TIM is Behind Ivy Bridge Temperatures After All

Interesting follow-up on the Ivy Bridge TIM debate. What do you guys think about this?

It's proven: the thermal interface material (TIM) used by Intel, inside the integrated heatspreader (IHS) of its Core "Ivy Bridge" processors are behind its higher than expected load temperatures. This assertion was first made in late-April by an Overclockers.com report, and was recently put to test by Japanese tech portal PC Watch, in which an investigator carefully removed the IHS of a Core i7-3770K processor, removed the included TIM and binding grease, and replaced them with a pair of aftermarket performance TIMs, such as OCZ Freeze and Coolaboratory Liquid Pro.

PC Watch findings show that swapping the TIM, if done right, can shave stock clock (3.5 GHz, Auto voltage) temperatures by as much as 18% (lowest temperatures by the Coolaboratory TIM), and 4.00 GHz @ 1.2V temperatures by as much as 23% (again, lowest temperatures on the Coolaboratory TIM). The change in TIM was also change the overclockability of the chip, which was then able to sustain higher core voltages to facilitate higher core clock speeds. The report concluded that Intel's decision to use thermal paste inside the IHS of its Ivy Bridge chips, instead of fluxless solder, poses a very real impact on temperatures and overclockability.


15

Sweden Calathea says:

The TIM is correctly applied and covers the core. If this would've been regular IHS to cooler contact then I'd call a 20 degree difference between TIMs BS. However between die and IHS, I'm not sure...

Norway knopflerbruce says:

Wow, 60 degrees is alot! (18% of ~330K). Wait... that puts it below ambient.

Would be cool (!) if someone could translate this percent nonsense for me:D

Australia Uncle Fester says:

I did this aswell to see if it would make 1 of my junk chips clock better under cold --- answer is no chip was still junk. I replaced the TIM under the IHS with gelid extreme and it was much better on air, i saw a drop of about 10c at idle.

Denmark M.Beier says:

Despite the fact that I officially have no right to comment (Right massman?).
I call this BS... TIM is not the reason, it does make a difference, sure, but, it is not the core of the issue... Less contact surface, and same heat that has to be removed, does not take a Ph.D. knowing that the chip will become hotter..

uncle fester said: I did this aswell to see if it would make 1 of my junk chips clock better under cold --- answer is no chip was still junk.
I replaced the TIM under the IHS with gelid extreme and it was much better on air, i saw a drop of about 10c at idle.

10c idle seems quite fantastic, could you share It0 It1 and Lt0 Lt1 ? (Idletemp0, Loadtemp0 etc.)

Belgium leeghoofd says:

Did you seal the heatspreader Fester again ?

Australia Uncle Fester says:

I just stuck the heatspreader back on i didnt glue it again, before removing and replacing TIM temps in bios no load were read as 32c while pot temp read as 22c ( empty no ln2 ofcourse ) then when i replaced the TIM with gelid extreme the temps in bios read 22c and pot read 22c. I really cbf benching it on air for any comparison as i dont give a fuck, i wanted to test if it would clock better subzero it had no cold bug before and after but didnt clock any better.

Norway knopflerbruce says:

I don't say the 10c lower readings are bullshit, but the sensor is probably way off. What's the power draw in idle mode? Can't be much at all, and definately not enough to cause a 10c difference between one TIM brand and another. I think the delta would be 40c+ under load, then.

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

Idle temps have considerably less relevance for judging anything. Sensor readings are calibrated to become more accurate the closer you get to TJmax. Load readings get closer to TJmax, and are better than idle readings to judge anything by.

That said, without ambient measurements, both idle/load are rather meaningless for talking about temperatures.

Uncle_Fester wasn't testing for temps though, he was trying to see if his dud chip became a better chip - nothing wrong with that.

For people that have tried testing temps and changing TIM on their own, test results have been all over the place. There are lots of ways to test poorly and get inaccurate and unreliably reproducible results. There are few ways to do it right, and it isn't easy.

M.Beier said: I call this BS... TIM is not the reason, it does make a difference, sure, but, it is not the core of the issue... Less contact surface, and same heat that has to be removed, does not take a Ph.D. knowing that the chip will become hotter.


It doesn't take a PH.D to read the ratings of the chips. Less contact surface, lower TDP. Look at multiple sites testing, and you'll see that even while overclocked the Ivy Bridge rigs are pulling lower wattage than Sandy Bridge at the same clocks. There is less power running through Ivy Bridge...

Bottom line - No one knows exactly what factor the smaller surface area and power density plays, but everyone pretty much agrees Ivy Bridge temperatures are higher than expected despite the process shrink. Now it looks like people are showing that just changing the paste under the IHS can actually improve temps...

My article didn't even make that claim, it was based on the premise that temps could be better if Intel went back to using the fluxless solder. As temps are found to be considerably better by just replacing the TIM, I would go on to suggest that they would be better yet if they were able to go back to the solder process. There may be various reasons why that isn't possible, or why Intel chose to not do that.

