btw, when i removed the IHS from my cpu, i found that there was a big hole in the middle of whatever thermal material had been applied between the IHS and the CPU die. i'm sure that the fact that the connection between the die and the IHS was terrible was a large part of the big temp reduction.
Yeah, PM you're talking about lapping one down. I'm talking about removing it completely. If you remove the heatspreader and your temps drop 6-10 degrees celcius. It's obviously hindering cooling.
Is that the ONLY thing you took away from his posts? A single line, and a bit out of context.
He's talking about the functionality of the IHS in terms of heat transfer, and why it is beneficial.
Just because one, or a few, get significant results (or get no results) from removing said IHS is hardly scientific. The results pre and post are not controlled, and too many factors can influence the results.
I'm guessing INTC went down the route of IHS as more of an insurance policy than a performance enhancement. Given the number of heatsink mfgs out there, and the range in quality, the last thing INTC wants is a user base getting the impression that that their P4 chips are inferior because the consumer uses a crappy 3rd party heatsink solution. Going to the IHS is just one more way to reduce the extremes in the range of thermal environments Intel's chips will face from 3rd party thermal solutions. It always can be improved on by customizing a particular P4 w/o IHS to a particular 3rd party heatsink, but as a ~30B$ company you can't rely on that being the last defense.
To me it is no different than going with locked multipliers and on-die thermal diodes, i.e. a business decision to minimize risk in field product variety.
You definitely work in a fab!
EDIT: spelling...
He hit it right on. It's there to move the hardest mating surface away from the end user so they can use cheap aluminum heatsinks and so they won't destroy the processor. It's not there to improve cooling.
And yes, I do believe that sometimes Wingznut can't see the forest for the fab.
MadRat
Lifer
- Oct 14, 1999
- 11,999
- 307
- 126
Two guesses:
1. The increase in the number of pins will aid in cooling the CPU; each pin transfers heat from the CPU packaging.
2. An AMD "Thermal Shield" will be for protective purposes, not for additional cooling.
Having direct contact with a copper core will be better, as someone said, because you eliminate one layer to transfer the heat. Just because the data hasn't been collected in a scientific study does not mean that it is not true. I'd rather challenge anyone to prove that the Thermal Shield is better for heat transfer than direct contact with the Copper Core of the heatsink. Its not.
The best way to study heat generation of the CPU would be to run it very slowly and aim a high-definition infrared scope at its surface. The bus and core would need to run at a proportionate speed so as to not create abnormal hot spots. (i.e. Running the FSB at its normal speed but using a low - say a 1/2 - multiplier means the test is disproportionately going to heat up components tied to the FSB.) Slowly increasing the speed while maintaining the same scale would lend information to hwo real actual speeds scale in heat generation. I'm sure Intel probably already does this and hunts down hotspots at the microscopic level in order to uncork the performance of their architecture.
1. The increase in the number of pins will aid in cooling the CPU; each pin transfers heat from the CPU packaging.
2. An AMD "Thermal Shield" will be for protective purposes, not for additional cooling.
Having direct contact with a copper core will be better, as someone said, because you eliminate one layer to transfer the heat. Just because the data hasn't been collected in a scientific study does not mean that it is not true. I'd rather challenge anyone to prove that the Thermal Shield is better for heat transfer than direct contact with the Copper Core of the heatsink. Its not.
The best way to study heat generation of the CPU would be to run it very slowly and aim a high-definition infrared scope at its surface. The bus and core would need to run at a proportionate speed so as to not create abnormal hot spots. (i.e. Running the FSB at its normal speed but using a low - say a 1/2 - multiplier means the test is disproportionately going to heat up components tied to the FSB.) Slowly increasing the speed while maintaining the same scale would lend information to hwo real actual speeds scale in heat generation. I'm sure Intel probably already does this and hunts down hotspots at the microscopic level in order to uncork the performance of their architecture.
MadRat, apparently you didn't read what pm said about the silicon not being flat. Nor did you do any further research on it.
Para, you are one of the last people who should comment on having an open mind... Especially when it comes to debating your "facts." Do some more research, using the suggestions that pm posted. You'll find a lot of good information, and maybe even learn something new.
You guys need to find more than a few posts on msg boards before you deem something as fact. Using that methodology, I could "prove" pretty much anything I wanted.
i.e. I could find literally hundreds (thousands?) of posts about how AMD cpu's are unstable. Does that make it a fact? Of course not.
I'm not saying that removing the IHS and using enough care, dilligence, and cost won't net you a gain in heat transfer. I'm just saying that you can't say that draw a definite conclusion based on one (or a few) post from someone in an uncontrolled experiment.
Para, you are one of the last people who should comment on having an open mind... Especially when it comes to debating your "facts." Do some more research, using the suggestions that pm posted. You'll find a lot of good information, and maybe even learn something new.
You guys need to find more than a few posts on msg boards before you deem something as fact. Using that methodology, I could "prove" pretty much anything I wanted.
i.e. I could find literally hundreds (thousands?) of posts about how AMD cpu's are unstable. Does that make it a fact? Of course not.
I'm not saying that removing the IHS and using enough care, dilligence, and cost won't net you a gain in heat transfer. I'm just saying that you can't say that draw a definite conclusion based on one (or a few) post from someone in an uncontrolled experiment.
Alright. I'll conduct a controlled experiement then. I'll be picking up a P4 on thursday and building a new system. I'll do that experiment. I'm tired of your Bullsh1t blind intel bias and idiot assumptions. We'll see how the experiment turns out. See you thursday
MadRat
Lifer
- Oct 14, 1999
- 11,999
- 307
- 126
Having control of the thermal transmission properties and actually improving them are two different things, Wingz.
