Wednesday, April 25th 2012

Ivy Bridge Temperatures Could Be Linked To TIM Inside Integrated Heatspreader: Report

PC enthusiasts with Ivy Bridge engineering samples, and reviewers at large have come to the consensus that Ivy Bridge is a slightly warmer chip than it should be. An investigation by Overclockers.com revealed a possible contributing factor to that. Upon carefully removing the integrated heatspreader (IHS) of an Ivy Bridge Core processor (that nickel-plated copper plate on top of the processor which makes contact with the cooler), the investigator found common thermal paste between the CPU die and the IHS, and along the sides of the die.

In comparison, Intel used flux-less solder to bind the IHS to the die on previous-generation Sandy Bridge Core processors in the LGA1155 package. Attempting to remove IHS off a chip with flux-less solder won't end well, as it could rip the die off the package. On the other hand, the idea behind use of flux-less solder in CPU packages is to improve heat transfer between the die and the IHS. Using thermal paste to do the job results in slightly inferior heat transfer, but removing IHS is safer. One can be sure that making it safe for IHS removal couldn't have been the issue behind switching back to conventional thermal paste, as everything under the IHS isn't user-serviceable anyway, and off limits for them. Perhaps Intel kept extreme overclockers in mind.

Source: Overclockers.com
Add your own comment

97 Comments on Ivy Bridge Temperatures Could Be Linked To TIM Inside Integrated Heatspreader: Report

#1
newtekie1
Semi-Retired Folder
by: PopcornMachine
The number of people who wish to overclock really is moot here.

The point is the, if the use of TIM instead of solder is the cause of the heat problems, why did Intel do this? What did they possibly gain? A half cent on each chip?

Why not use the same tried and true method they always have?

If they manufacture the same way, then it doesn't matter who overclocks and who doesn't.

This is the real point. Why would they mess with something they didn't need to mess with?
At this point, I'd say using paste instead of solder saves a little more than half a cent per chip. The solder requires special equipment, and and is more expensive than paste.

Even if we say the savings is only say 1 cent per processor, Intel has shipped 75 Million+ Sandybridge processors, and likely will ship close to that number if not more Ivy Bridge processors. So that is a huge savings. One cent on 75 Million processors is $750,000. That is a nice chunk off the bottom line for a change that won't affect but maybe 10% of users.
Posted on Reply
#2
PopcornMachine
by: newtekie1
At this point, I'd say using paste instead of solder saves a little more than half a cent per chip. The solder requires special equipment, and and is more expensive than paste.

Even if we say the savings is only say 1 cent per processor, Intel has shipped 75 Million+ Sandybridge processors, and likely will ship close to that number if not more Ivy Bridge processors. So that is a huge savings. One cent on 75 Million processors is $750,000. That is a nice chunk off the bottom line for a change that won't affect but maybe 10% of users.
Uh, 750G is a drop in the bucket to Intel. If they are causing this much heat to save that amount, then they are just greedy beyond all understanding.

But that is the way of the world today. Money, money, money.

Common sense be damned.
Posted on Reply
#3
newtekie1
Semi-Retired Folder
by: PopcornMachine
Uh, 750G is a drop in the bucket to Intel. If they are causing this much heat to save that amount, then they are just greedy beyond all understanding.

But that is the way of the world today. Money, money, money.

Common sense be damned.
This much heat? It is a few degrees C, maybe 10 at most at stock speeds, where most of these will be run. And the temps aren't anywhere near worrysome. As the reviewers put it it is "slightly warmer than it should be". And not all of that comes down to the TIM, the higher thermal density plays a large role as well.
Posted on Reply
#4
CaptainFailcon
you can always remove the IHS you're self and run it naked :) just don't crush the die .....
Posted on Reply
#5
cadaveca
My name is Dave
by: PopcornMachine
Common sense be damned.
I think perhaps the fact that the solder uses relatively rare minerals might play a role in them making this decision, which actually holds a lot more common sense than first meets the eye.
Posted on Reply
#6
CaptainFailcon
by: cadaveca
I think perhaps the fact that the solder uses relatively rare minerals might play a role in them making this decision, which actually holds a lot more common sense than first meets the eye.
with all the flooding in east Asia you bet you're biscuits the price of neodymium is still though the roof
Posted on Reply
#7
cadaveca
My name is Dave
by: CaptainFailcon
you can always remove the IHS you're self and run it naked :) just don't crush the die .....
the socket retention mechanism relies on the IHS as a shim to have adequate pressure applied to the chip so that the pins touch properly. Running without the IHS isn't as easy as it sounds.
Posted on Reply
#8
PopcornMachine
by: newtekie1
This much heat? It is a few degrees C, maybe 10 at most at stock speeds, where most of these will be run. And the temps aren't anywhere near worrysome. As the reviewers put it it is "slightly warmer than it should be". And not all of that comes down to the TIM, the higher thermal density plays a large role as well.
If "everyone" is saying it is "slightly" warmer, why are we talking about it?