Denmark M.Beier says:

I.M.O.G. said:
It doesn't take a PH.D to read the ratings of the chips. Less contact surface, lower TDP. Look at multiple sites testing, and you'll see that even while overclocked the Ivy Bridge rigs are pulling lower wattage than Sandy Bridge at the same clocks. There is less power running through Ivy Bridge...

Bottom line - No one knows exactly what factor the smaller surface area and power density plays, but everyone pretty much agrees Ivy Bridge temperatures are higher than expected despite the process shrink. Now it looks like people are showing that just changing the paste under the IHS can actually improve temps...

My article didn't even make that claim, it was based on the premise that temps could be better if Intel went back to using the fluxless solder. As temps are found to be considerably better by just replacing the TIM, I would go on to suggest that they would be better yet if they were able to go back to the solder process. There may be various reasons why that isn't possible, or why Intel chose to not do that.


Last time I checked, they Ivy rigs pulled more.... I believe it was OBR.... Please linkie to a review that indicates higher powerconsumption with sandy in OC'ed state...

NB; I will check this myself within first 48 hrs when my Z77 mobo arrives.

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

Friends don't let friends read OBR looking for reliable information. Let's be friends? Their information can be early and interesting, but the journalism portrayed there doesn't exemplify a loose standard of reliability. Different sites target different goals, and every site is useful in its own way... I like reading early reports from places like OBR, fudzilla, semi-accurate, etc. I'm not bashing them... I just don't use them as a reference when I'm trying to make a point. I have just been paying attention to what people have tested in communities. Ivy Bridge consuming less power isn't really something that is under debate. That hasn't stopped a lot of people from talking about power density and other things blindly without paying attention to actual data. There is review testing power consumption results here, and you can find more corroborating info across the communities if you look at what people are getting who went from SB to IB (I've seen similar results verified by Overclockers.com forum members): http://www.bit-tech.net/hardware/cpus/2012/05/01/intel-core-i5-3570k-cpu-review/7 Note that their overclocked 3770 is only at 4.8GHz, because they couldn't do 5GHz as they did for other results, so the loaded power consumption for 3770K is a little lower than it would be at 5GHz. The 3570 can be compared directly against 2500K however.

Denmark M.Beier says:

I.M.O.G. said: Friends don't let friends read OBR looking for reliable information. Let's be friends?

Their information can be early and interesting, but the journalism portrayed there doesn't exemplify a loose standard of reliability. Different sites target different goals, and every site is useful in its own way... I like reading early reports from places like OBR, fudzilla, semi-accurate, etc. I'm not bashing them... I just don't use them as a reference when I'm trying to make a point.

I have just been paying attention to what people have tested in communities. Ivy Bridge consuming less power isn't really something that is under debate. That hasn't stopped a lot of people from talking about power density and other things blindly without paying attention to actual data. There is review testing power consumption results here, and you can find more corroborating info across the communities if you look at what people are getting who went from SB to IB (I've seen similar results verified by Overclockers.com forum members): http://www.bit-tech.net/hardware/cpus/2012/05/01/intel-core-i5-3570k-cpu-review/7

Note that their overclocked 3770 is only at 4.8GHz, because they couldn't do 5GHz as they did for other results, so the loaded power consumption for 3770K is a little lower than it would be at 5GHz. The 3570 can be compared directly against 2500K however.


For what its worth, your opening line sure made me laugh :)
Thanks for the link, however, I will test it myself as soon as possible, been 2 weeks delay so far, dead CPU then dead motherboard, I feel really 'blessed'.

Germany lrwr says:

note that your percentage calculation is wrong.
for example a decrease from 60°C to 30°C is NOT a 50% decrease, but a 9% decrease
when dealing with a difference in absolut temperature you have to calculate in Kelvin not in °C...
what i would like is a graph of temperature DIFFERENCE between room temperature an cpu temp...

United States Hondacity says:

Decrease from 60c - denotes reference from 60c, hence his statement is correct.

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

lrwr said: note that your percentage calculation is wrong.
for example a decrease from 60°C to 30°C is NOT a 50% decrease, but a 9% decrease
when dealing with a difference in absolut temperature you have to calculate in Kelvin not in °C...
what i would like is a graph of temperature DIFFERENCE between room temperature an cpu temp...


Ya, we talked about that too on OCF. Basically, the frame of reference is important with the percentage - most people like Hondacity mentioned, using the percentage if you reference the ranges makes sense to them.

Technically, you are right that Kelvin is better to use to talk about an actual difference in the terms of percentages... Kelvin is pretty foreign to a lot of people however, and most everyone works in Celsius ranges when comparing computer temps. Realistically, to fit the audience, I think its best to stick with Celcius... And maybe get away from using percentages.

Everyone talks about it this way though, and I've seen your point brought up a couple times now.

Norway knopflerbruce says:

+1. I mentioned this already. i don't see how you can use percent when dealing with Celsius at all.. better get rid of it. Those who know physics and math won't have a clue what they're talking about. The others THINK they got what they meant, but in most cases they're wrong:D

Please log in or register to comment.