Saturday, May 12th 2012

TIM is Behind Ivy Bridge Temperatures After All

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.

Source: PC Watch
Add your own comment

219 Comments on TIM is Behind Ivy Bridge Temperatures After All

#1
Disruptor4
I've got a 3770k and am happy with it. Sure, it idles at 30C and can get up to around 60C on full load, but knowing that they skimped out on A) the quality of TIM used and B) the fact that they wanted to save a few bucks by not soldering the IHS, and then charge more for the CPU, it does annoy me quite a bit. If they change that with a new revision that's dodgy as well imo.
Posted on Reply
#2
Xzibit
by: Disruptor4
I've got a 3770k and am happy with it. Sure, it idles at 30C and can get up to around 60C on full load, but knowing that they skimped out on A) the quality of TIM used and B) the fact that they wanted to save a few bucks by not soldering the IHS, and then charge more for the CPU, it does annoy me quite a bit. If they change that with a new revision that's dodgy as well imo.
Thats actually better then the OP test but I dont get the big deal maybe its just me

If the 3770k is is doing 34C on idle and 61C on load i thinks its fine.

61C @ 3.9ghz on stock cooler

84C @ 5ghz on stock cooler

Unless everyone missed the part where he left Turbo Boost on the chip turned on.

For Air cooling I think thats fine
Posted on Reply
#3
eidairaman1
by: jihadjoe
I don't get the aversion to using shims. Re-lidding the CPU is very inefficient compared to direct touch. Anyways we all did that shimming thing back in the day, and I'm willing to bet most people here didn't have any problems.

Those old Athlon XP/Duron days were awesome.
I shimmed mine, Ive even crushed a CPU before because I bent the HS Clip to add more pressure
Posted on Reply
#4
Disruptor4
by: Xzibit
Thats actually better then the OP test but I dont get the big deal maybe its just me

If the 3770k is is doing 34C on idle and 61C on load i thinks its fine.

61C @ 3.9ghz on stock cooler

84C @ 5ghz on stock cooler

Unless everyone missed the part where he left Turbo Boost on the chip turned on.

For Air cooling I think thats fine
Well I am using a Coolermaster Hyper 212 Evo with Arctic Silver 5 paste. It's always nice to have cooler temps though and I would've liked for it to be a bit cooler but really not too fussed. More just in my head thinking it's bad because of all the press, but I'm actually happy with it though and I won't be kicking up a stink tbh because I moved from an AM2 4200+ system!

As a side note, Mussles, you live a couple hours away from me. Wanna take my IHS off, reapply some good TIM and then put the IHS back on and see if we see a nice drop in temps? :P
Seriously though, if Intel offered a optional recall to have better TIM on it as a revision, i'd take it in a heart beat. Free new CPU with cooler temps!
Posted on Reply
#5
Wile E
Power User
by: cadaveca
You cannot use retention bracket with IHS removed, which makes it hard to give CPU correct pressure. It is also not quite that easy to get even pressure across the core when using block mountings. With very careful block palcement, it could be better, for sure, but it's much easier to just replace TIMM, put IHS on top, and crank it down as far as it will go..with no fear of cracking the core.
Easier, but not better. I can say I honestly don't know how people manage to crack their cores. If you use spring loaded bolts and nuts, and tighten 2 of the heatsinks/block's mounting screws diagonal from each other at the same rate at the same time, you'll never crack a core. Then you do the same to the other 2 until the reach the same tension. You don't need to crank them down on a bare core, just need them snug enough to not move.

I would totally do this to a $300 or less cpu. Not to a more expensive one though, at least not until it's already out of warranty. That's the entire reason my 980 still has it's lid. lol.

by: erocker
But if you have a 2600k/2700k, there's no reason to get Ivy Bridge.. Unless your current SB chip is a complete dud.
I'd have to agree there. If I were building a new system, I would definitely go IB though. The only way I would go from SB to IB is if I were going from a 2500 to a 3770 to gain hyperthreading.
Posted on Reply
#6
douglatins
by: Xzibit
Thats actually better then the OP test but I dont get the big deal maybe its just me

If the 3770k is is doing 34C on idle and 61C on load i thinks its fine.