The Thermal Shield helps to control the transfer, but it adds a new barrier to the transfer; I just stated the obvious. Nowhere did pm suggest that each shield is carefully crafted to fit the contorsions created when the core is fastened to the packaging, he only said Intel found a benefit from using a shield. It would be a trick to stamp a shape into each shield specifically for the curvatures of the core, and this stamping process likely didn't happen. I did search according to what he said (i.e. flip chip encapsulation) and the studies seems to be more concerned with failure of the contact points from fatigue rather than the specifics on warpage of the core. In no way am I saying that pm's claim that the bulge in the center wouldn't cause premature failure of the core. All I am saying is that the methodology hasn't been done to prove direct contact with the heatsink is better, but that doesn't mean they are not correct in their assumptions.
I think Intel uses the Thermal Shield to control the properties of the torsion forces placed on the core and not specifically for its properties of thermal transmission.
The Thermal Shield helps to control the transfer, but it adds a new barrier to the transfer; I just stated the obvious. Nowhere did pm suggest that each shield is carefully crafted to fit the contorsions created when the core is fastened to the packaging, he only said Intel found a benefit from using a shield. It would be a trick to stamp a shape into each shield specifically for the curvatures of the core, and this stamping process likely didn't happen. I did search according to what he said (i.e. flip chip encapsulation) and the studies seems to be more concerned with failure of the contact points from fatigue rather than the specifics on warpage of the core. In no way am I saying that pm's claim that the bulge in the center wouldn't cause premature failure of the core. All I am saying is that the methodology hasn't been done to prove direct contact with the heatsink is better, but that doesn't mean they are not correct in their assumptions.
I think Intel uses the Thermal Shield to control the properties of the torsion forces placed on the core and not specifically for its properties of thermal transmission.
I do not get it... why do people always accuse wingnut, and to a lesser extent, pm, for the "intel bias"??? They seem to be more open to new ideas that many people on this board. They have been honest enough to let us know that they DO work for the company, and I find that really wonderful of them. People have their opinions, and I think both these people back most of their thoughts with links to reputable articles on the net, unlike some people we know (hint... hint... ).
I would say they have looked at it pretty hard, and balanced out the good and bad... i.e., the average user is seeing a benefit from this.
Also, I would say that the end-price of a whole intel system has come down a bit because of this, as OEMs do not have to pay much attention to the quality of heatsinks used....
/EDIT: and THAT makes me a senior member!!!
Wings has an amazing capacity to ignore the point of an argument and focus on anything which might look bad for intel whether it's relevant or not. This doesn't apply here but I"m tired of his idiot Intel bios which permeates everything he says. It sad that even he's aware of it and it's sad that he can convince others it doesn't exist by stating that he's not bias and that by no means would he ever blah blah blah. But then if you look at what he's actually saying...
And yes, this experiment will be quite scientific.
Tone down the name calling or you'll be on vacation
AnandTech Moderator
EDITED by paralazaguer: SUCK IT
This is the type of behavior that will keep me from ever purchasing an Anandtech forum subscription.
Originally posted by: MadRat
So what you are really intending is to troll someone with this comment, eh?
ummm.. take it any way you want... I meant it as a quick poke at him, which I think the Moderators have done too....
This is getting way off topic, so I'll make this short...
UlricT... Thanks for the kind words.
para... Have I ever said that I had zero bias? Not that I recall. And to the contrary, that's one of the (secondary) reasons that I put my employer in my sig: So people understand "where I'm coming from." (FYI, the primary reason is the fact that Intel asks us to... Not us specifically, but just in their general guidelines.)
What I find ironic is that you feel the need to flame us because we disclose who we work for, yet you choose not to reveal who you work for. (Of course, that might explain some of the obvious ire you have towards Patrick and I.) Don't get me wrong, I'm not suggesting you need to. It's a personal choice, and one that everyone has the right to.
ReMeDy{WcS}... Don't let one (or a few) person's attitude affect your overal view. There are probably hundreds of good attitudes for every bad one. (Off Topic not included. )
UlricT... Thanks for the kind words.
para... Have I ever said that I had zero bias? Not that I recall. And to the contrary, that's one of the (secondary) reasons that I put my employer in my sig: So people understand "where I'm coming from." (FYI, the primary reason is the fact that Intel asks us to... Not us specifically, but just in their general guidelines.)
What I find ironic is that you feel the need to flame us because we disclose who we work for, yet you choose not to reveal who you work for. (Of course, that might explain some of the obvious ire you have towards Patrick and I.) Don't get me wrong, I'm not suggesting you need to. It's a personal choice, and one that everyone has the right to.
ReMeDy{WcS}... Don't let one (or a few) person's attitude affect your overal view. There are probably hundreds of good attitudes for every bad one. (Off Topic not included.
)
There was no name calling. What I said was: " but I"m tired of his idiot Intel bias which permeates everything he says. "
That means that the bias is idiotic, not the person with it.
And yes, wings, you have claimed to be 100% non-biased in the past.
I'm absolutely amazed at the ammount of intelligence on these forums. I really couldn't care less if I got a "vacation." From all the time I've been spending in this forum, my head's starting to hurt. It's amazing that I don't run into this level of stupidity, bias, and ignorance on any other forums that I visit. This place is a joke and the moderators are worse than the members.
Thanks, we appreciate the input
Bye bye
AnandTech Moderator
i lurk everyday ( i never post because i have nothing to say to contribute ) and para hit me as annoying yet arrogant. it was time for him to go.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads
- Started by Tigerick
- Replies: 23K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter