Ivy Bridge Temperatures – It’s Gettin’ Hot in Here

Interesting editorial by I.M.O.G. ... worth a read!

Why is Ivy Bridge so hot? Ask that question in any forum currently, and you are likely to receive one of two different popular (but not entirely correct) answers that everyone has been parroting:

  • "Power density is greater on Ivy Bridge than Sandy Bridge"
  • "Intel has problems with tri-gate/22nm"

The first answer is correct, but wrong at the same time – power density is greater, but it isn’t what is causing temperatures to be as much as 20 °C higher on Ivy Bridge compared to Sandy Bridge when overclocked. The second answer is jumping to conclusions without sufficient evidence. If you aren’t in the loop, there’s evidence of a considerable temperature difference nearly everywhere you look – we confirmed it by mirroring settings in our Ivy Bridge review, and we have read similar reports in solid testing at Anandtech as well as from other sites.

So why is Ivy Bridge hot?

Intel is using TIM paste between the Integrated Heat Spreader (IHS) and the CPU die on Ivy Bridge chips, instead of fluxless solder.

...


38

Germany Masterchief79 says:

Ouch, someone damaged that chip pretty badly^^

To the news, simple question: WHY?

TaPaKaH says:

deliberate cripple to prop up Sandy sales? :)

now I want to know - if you remove the IHS, do temperatures drop significantly?

United Kingdom borandi says:

Cripple Ivy so next revision of chips uses the solder? New ideal chip for daily use all over again.

United States Redwoodz says:

maybe cold bug fix?

Canada Vinster says:

Sam OCX said: deliberate cripple to prop up Sandy sales? :)

now I want to know - if you remove the IHS, do temperatures drop significantly?


Wondering the same thing.

Vin

K404 says:

Until someone does a With & without IHS comparison on temperatures and MHz, this is FUD and this story is popping up all over the internet. It's just a theory....

United States sin0822 says:

borandi said: Cripple Ivy so next revision of chips uses the solder? New ideal chip for daily use all over again.


IMo i think i agree with you that Intel crippled Ivy with its TDp rating and launch delay and pricing and this.

BTW who wants to take off their IHS first? LOL

that is P1T1's picture right?

K404 says:

Intel could have crippled Ivy with an invisible multi cap. This is too complicated. Why cripple chips in a way that makes them, theoretically, more likely to fail?

United States Hondacity says:

Sam OCX said: deliberate cripple to prop up Sandy sales? :)

now I want to know - if you remove the IHS, do temperatures drop significantly?


I will lap my ivy if she is 6.7ghz :)

United States sin0822 says:

K404 said: Intel could have crippled Ivy with an invisible multi cap. This is too complicated. Why cripple chips in a way that makes them, theoretically, more likely to fail?


not cripple for us, but for the general public, do you think it really would hurt sales if they put multi max at 57x? lol

United States I.M.O.G. says:

sin0822 said: that is P1T1's picture right?


Correct, its referenced as such in the article, and used with permission.

K404 said: Until someone does a With & without IHS comparison on temperatures and MHz, this is FUD and this story is popping up all over the internet. It's just a theory....


Agree its a theory, one that is tangible which I think makes it interesting considering the other possible explanations. FUD, I don't think so - the heat issue is there in most tests it seems, this is just one way to explain it that no one had been so much as even looking at previously, let alone mentioning. I would hope the article shouldn't inspire FUD - the facts are the facts on the temps - this is just another potential explanation for what we are seeing in the communities.

FWIW, at stock TDP W/mm^2 is .439 for SB, .481 for IB - power density isn't that different. There are tests showing some OC'd/loaded wattage for SB/IB in some reviews, but I don't want to talk about that as those numbers may not be very accurate in how they were obtained (if they are accurate, the power density comes out roughly the same anyways).

