

I don't know how you got that number for 50In50Sn. Last time I checked from Indium Corp, their Indalloy 1E ( 52In48Sn) is only 34W/(mK). Their Indalloy 4 (99.99% Indium) is around 86 and Indalloy 290 (97% In and 3% Ag) is 73. Given the price for a 3-foot wire of $95 regardless of the alloy's composition, this is not a cheap mod.

I am thinking if I ever decide to do one of these, I would heat up the IHS and the wire on a laboratory hot plate. While it is cooling off, I can just attach the wafer/chip on top, and voila.

Edit: A quick search on eBay does show much cheaper alternatives. I am tempted now.
 
Both surfaces need to be heated for the solder to bond. Your method would result in a 'cold' joint which would be worse than the Tim.
 
Seems like such an easy mod to cool down up to 20C.

I would be all over this mod if Intel would still honor the warranty.
 
Fuck this shit,

I'm trying this! mainly due to the fact of just not being able to reach 5Ghz stable on air with decent temps.

Just need to find some real sharp thin knife.
I'm also wondering if I could use some Indigo Xtreme I still happen to have lying around, I'd need to get it off of the plastic plate somehow.
Or maybe its just a better idea to get some coolaboratory liquid pro..
 
It looks to me like the best results including the IHS have been with Coollaboratory Liquid Pro.

They're not bonding the IHS to the die. The IHS is not "glued." It seems that they are simply fitting the IHS back on carefully after having applied their choice of thermal interface material, and allowing the pressure of the mount to secure the installation.

I do wonder whether indigo Xtreme would work between the IHS and the die, but CLLP is much less expensive, and easier to install.

As for running the die bare: I'm more comfortable with the protection afforded by an IHS.

It could easily be that the baseplates are not optimally thick on most of our modern heatsinks and waterblocks because they were designed to cool the IHS and not the bare die. Given a thicker base, perhaps even one that is "stepped" (essentially, machined to integrate an IHS, perhaps up to a 1mm protrusion) for this purpose, the bare-die option should perform better given the same thermal interface.

Hmm...maybe someone could ask DT waterblocks whether or not a special, stepped edition of the DT 5noz would improve the situation. I figure that'd be a $20-$50 premium ... more copper, custom machining, small batch size, and optimized specifically for Ivy Bridge...sounds like a bad plan eh?

Even then, I'd prefer some protection around the die to prevent chipping / cracking caused by uneven mounting. Back when AMD was selling bare-die CPUs, we were using little dots of dense clay at the four corners. A similar approach might be a good idea for those so determined.

Heh - I briefly scanned through some of the threads on this. It looks like people were doubting that the TIM had anything to do with the increased temps...I'm not going to review the evidence for those beliefs because now it seems we have enough data to conclude that the TIM does indeed make a large difference.
 
Last edited:
Ok i see you guy's struggling with the temperatures. Here are my results.

5 Day's old AS5 stock IHS:
Room Temperature 24,2c:
Core 0: 75c
Core 1: 84c
Core 2: 84c
Core 3: 78c
Package: 87c

2x MX4 Paste:
Room Temperature 24,2c:
Core 0: 63c
Core 1: 73c
Core 2: 73c
Core 3: 70c
Package: 75c

Profit:
Core 0: 12c
Core 1: 11c
Core 2: 11c
Core 3: 8c
Package: 12c

The gain was quite good over here :) i'm using a Thermalrigh Archon CPU Cooler. And was running my CPU @ 4,5 GHz 1.25v.
 
Impressive gains, I'm not personally willing to attempt this and am happy with my clocks, but hats off to those that do, this is [H] afterall.
 
I just ordered some liquid pro. I hate you guys. I was perfectly content before reading this thread over and over. Think I'll go AS5 between die and IHS as liquid pro for hsf mounting.
 
im curious to see how much of a gain there is to have with just something along the lines of AS5 (or equivalent) with a lapped Surface.
 