I know what I've read. We're not talking about stock speeds.
Posted on Reply
#9
NHKS
by: cadaveca
I think perhaps the fact that the solder uses relatively rare minerals might play a role in them making this decision, which actually holds a lot more common sense than first meets the eye.
probably indium based.. which, as u said is rare

Wiki: "Indium is used as a thermal interface material in the form of pre-shaped foil sheets fitted between the heat-transfer surface of a microprocessor and its heat sink. The application of heat partially melts the foil and allows the indium metal to fill in any microscopic gaps and pits between the two surfaces, removing any insulating air pockets that would otherwise compromise heat transfer efficiency."

Intel Patent(2010)-Methods of fabricating robust integrated heat spreader designs
Posted on Reply
#10
newtekie1
Semi-Retired Folder
by: PopcornMachine
If "everyone" is saying it is "slightly" warmer, why are we talking about it?

I know what I've read. We're not talking about stock speeds.
When Intel is making the decision, you better believe it is based on stock speed, regardless of what you are talking about. Their decisions have to be based on what the large majority of customers will experience, and since 90% of customers will run stock speeds, decisions have to be made based on stock speeds. What doesn't make sense is throwing away $750,000 just to cater to 10% of your market, and even then of that 10% of the market, maybe 10% will really care enough that the processor runs warmer that they won't buy it. How many people have you seen say "OMG, they run so much hotter, I'm not buying one!"?
Posted on Reply
#11
PopcornMachine
by: newtekie1
When Intel is making the decision, you better believe it is based on stock speed, regardless of what you are talking about. Their decisions have to be based on what the large majority of customers will experience, and since 90% of customers will run stock speeds, decisions have to be made based on stock speeds. What doesn't make sense is throwing away $750,000 just to cater to 10% of your market, and even then of that 10% of the market, maybe 10% will really care enough that the processor runs warmer that they won't buy it. How many people have you seen say "OMG, they run so much hotter, I'm not buying one!"?
Why do you continue talking about it? There's no problem. It's only a few degrees. No one cares about overclockers.

Personally, I am very happy to have my 2500K with proper thermal material in it. Even if it doesn't matter.
Posted on Reply
#12
newtekie1
Semi-Retired Folder
by: PopcornMachine
Why do you continue talking about it? There's no problem. It's only a few degrees. No one cares about overclockers.

Personally, I am very happy to have my 2500K with proper thermal material in it. Even if it doesn't matter.
Because you keep talking about it like it matters, I'm saying it doesn't.
Posted on Reply
#13
PopcornMachine
I will just have to agree to disagree then.

It does in fact matter. Don't think people would have started this thread, or written articles about it if it didn't matter to at least a few people.

If it doesn't matter to you, I can live with that. But I'm not going to agree with you.

For the cost of the salary of a few overpaid executives, they cause what should be an easily overclockable chip to become a very hot chip.

So I don't agree that it makes sense for them to do this.
Posted on Reply
#14
newtekie1
Semi-Retired Folder
by: PopcornMachine
For the cost of the salary of a few overpaid executives, they cause what should be an easily overclockable chip to become a very hot chip.

So I don't agree that it makes sense for them to do this.
And it is still easily overclockable, reaching the same speeds as SandyBridge with significantly less voltage no less. Yes, it runs hotter, but it also has a higher thermal limit. And again, this isn't all due to the TIM, the higher thermal density plays a huge part.

So, yes, if it saves them money and it doesn't affect the overwhelming majority of users, it makes perfect sense.
Posted on Reply
#16
newtekie1
Semi-Retired Folder
by: PopcornMachine
Like I said, I can live with you being wrong.
Except I'm not. And to answer your question about why someone made this thread if it doesn't matter, people bitch about stuff that doesn't matter all the time.
Posted on Reply
#17
Wile E
Power User
by: newtekie1
I complain now just like I complained then. And I thought only the ones with turbo-boost got the 4 extra bins(beyond the turbo-boost bins), or do they all get 4?