There will be problems actually testing the theory. First, can't test the soldered indium IHS on IB, because Intel doesn't appear to be making them that way as far as anyone has found so far. Second, if you remove the IHS, you may need to account for clearance with the clamp/socket as the die sits lower than the IHS. Third, if you remove the IHS, you may need to do custom mounting retention to ensure the pressure is the same/similar - very hard to do accurately. Fourth, in order for the test to be good it depends on the cooler - older coolers used to be designed to cool dies directly, current ones are designed to cool IHS and not for direct die cooling (especially waterblocks, compare storm versus modern block). Fifth, these numbers are getting high and I'm having trouble counting... There are many issues with someone actually doing good/quality tests on this. Some may try, but a test is only as good as the methodology and control.

That said, I would expect a slight improvement by cooling the die directly even with a cooler not designed for cooling the die directly. I think if Intel would be using solder under the IHS, that could be more ideal in light of currently available cooling solutions. Solder IHS was good for E6XXX when TIM paste sucked for E4XXX; solder IHS was good for SB and maybe TIM paste sucks for IB.

Just ideas, and I thought the topic was interesting. Certainly got people talking. I have some other ideas now about why Intel did it, but none as nefarious as some other people are thinking... I think its silly to think they did it to cripple anything.

Canada Vinster says:

IMOG, the logic you have I think is sound imo, and I think you're right, there is no right way to compare all these results given the variances that could be present. this is a tricky one to say the least. but I hope that the TIM version is for the ES versions and that once the testing starts on the production version they will be soldered. Only time will tell. Vin

K404 says:

People have been removing IHS for years and successfully running their chips that way. IMO, that removes reasons 2,3 & 4. Reason 5 comes down to who you trust. We trust all these review site to get meaningful numbers that tell us the chips are hot in the first place. Reason 1 is true, but we don't need to re-create that scenario. Remove Ivy IHS. Get "tweaked" cooling system mounted that the user is satisfied is safe and even. Check temperatures. Have the temperatures dropped by any noticeable amount? Yes? Thermal path through IHS is to blame. No? Thermal paste is NOT the problem, options go back to process and thermal diode. Could it be mis-calibrated? Reviewers have checked that high temps are matched by a hot heatsink, right? :) @Sin, you're right. x57 is too much. If Intel don't want SB-E to look ridiculous (oops, too late) cap the multi at x45. Could even offset the thermal diode by +30 degrees so thermal protection hits early. Intel don't even need to offer massive unlocked multi anyway if we're so unimportant. Could almost accuse Intel are falling between stools.

United States I.M.O.G. says:

K404: I agree with your thinking generally. However, my concern is kinda demonstrated best with an example. Do you trust ANY of the thermal paste roundup testing done in "reviews" anywhere? In my opinion, no one should. The testing was faulty in every case, though I respect the effort and think it should be commended - ultimately the results are misleading when people interpret them however. Thermal tests involving the quality of the interface are very difficult - no good way to evaluate/analyze/report bond line thickness, total wetted area, corresponding wetted area, forget about actually cleaning out residue on a microscopic level so every sample is tainted after you initially tint the surface, etc, etc. These are issues which make respected industry laboratory testing inconsistent and incomparable from one lab's result to the next. A lot of unknowns and important factors that people don't have the ability to account for... Thermal conductivity of the TIM is only one of many factors that result in actual performance and resulting temperatures. User testing by modifying the IHS and seeing results is interesting, and I'm paying attention to what people are seeing... But I have a hard time putting much faith in the actual outcomes, good or bad, due to the number of problems inherent to testing it. If anything, my article was an attempt at another possible explanation of what we're seeing with Intel's IB chips as we've seen them. Making the jump to testing it, "fixing it" or "improving upon it" is a big leap in my opinion. Another problem is basic understanding of thermal behaviors. If the chip is hot, that doesn't mean reviewers would observe a match in hotness in the heatsink. After all, the core of the idea here is that the chip temps could be a problem with transmitting heat from the chip to the heatsink. If its a problem with the thermal path through the IHS, a reviewer would find the heatsink isn't as warm as it should be, they wouldn't find the heatsink to be hotter (if it were measurable). I don't mean to come off as "difficult" or "contrarian". Just sharing some thoughts and personal perspectives.