61C @ 3.9ghz on stock cooler

84C @ 5ghz on stock cooler

Unless everyone missed the part where he left Turbo Boost on the chip turned on.

For Air cooling I think thats fine
I doubt it. 5ghz was going over 100C in reviews. With WC.

All this TIM thing is a bad move, that ihs is for protection against damage and uneven pressure that could crush the transistors.

For average users those temps are going to escalate in time since tim gets bad after long periods to the point that will harm the CPU. For the average overclocker it's the worst, its hard to have the guts to maybe potentially destroy the CPU. Removing that thing is tricky.
Why would intel release a product that is worse for 99% of customers and better for 1%?

removing the ihs and replacing the tim and putting back the ihs is just proving that it was the worst idea ever. Fixing their job, really? There is only 2 options, use it as is and have bad temps, and remove it risking losing the high value chip for better temps.
Sure the 3 option is remove it, place new tim and place it again, but again this is just proving how bad this design is.
I hope this won't happen ever again and is fixed with new revisions. If they wish to get that 1% release batches with tim, like 3750kk? or something
Posted on Reply
#7
CaptainFailcon
sombody needs to start breaking some dies and figuring out exactly what kind of forces they can take as well as start machining shims
Posted on Reply
#8
Xzibit
by: douglatins
I doubt it. 5ghz was going over 100C in reviews. With WC.
I got the stock cooler part wrong

A 3770k on Turbo Boost will upscale 400mhz.

Disruptor4 said he was using a Coolermaster Hyper 212 Evo with Arctic Silver 5 paste and the reviewer is using a Thermalright Silver Arrow

So before he take out the TIMMY!!!!:rockout:
Hes getting 5ghz on load when he leaves Turbo Boost on at 4.6ghz@1.2V witha max temp of 84C

Focus on the top bar :twitch: on each run





If you go to this chart before he removes poor one legged TIMMY!!!!:rockout:
Hes able to base clock it to 4.9ghz@1.45V and run Cinbench without a Themal Shutdown so if Turbo Boost is on hes getting 5.3ghz before reaching 105C

Thats not bad Air Cooled considering Sandy-Bridges had issues getting to 4.5ghz stable. Factor in the SB to IB differance and its like running a Sandy-bridge at 5.6ghz air cooled.

Focus on first table :twitch:
Posted on Reply
#9
eidairaman1
This should be expected from a tick anyway (Tock being an improvement in architecture, tick being a shrink) Im honestly not worried about the temps anyway, this chip looks catered to overclockers anyway. I wonder if 2011 has this chip per se

by: Xzibit
I got the stock cooler part wrong

A 3770k on Turbo Boost will upscale 400mhz.

Disruptor4 said he was using a Coolermaster Hyper 212 Evo with Arctic Silver 5 paste and the reviewer is using a Thermalright Silver Arrow

So before he take out the TIMMY!!!!:rockout:
Hes getting 5ghz on load when he leaves Turbo Boost on at 4.6ghz@1.2V witha max temp of 84C

Focus on the top bar :twitch: on each run
http://www.techpowerup.com/img/12-05-11/114b.jpg




If you go to this chart before he removes poor one legged TIMMY!!!!:rockout:
Hes able to base clock it to 4.9ghz@1.45V and run Cinbench without a Themal Shutdown so if Turbo Boost is on hes getting 5.3ghz before reaching 105C

Thats not bad Air Cooled considering Sandy-Bridges had issues getting to 4.5ghz stable. Factor in the SB to IB differance and its like running a Sandy-bridge at 5.6ghz air cooled.