And while I agree somewhat with the getting tomorrows performance today, it was still used to see real world performance gains, even the high end processors saw real world performance gains, because hardware was behind software. That isn't the case today, hardware(especially Intel's) has far outpaced the software. Whether it was overclocking a high end processor or a low end, it was for a pretty noticeable performance gain. I still remember overclocking Ahtlon XP 3200+ chips to shave up to half an hour+ of video rendering, now doing the same to an i7 shaves maybe 30 seconds.

And, yes, I completely agree that most people do not overclock. But I'm just talking about the people that would care that there is TIM under the heatspreader vs solder. The idea was said to be to help with extreme overclockers while shafting everyone else, which I assumed to mean everyone else that overclocks, not everyone in the world that will use the processor. Obviously, as you said, this doesn't affect 90% of the people that will end up using these processor.
High-end BD encodes benefit greatly from OCing.

But to be honest, if I'm not encoding, I just run at stock with the rad fans turned down for silence. I only load up my OC profile for heavy encoding anymore. Even my slower IPC 980x is more than enough for most tasks (including gaming @ 1920x1200) for me not to bother.

by: cadaveca
the socket retention mechanism relies on the IHS as a shim to have adequate pressure applied to the chip so that the pins touch properly. Running without the IHS isn't as easy as it sounds.
You remove the metal socket retention mechanism and use a bolt through kit when you delid. The die sticks up past the actual socket once the bracket is removed.
Posted on Reply
#18
cadaveca
My name is Dave
by: Wile E
The die sticks up past the actual socket once the bracket is removed.
I hear ya, but the pros tell me otherwise:

by: shamino
temps are not as good with IHS without exertion of force by ILM on the core but with good thermal paste and remounted IHS, temps are better and may bring the CPU to cold bug
I mean, I never say stuff like this unless it's been confirmed by those that know far more than I do!

http://kingpincooling.com/forum/showthread.php?t=1730
Posted on Reply
#20
TheMailMan78
Big Member
Maybe they just use paste to save a buck as the Sandy Bridge is already over kill for 90% of applications today? Why waste the time/cash with solder on making something "cooler" when its already cool enough AND over powered for everything we pretty much throw at it? Sure I can hear them in board meetings now "Listen guys we have to add MILLIONS to our overhead in solder so less then 1% of our market can get a better score in 3DMark."
Posted on Reply
#22
qubit
Overclocked quantum bit
by: phanbuey
Nice... i could see it. Could it also be that the thermal sensors are in a different spot relative to the core? Closer to the GPU? etc...

dunno, I just remember having a 92C E4300 that is still alive today in a friends rig... like 5 years later.
That's one hard bastard. :laugh:
Posted on Reply
#23
xenocide
by: PopcornMachine
Personally, I am very happy to have my 2500K with proper thermal material in it. Even if it doesn't matter.
A 2500K at 4.8GHz is roughly equal to a 4.4GHz 3570K in terms of performance, the only trade off is the 2500K uses noticably more power, requires higher Voltage, and runs maybe 5c colder. This is really a non-issue...
Posted on Reply
#24
Wile E
Power User
by: cadaveca
I hear ya, but the pros tell me otherwise:



I mean, I never say stuff like this unless it's been confirmed by those that know far more than I do!

http://kingpincooling.com/forum/showthread.php?t=1730
I'm not seeing where it contradicts my statement. Though I don't know what ILM means. If it's like every other Intel I've seen, the pcb sits pretty flush with the socket itself, and the core sits above that. Hence, once you completely remove the metal retaining bracket and base (leaving only the bare socket), and bolt the cooler down, you get direct core contact.
Posted on Reply
#25
larsjrg
Warmer? My new i5-3550 runs same temperatures with the intel stock cooler at the 2500K do, about ~62-63C on load at stock clock speeds. I'm a happy buyer with a Ivy Bridge CPU using less power than my previous Athlon II x4, which was about ~48C on load with AMD stock cooler...

So what is all the fuss I read about increased temperatures? I certainly do not see the proof when it comes to my Ivy Bridge compared vs. old reviews on Sandy Bridge. (Different intel stock coolers might contribute a few degrees differences)
Posted on Reply
Add your own comment