RamonGTP said:
I just ordered some liquid pro. I hate you guys. I was perfectly content before reading this thread over and over. Think I'll go AS5 between die and IHS as liquid pro for hsf mounting.
Click to expand...

No, use the liquid pro between the core and the IHS. Otherwise why even bother with it? It has much higher thermal conductivity than AS5.
 
Forceman said:
No, use the liquid pro between the core and the IHS. Otherwise why even bother with it? It has much higher thermal conductivity than AS5.
Click to expand...

I thought about that too. My only fear is that the application of liquid pro calls for spreading a very thin layer, versus a dab of AS5 and relying on pressure to spread it. I'm afraid that there may not be physical contact with the IHS
 
RamonGTP said:
I thought about that too. My only fear is that the application of liquid pro calls for spreading a very thin layer, versus a dab of AS5 and relying on pressure to spread it. I'm afraid that there may not be physical contact with the IHS
Click to expand...

You came to the same exact conclusion that I did. Also, I feared that if I put enough on to make contact, over time, it might flow and cause all sorts of issues.

You can sand down the IHS so that it makes better contact though. Pretend you're lapping the IHS, but instead flip it over. Anything you remove there will make it sit closer to the die. I did this and don't really have a gap anymore, but I'm still afraid of a liquid TIM. Especially after I had random issues with the topmost core suddenly growing very hot (as in 19C higher than it previously had) and had to take everything apart (ruining my IX) and redo it all. The computer sat, unmoved for 2-3 weeks and then I decided to make sure all was well and noted that one core the temps were completely out of whack from what they had been.
 
ferzerp said:
...but I'm still afraid of a liquid TIM. Especially after I had random issues with the topmost core suddenly growing very hot (as in 19C higher than it previously had) and had to take everything apart (ruining my IX) and redo it all. The computer sat, unmoved for 2-3 weeks and then I decided to make sure all was well and noted that one core the temps were completely out of whack from what they had been.
Click to expand...

Under what conditions did this occur?

Were you using a liquid metal TIM like CLLP? And then specifically between the die and IHS?

Or was this just a general experience with liquid metal TIMs?

I noticed that you mentioned IX, so I assume this refers to a Die/CLLP/IHS/IX/Heatsink configuration ?

If so, then that blows :(

If your IX simply failed... well that also blows, but that would appear to be an isolated case, no?

Regardless - people are getting 8-12C improvements at overclocked load using just traditional pastes. Go for it.
 
Archmage said:
Under what conditions did this occur?

Were you using a liquid metal TIM like CLLP? And then specifically between the die and IHS?

Or was this just a general experience with liquid metal TIMs?

I noticed that you mentioned IX, so I assume this refers to a Die/CLLP/IHS/IX/Heatsink configuration ?

If so, then that blows :(

If your IX simply failed... well that also blows, but that would appear to be an isolated case, no?

Regardless - people are getting 8-12C improvements at overclocked load using just traditional pastes. Go for it.
Click to expand...

No, I didn't go with liquid anywhere. It was die-NTH1-IHS-IX-H100. I'm not sure what the cause of the failure after a few weeks. I was just saying I was afraid of something liquid when the gap was as large as it is. Right now I'm running it all with NT-H1 in both spots for a while. I don't want to waste my last IX if it's just going to mess up again.
 
Wait just a damn minute! Is this a thread attempting to confirm that taking the IHS off and using better TIM with the cooler having direct contact with the chip itself makes a difference in cooling Ivy Bridge!? That maybe Intel did in fact go cheap and it actually does matter that they chose not to use solder?

No this cannot be! Intel would never do such a thing as go cheap and they have engineers and shit. And anyway, it has been "tested" (read: proven) (because one dude somewhere on the internets did it right around launch) that removing the IHS doesn't make any difference. It's just the chip itself, shit is too dense and stuff and there is nothing really to be done. The thing is just hot and stuff and soldering the IHS onto the chip wouldn't have made any difference and Intel and IB are the BESTEST!