Focus on first table :twitch:
http://www.techpowerup.com/img/12-05-11/114a.jpg
Posted on Reply
#10
H82LUZ73
by: cadaveca
Bullshit. A shim is a bit of metal that will brace the CPU agains the heatsink, and prevent damage to the core. NO R&D involved, jsut a couple of measurements, and cutting some metal, or even plastic.

like this:

http://www.overclockersonline.net/images/articles/coppershims/durontop.jpg



http://www.tweaknews.net/reviews/tsr/img/shim10.jpg


You can pretty much guarantee that extreme guys are gonna want one.
Good old Socket A 462 Duron .Thunderbird/Thoroughbred shims.....:respect::respect::respect:

For Socket 939 was not needed also The 462 Palomino chips,Just clarify the year 2000-2001 ;) We seen the first IHS on AMD when socket(first gen x64 cores) 754 I think was out for a short time before the X64 send gen 939 boards.

http://en.wikipedia.org/wiki/Duron

http://en.wikipedia.org/wiki/Socket_754

Hm for Intel to do this ,They must have some reason,Do I think it is bad?, maybe Seeing how some user want to overclock the IB chip,But why I think SB is the better of the two.
Posted on Reply
#11
Jurassic1024
by: qubit
Eh? :confused: An Athlon XP dates from 2003, so I'm talking about a decade ago, lol. Dang, TPU didn't even exist then!
That made no sense whatsoever. Your original comment was confusing enough, and now you just took it to a whole new level. aye yi yi.
Posted on Reply
#12
Yellow&Nerdy?
Apparently it's not only AMD that has non-competent marketing people... I have no idea why Intel chose to use cheap TIM instead of fluxless solder. I understand if it's cheaper, that they would use it on non K-series CPU's, but for the love of god, why didn't they go with fluxless solder on the unlocked CPU's? It can't be much cheaper to use TIM, and I for one would be glad to pay 10 - 15 bucks more for better thermals. Not that it's a critical problem, since a IB CPU overclocked to it's max is usually faster than a SB CPU clocked to it's max, even if the SB CPU is clocked higher. But it would still be nice to see this fixed in a later revision...
Posted on Reply
#13
Octavean
So who is this Tim and why is everyone so ticked off at him,…..?

LOL, just kidding. I expect Intel may address this with later revisions and start using more traditional fluxless solder. It is something of a blessing to some though, just don’t crush your core in the process.

My theory was that the chip got too hot even with fluxless solder so fusing it to the heat spreader may result in weakening its electrical connections as the heat spreader heats and cools (expands / contracts). However, that’s probably not likely if the solder can help keep the temp down, which this artical seems to suggest,....
Posted on Reply
#14
badtaylorx
oi oi oi....you know taking off the heatspreader has been found counter-productive since the beginning of multi core architecture.....it's there for a reason...

IT WORKS

will replacing the tim help,,, yes.

but ALL of the air heatsinks and water cooling cpu blocks were designed for use with a heatspreader...if only 1 heatpipe crosses your poor lil ivy chip, what good is that???
Posted on Reply
#15
Casecutter
Oops - Intel went cheap basically constrained or hobbled it this round to be able to later offer the same chips (with a simple fix) and a boost in clocks. Intel got caught! :slap:
Posted on Reply
#16
Widjaja
Intel can use TIM under their new chips as they do not NEED to try because they have no competition.

Sure they may get hotter with over clocking in comparison to Sandy Bridge but still beats the offerings of any chip of any other manufacturer.