United Kingdom El Gappo says:

I.M.O.G. said:

Another problem is basic understanding of thermal behaviors. If the chip is hot, that doesn't mean reviewers would observe a match in hotness in the heatsink. After all, the core of the idea here is that the chip temps could be a problem with transmitting heat from the chip to the heatsink. If its a problem with the thermal path through the IHS, a reviewer would find the heatsink isn't as warm as it should be, they wouldn't find the heatsink to be hotter (if it were measurable).


I wasn't seeing this.. Small delta between core temp and pot temp/ evap temp as well as my phase unit getting a real good kicking at 5ghz. These chips are kicking out a lot of heat whichever way you look at it.

K404 says:

I.M.O.G. said: K404: I agree with your thinking generally. However, my concern is kinda demonstrated best with an example. Do you trust ANY of the thermal paste roundup testing done in "reviews" anywhere? In my opinion, no one should. The testing was faulty in every case, though I respect the effort and think it should be commended - ultimately the results are misleading when people interpret them however. Thermal tests involving the quality of the interface are very difficult - no good way to evaluate/analyze/report bond line thickness, total wetted area, corresponding wetted area, forget about actually cleaning out residue on a microscopic level so every sample is tainted after you initially tint the surface, etc, etc. These are issues which make respected industry laboratory testing inconsistent and incomparable from one lab's result to the next.

A lot of unknowns and important factors that people don't have the ability to account for... Thermal conductivity of the TIM is only one of many factors that result in actual performance and resulting temperatures.

User testing by modifying the IHS and seeing results is interesting, and I'm paying attention to what people are seeing... But I have a hard time putting much faith in the actual outcomes, good or bad, due to the number of problems inherent to testing it. If anything, my article was an attempt at another possible explanation of what we're seeing with Intel's IB chips as we've seen them. Making the jump to testing it, "fixing it" or "improving upon it" is a big leap in my opinion.

Another problem is basic understanding of thermal behaviors. If the chip is hot, that doesn't mean reviewers would observe a match in hotness in the heatsink. After all, the core of the idea here is that the chip temps could be a problem with transmitting heat from the chip to the heatsink. If its a problem with the thermal path through the IHS, a reviewer would find the heatsink isn't as warm as it should be, they wouldn't find the heatsink to be hotter (if it were measurable).

I don't mean to come off as "difficult" or "contrarian". Just sharing some thoughts and personal perspectives.


I don't interpret you as difficult :)

It's even scienticifally correct to add in all the minute problems to make a comparison truely "proper" with complete control over all the variables.

My personal take is that if some has *reasonable* attention to detail and conducts all their testing following the same methodology, even if the results are skewed/non-perfect, they will all be skewed by the same amount, so the testing can still tell us something useful.

The other thing about a normal user doing the testing is that it reflects a real user in the real world.

Give us all the "physics defines rules from work of this quality" numbers in the world, if a real-world user cannot replicate it, is the work accurate or relevant?

Same as Asus/Gigabyte/eVGA/whoever showing OMFGWTFBBQ PR numbers. Doesn't mean jack to Joe Average, chances are, we won't see the same MHz. SO.... are the PR numbers representitive of the real world? :)

United States sin0822 says:

El Gappo said: I wasn't seeing this.. Small delta between core temp and pot temp/ evap temp as well as my phase unit getting a real good kicking at 5ghz. These chips are kicking out a lot of heat whichever way you look at it.


I see a large delta under LN2 like you see the spike too, but the temp is easy to control like you don't have to keep pouring LN2 like you would for SBe.

K404 says:

http://www.overclock.net/t/1249419/pcevaluation-intel-i7-3770k-temperature-measured-without-ihs Well, that doesn't tell us much. Bad mount?

Canada Vinster says:

K404 said: http://www.overclock.net/t/1249419/pcevaluation-intel-i7-3770k-temperature-measured-without-ihs

Well, that doesn't tell us much. Bad mount?