Were you all not here when that was decided and anyone who disagreed was ridiculed? Where are "snoops" et al. when you need them to clarify things?
 
just curious, has anyone tried Arctic Silver Thermal Adhesive? Wonder how well that would work.
 
beto said:
Wait just a damn minute! Is this a thread attempting to confirm that taking the IHS off and using better TIM with the cooler having direct contact with the chip itself makes a difference in cooling Ivy Bridge!? That maybe Intel did in fact go cheap and it actually does matter that they chose not to use solder?

No this cannot be! Intel would never do such a thing as go cheap and they have engineers and shit. And anyway, it has been "tested" (read: proven) (because one dude somewhere on the internets did it right around launch) that removing the IHS doesn't make any difference. It's just the chip itself, shit is too dense and stuff and there is nothing really to be done. The thing is just hot and stuff and soldering the IHS onto the chip wouldn't have made any difference and Intel and IB are the BESTEST!

Were you all not here when that was decided and anyone who disagreed was ridiculed? Where are "snoops" et al. when you need them to clarify things?
Click to expand...

Don't know, but unless you attach it with something with extremely low thermal resistance (like the solder on SB), there is physically no way a "heat spreader" assists cooling compared to a properly designed cooler directly on the die (and even then, if you could solder the cooler like the SB IHS, the direct die cooling would still be better).

You might have a heatsink that doesn't work on the smaller die due to being designed for a different type of contact (especially the heat pipe types), but anyone who says adding an extra layer of high thermal resistance between your cooler and the die is beneficial just doesn't understand physics. They may be able to type a lot and ridicule, but they exhibit a fundamental lack of understanding. If "spreading" heat by placing a mini heat-sink with poor TIM (as is the case in IB) was beneficial, then adding another layer would be even better. Why we should all make layer after layer of slightly wider plates of metal and TIM before finally putting on a traditional cooler. Man that heat will really be spread out by then!

Of course, anyone will say that's a silly notion, but it is the same "logic" that would have to be used to claim that the way IB is set up assists cooling and doesn't impede it.

Direct die cooling is hard, because getting the proper contact is hard, and not breaking the die is hard, and because you may need a slightly different cooler/block design. Because one person couldn't do it properly (and the source of that "information" was a single random forum post that was re--posted over and over until it looked like data), doesn't mean that it's a bad idea.
 
It think it really comes down to how good the thermal interface is between the core and the IHS. If you get 27 W/(m·K) between the die and a direct contact heatsink (numbers made up) but you can get 85 W/(m·K) between the die and a soldered-on IHS, and then 27 W/(m·K) between the IHS and a heatsink, you are better off in the second example because you've increased the surface area by basically making the die larger.
 
Forceman said:
It think it really comes down to how good the thermal interface is between the core and the IHS. If you get 27 W/(m·K) between the die and a direct contact heatsink (numbers made up) but you can get 85 W/(m·K) between the die and a soldered-on IHS, and then 27 W/(m·K) between the IHS and a heatsink, you are better off in the second example because you've increased the surface area by basically making the die larger.
Click to expand...

But you would always be better off with direct die cooling if you can get the same low thermal resistance, or at least comparable resistance, and in the case of IB, that's not very hard (because with a little work, we can improve it so much). I see screenshots of people using the same voltage I do and hitting mid-high 80's, while my proc never even gets to 70C. Now, an H100 is an ok cooler, but it's not a 15-18C delta from the stuff these people are using. It's a shame that my proc is such a poor OC'er.
 
Last edited:
ferzerp said:
But you would always be better off with direct die cooling if you can get the same low thermal resistance, or at least comparable resistance, and in the case of IB, that's not very hard (because with a little work, we can improve it so much). I see screenshots of people using the same voltage I do and hitting mid-high 80's, while my proc never even gets to 70C. Now, an H100 is an ok cooler, but it's not a 15-18C delta from the stuff these people are using. It's a shame that my proc is such a poor OC'er.
Click to expand...

But you can't always get the same thermal resistance - that's the point. In the Ivy case you might be able to, but there is no way you are going to get the same thermal resistance between a direct die and a heatsink that Intel is able to get with a soldered connection between the die and an IHS. So you can't say it is always better.
 
I think if you go back and read, you'll see that I only claimed it is always better when you can match or come close to the thermal resistance of whatever they use. In theory, yes, you could solder a heatsink directly on to a SB die and that would match it.

On IB, I think I could use peanut butter and meet the resistance they have :p
 
Why do you sit the IHS again ? I mean you guys took all the trouble opening it... and you sit it back on ? Why ?

IMO, direct die contact with HSF will give you better temps always. The IHS is like 3mm thick, that's some thermal resistence right there + the loss of thermal transfer due to the IHS beeing concave.
 
ShePearl said:
Why do you sit the IHS again ? I mean you guys took all the trouble opening it... and you sit it back on ? Why ?

IMO, direct die contact with HSF will give you better temps always. The IHS is like 3mm thick, that's some thermal resistence right there + the loss of thermal transfer due to the IHS beeing concave.
Click to expand...

Your question has been answered a while back. If you don't put it back on, you have to get creative on how you're going to secure the CPU in the socket and ensuring physical contact between the HSF and die is tricky since the clearances are intended for it to be mounted to the IHS. Assuming you find a creative way to keep the CPU securely in place, and modify your HSF to make physical contact with the die, you then have the added problem of putting too much pressure on it and cracking/chipping it. Thats why you put it back on, quite simply, it's a good idea.

The removal of IHS isn't for the sport of it, its to apply a better thermal compound between it and the die, and to maybe sand it down slighly to make the gap smaller.
 
RamonGTP said:
Your question has been answered a while back. If you don't put it back on, you have to get creative on how you're going to secure the CPU in the socket and ensuring physical contact between the HSF and die is tricky since the clearances are intended for it to be mounted to the IHS. Assuming you find a creative way to keep the CPU securely in place, and modify your HSF to make physical contact with the die, you then have the added problem of putting too much pressure on it and cracking/chipping it. Thats why you put it back on, quite simply, it's a good idea.

The removal of IHS isn't for the sport of it, its to apply a better thermal compound between it and the die, and to maybe sand it down slighly to make the gap smaller.
Click to expand...
Aha. thanks for the reply. Did you overclock your 3770K ? If so, what frequency are you running it at ?
 
I have a very mild OC of 4.2GHz. I have not de-lidded my CPU yet. I have some liqid pro on the way and will delid soon with the intentions of going to 4.4-4.6GHz
 
I really admire the effort here but I just cannot see the justification in voiding the warranty of a 350 dollar processor. This is right up with pin mods and pencil tricks. Just use good cooling buy the cheap Intel warranty they offer for ocing and crank it up! I have never burned out an Intel CPU and I ran old P4 2.4c @ almost 4 GHz and PD [email protected] they never went bad. With all the good cooling and pastes out now this just seems like it really isn’t worth the trouble. Maybe I’m getting old
 
HisSvt76 said:
I really admire the effort here but I just cannot see the justification in voiding the warranty of a 350 dollar processor. This is right up with pin mods and pencil tricks. Just use good cooling buy the cheap Intel warranty they offer for ocing and crank it up! I have never burned out an Intel CPU and I ran old P4 2.4c @ almost 4 GHz and PD [email protected] they never went bad. With all the good cooling and pastes out now this just seems like it really isn’t worth the trouble. Maybe I’m getting old
Click to expand...

The good cooling isn't being utilized efficiently with IB, that's the problem. De-lidding fixes that. The intel warranty isn't going to help us get more performance considering the CPU will throttle at the high end of a moderate overclock with or without the warranty.
 
RamonGTP said:
The good cooling isn't being utilized efficiently with IB, that's the problem. De-lidding fixes that. The intel warranty isn't going to help us get more performance considering the CPU will throttle at the high end of a moderate overclock with or without the warranty.
Click to expand...

I really understand this but if the product is not up to the task don't buy or return it and raise hell with Intel. Money talks if it sucks don't buy it. It is like buying a mustang with a fiesta motor then swapping in a mustang motor. Just seems really stupid for the price you have to pay. To me that just makes IB a skip and hope the next one is better. But I know I used to have the PC part disease once too.
 
HisSvt76 said:
I really understand this but if the product is not up to the task don't buy or return it and raise hell with Intel. Money talks if it sucks don't buy it. It is like buying a mustang with a fiesta motor then swapping in a mustang motor. Just seems really stupid for the price you have to pay. To me that just makes IB a skip and hope the next one is better. But I know I used to have the PC part disease once too.
Click to expand...

I don't think that you do. The product doesn't suck, we just want it to be better and found a relatively easy way to accomplish that. There isn't anything better out there, so not buying it really isn't an option for people who want an upgrade to the latest and fastest tech. Intel doesn't care about overclockers so complaining to them will be a waste of time, we're a small market so it's up to us.

Your suggestion will do nothing but waste our time with complaining to Intel, and relegate us to inferior products by choosing not to buy it... I think i'll de-lid. :)
 
RamonGTP said:
I don't think that you do. The product doesn't suck, we just want it to be better and found a relatively easy way to accomplish that. There isn't anything better out there, so not buying it really isn't an option for people who want an upgrade to the latest and fastest tech. Intel doesn't care about overclockers so complaining to them will be a waste of time, we're a small market so it's up to us.

Your suggestion will do nothing but waste our time with complaining to Intel, and relegate us to inferior products by choosing not to buy it... I think i'll de-lid. :)
Click to expand...

Alright i won't say anything else but i really do get it i do this for a living as well as a hobby i just think it's a stupid thing to do. But hey if money is no object and warranty doesn't matter have at it. But don't try to rma your chip after you mess it up. bullshit warranty claims just make it worse for the regular consumer to get a legit replacement. If the product does not do what it should do and cool properly then it does suck. I love Intel but they went cheap and you guys just say "Oh well it's still the best even though they half assed part of the product" Again i admire all your effort but it should not be needed!! Esp with a INTEL product! so seriously i'm not knocking you guys but damn intel should step up and fix this .
 
Last edited:
Thanks for your input, but please dont lecture me on warranty claims. If I make an unethical warranty claim what you say here isn't going to stop me. If I dont, it won't be because of what you said.

Did intel go cheap? Yes. Do they need to fix it? Not a valid question since it isn't broken. Can they make it better? Absolutely. Will they? Not likely considering their closest competition is their own previous gen CPU and the closest to that is their own previous gen CPU again.

The product IS doing what it should. So by your own definition it doesn't suck.

You think it's stupid, fair enough. Everyone is entitled to their opinion. I think writing a letter to intel is infinitely stupid. One can perhaps delid and write a letter, at least then they did one thing that makes a difference. ;)
 
ShePearl said:
Why do you sit the IHS again ? I mean you guys took all the trouble opening it... and you sit it back on ? Why ?

IMO, direct die contact with HSF will give you better temps always. The IHS is like 3mm thick, that's some thermal resistence right there + the loss of thermal transfer due to the IHS beeing concave.
Click to expand...

they put it back on because they're not equipped to attach the heatsink directly to the die.

the IHS messes up the heat transfer, but not because of the thickness. the very existence of another material to transfer to is the problem. if the IHS was a foot thick it would work great

also, the intel heatsink spec uses non-flat surfaces because it is easier to make good connections that way
 
You must log in or register to reply here.
Back
Top