So they can cheap out and still be ahead.
Posted on Reply
#17
Random Murderer
The Anti-Midas
by: eidairaman1
I wonder if 2011 has this chip per se
i know the SB-E has fluxless solder under its hat and not TIM, but i don't think that's what you're asking.
would you mind elaborating a little on what it is you're asking? i'm sure between cadaveca, Aquinus, and me we'll be able to answer.
Posted on Reply
#18
kajson
I kinda wonder if maybe a company like Apple, who want to put these babies into their new macbook airs, will call Intel out on this and force em to apply a higher quality thermal solution... I've seen a guy post intel burn test temps of his brand new MSI Ivy laptop at 95 celcius, you cant really expect to pay $2000 for a brand new laptop just to, best case scenario have it's lifespan shortened conciderably / worst case demolish your fertility.
Posted on Reply
#19
TheLostSwede
by: kajson
I kinda wonder if maybe a company like Apple, who want to put these babies into their new macbook airs, will call Intel out on this and force em to apply a higher quality thermal solution... I've seen a guy post intel burn test temps of his brand new MSI Ivy laptop at 95 celcius, you cant really expect to pay $2000 for a brand new laptop just to, best case scenario have it's lifespan shortened conciderably / worst case demolish your fertility.
Notebook CPUs don't have heat spreaders...
Posted on Reply
#20
n-ster
by: TheLostSwede
Notebook CPUs don't have heat spreaders...
unless they have the desktop version in the laptop :p
Posted on Reply
#21
Gradius2
This is totally UNACCEPTABLE !
Posted on Reply
#22
Wile E
Power User
by: badtaylorx
oi oi oi....you know taking off the heatspreader has been found counter-productive since the beginning of multi core architecture.....it's there for a reason...

IT WORKS

will replacing the tim help,,, yes.

but ALL of the air heatsinks and water cooling cpu blocks were designed for use with a heatspreader...if only 1 heatpipe crosses your poor lil ivy chip, what good is that???
The results achieved going lidless suggest otherwise.
Posted on Reply
#23
largon
by: badtaylorx
oi oi oi....you know taking off the heatspreader has been found counter-productive since the beginning of multi core architecture.....it's there for a reason...
Laws of physics disagree.
Posted on Reply
#24
jaredpace
by: Gradius2
This is totally UNACCEPTABLE !
Yup, this heat issue sucks.
Posted on Reply
#25
badtaylorx
physics....thats funny.

thats not what the physics say at all!!!

here...this is from XtreemSystems......there are some seriously smart mf'rs on this site....
Some people will read get 10-25C better temps by removing the IHS, then do so, and get disappointed when they see 15+C higher temps and realize they ruined their cpu.

Nobody gets 10-25C better temps by removing the IHS on a modern high power density 100W+ cpu, though a few claimed it without proof. I have removed the IHS on a few and my temps were worse as has another person that posted pics in previous thread with before/after temps. Old 40w cpus that just used paste, yes, those you could remove the ihs and get better temps, since replacing paste with paste and removing a layer.

2/3 or more of the gradient in modern cpus is from hot spot through the die substrate, that is via stanford article linked and pics in post here, also from drilling holes through ihs with calibrated thermocouples on ihs and in die. The copper laden die has thermal conductance of ~125W/MK, solder attach 80 w/mK but is only 20 microns thick, and copper is 400 w/mk, hence one would expect most of gradient is through the die. Also you can load 2 cores with coredamage/prime (set affinity in task bar to load just 2 cores of 4) and you will see 20C gradient from loaded core to idle core, which demonstrates high gradient through die substrate, like pic below in slide from ESL.


You could get ~10C better temps (assuming 150W tdp and 30C gradient from core to IHS temps) by removing IHS, If you then solder on a waterblock with 80w/mk solder assuming you could avoid voids like intel, and assuming the same bondline thickness of ~20 microns). Then the waterblock is the heatspreader + eliminate user tim. The IHS isnt just protection, but for spreading heat to a much larger area at 80W/mk through only 20 microns thickness then at 400w/mk via copper, ie relatively rapidly to a relatively large surface area....before the end user puts on relative crappy 3-4W/mk tim paste at thickness higher than 20 microns, then tries to cool it with water 0.6w/mk or air, even worse.

But removing 80w/mk solder and replacing with 3-4 w/mK paste at a stage when heat is still confined to relatively small surface area, is a disaster that will result in 10-15C+ higher temps at stock alone.


These results DO NOT point at - IHS=lower temps

they point out a bad decision coupled by poor implementation on Intel's part
Posted on Reply
Add your own comment