First post was good, I think the cooler may be to blame, when going directly on the IC shouldn't you use a Direct contact heat pipe cooler? he wasn't..

The rest of the thread was a little silly... too many AMD fanboys and people thinking IB was going to be a huge upgrade to SB...

Also that was an ES, not a retail, so it might just be a sample thing.. I have yet found a photo of a retail chip with IHS off. have any of you?

Vin

K404 says:

I personally wouldn't because of the channels in-between the heatpipes. Depending on the number of pipes the cooler has, you either have two of them in partial contact with the die, with a gap in the middle OR.... just one pipe along the length of the die. (or should the heatpipes meet the die at right angles??... in that case, 3 might make contact, with gaps in between) I've always been of the opinion that direct contact heatpipes NEED an IHS in order to work well BUT if i'm wrong on that, please link me to a comparison that shows it :)

Canada Vinster says:

K404 said: I personally wouldn't because of the channels in-between the heatpipes. Depending on the number of pipes the cooler has, you either have two of them in partial contact with the die, with a gap in the middle OR.... just one pipe along the length of the die. (or should the heatpipes meet the die at right angles??... in that case, 3 might make contact, with gaps in between) I've always been of the opinion that direct contact heatpipes NEED an IHS in order to work well

BUT

if i'm wrong on that, please link me to a comparison that shows it :)


actually that makes sense, I didn't think of it that way, and thinking of it more... you're right... that was a dumb post. my bad.

Vin

United States I.M.O.G. says:

K404 said: http://www.overclock.net/t/1249419/pcevaluation-intel-i7-3770k-temperature-measured-without-ihs

Well, that doesn't tell us much. Bad mount?


Looks like that is a real possibility on the bad mount:



I would hope to see a different TIM pattern on the CPU/cooler base, preferably not one in which there is a thicker/thin patch in corresponding areas of CPU/cooler. Looks like the cooler base may not be so flat perhaps, not sure how you end up with a pattern like that.

United States sin0822 says:

what is wrong with the pattern? Maybe there just is too much TIM. Either way the guy can just test with another cooler. I don't think the difference will be that much. I personally think perhaps he was scared to crack it so he tried very hard to have even pressure, IMO she should have put some pads on the edges of the CPU to help even the wieght load of the cooler. IMo we need more targeted cooling, I am glad though that your article brought to light doubts about Intel's IHS cooling methods.




IMO so where is the best place to put the temp sensor during LN2 with no lid? Right next to the cores, but how about when you still have the IHS on there?

Sweden ME4ME says:

I.M.O.G. said: I would hope to see a different TIM pattern on the CPU/cooler base, preferably not one in which there is a thicker/thin patch in corresponding areas of CPU/cooler. Looks like the cooler base may not be so flat perhaps, not sure how you end up with a pattern like that.


Seriously, stop being so defensive about your teori and start accepting it failed. Nearly all evidence from people who actually did some testing says replacing thermal interface is not solving any heat-related issues.

K404 says:

ME4ME said: Nearly all evidence from people who actually did some testing says replacing thermal interface is not solving any heat-related issues.


Please can you post links? I'm interested in the numbers :) (i'm sure plenty of other folk are too!)

United States Hondacity says:

me4me said: seriously, Stop Being So Defensive About Your Teori And Start Accepting It Failed. Nearly All Evidence From People Who Actually Did Some Testing Says Replacing Thermal Interface Is Not Solving Any Heat-related Issues.



he's sticking with his TIM theory lol

United States sin0822 says:

teori lol its theory.

K404 says:

Not in Swedish it isn't ;)

United States sin0822 says:

oh i wasn't aware we were speaking Swedish :D

United States I.M.O.G. says:

ME4ME said: Seriously, stop being so defensive about your teori and start accepting it failed. Nearly all evidence from people who actually did some testing says replacing thermal interface is not solving any heat-related issues.


Breath in-breath out home slice. At no point have I felt defensive, perhaps you are interpreting my comments incorrectly?

I would be interested in seeing a test performed with good methodology. A lot of people have contacted me with their results, and I've been sharing feedback on what I would do differently.

Keep in mind as well, no where did my article recommend people to go out and test this themselves. Personally, I think Ivy Bride temperatures would be better if it had a soldered IHS like Sandy Bridge did. Secondly, I think temperatures would probably be a little better if the current IHS was removed, but probably not anything groundbreaking... I probably won't remove mine, as I haven't seen anyone test and show that it improves LN2 cooling.

@sin0822: I would like to see an even distribution in paste. If you can see a definite difference in paste thickness by the naked eye, that implies 2 important things. First, the bond line thickness varies A LOT across the die (it is always going to vary, but ideally it would vary by some measurable amount less than can be seen by the naked eye). Second, the mounting pressure or surface finish may not have been very good to leave a pattern like that.

Sweden ME4ME says:

Hey man, im not trying to attack you :) Just saying, the whole story is getting dragged out of proportion. Since that article at OC.com, everyone is talking about it, saying Intel is to blame, they did something wrong, etc. Now that finally people have started to do some testing, and report that the thermalpaste is not the problem, i think that closes this chapter. What bothered me about your previous post is that your still questening, while severel independant people have come to the same conclusion. If you really want to find out, do some testing your self. I think there have been lots of words, but very little facts from you and overclockers.com

K404 says:

http://www.maximum-tech.net/intel-admits-that-ivy-bridge-runs-hotter-because-of-22nm-shrink-12410/ "Intel" say it's the process/ die size

United States Hondacity says:

AFQ the author just got those information from the web(inquirer.net and xs forums) lol

Belgium Massman says:

It kinda makes sense the new tri-gate is giving some temperature issues. Let's check the temperature issues in maybe 6 months when the process has matured a bit.

United States Hondacity says:

Intel has had this Ivy for more than 8months now. I think they'll spend a few more months on it but i think they're more inclined improving haswell[power wise]. haswell[mobile] is due 2013.

United States I.M.O.G. says:

ME4ME said: Just saying, the whole story is getting dragged out of proportion. Since that article at OC.com, everyone is talking about it, saying Intel is to blame, they did something wrong, etc. Now that finally people have started to do some testing, and report that the thermalpaste is not the problem, i think that closes this chapter. What bothered me about your previous post is that your still questening, while severel independant people have come to the same conclusion. If you really want to find out, do some testing your self. I think there have been lots of words, but very little facts from you and overclockers.com


I couldn't agree more with it getting dragged out of proportion. A lot of people have taken what is in the story and ran with it, blamed Intel, or jumped to a lot of conclusions.

It does not irritate me that you put more faith in tests you have seen than how much faith I place in the tests I have seen. I think everyone has a right to make up their own minds. My mind just isn't made up yet. It looks like there could be some flaws in the few user tests I've seen. I probably need to look around more.

I don't intend to do any testing myself with delidding however, I'm just going to pour some LN2 on one of these buggers and freeze the crap out of it. That is more my thing. I've actually been traveling since I published the story, though I did buy a 3770K on the way home yesterday.

@Hondacity: I didn't take anything from the inquirer or XS for this article. I confirmed the W/mK from rge's comments on OC forums and research. I don't know if the inquirer wrote anything about this even?

Christian Ney says:


United States I.M.O.G. says:

Ya, cool to see more tests being done. Hokie posted that over in our Overclockers discussion thread, after seeing it referenced on VR-Zone. The site that did the testing actually replaced the thermal paste and still used the heatspreader (sticking it back on with 2 sided sticky tape), and seem to have gotten good results - a lot cooler when overclocked using different paste under the IHS. Their results show more improvement than I would expect from just replacing the paste.

Please log in or register to comment.

Leave a Reply: (BBCODE allowed: [B], [QUOTE], [I], [URL], [IMG],...)