I agree, the entire cooling/protection system engineered around the P4 platform (the IHS, the heatsink retention mechanism and the integrated on-die thermal protections mainly) is extremely well concived and well executed. There's almost zero risk associated with handling and installing of a P4, and the cool temperaures that result from the stock system are pretty remarkable, considering the power the newest chips consume. Whether the IHS directly impacts temperatures by a degree or two or not, I think it's one of the best innovations Intel has brought to the table in terms of CPU packaging in a long time.
I agree, the entire cooling/protection system engineered around the P4 platform (the IHS, the heatsink retention mechanism and the integrated on-die thermal protections mainly) is extremely well concived and well executed. There's almost zero risk associated with handling and installing of a P4, and the cool temperaures that result from the stock system are pretty remarkable, considering the power the newest chips consume. Whether the IHS directly impacts temperatures by a degree or two or not, I think it's one of the best innovations Intel has brought to the table in terms of CPU packaging in a long time.
Yeah. I agree. I think it's a good thing overall. I just think that people who believe that it somehow spreads heat are crazy. They don't realize that the only way to do that is by having a larger die. The thing that spreads heat is the heatsink. Another tiny heatsink in between which makes good contact with the CPU will not help spread the heat any further. All it does is add two areas of TIM and contact rather than one. Overall, I think it's a positive thing though.
Althought I'm really looking forward to bumpless buildup layer CPU packaging. no worried with that.
Which is why Intel calls it an Integrated Heat Sink, not Integrated Heat Spreader.
Once again, it's not intended to spread heat but draw it off the CPU with better contact than you would get from a standard heatsink/fan combo. The increased area of the IHS creates a larger surface to transfer heat to the HS/F creating a better likelihood of greater heat transfer than you would get with a HS/F combo that has poor or even decent contact with the core itself. I don't see why you are arguing this it's pretty much common sense. Sure someone could engineer a better cooling solution that doesn't use Intel's IHS, but considering the P4 doesn't require exotic cooling there's no reason to bother with anything better.
actually, the do call it an integrated heat spreader and the processor runs cooler when the die has direct contact with the heatsink so I guess your common sense is actually B.S. Two mating surfaces will always produce more thermal resistance than one. Having one make very good contact with the cpu doesn't change the fact that you still have to make contact with the heatsink on top of that and now you've got twice the resistance.
What do you mean? P4 runs considerably hotter than Athlon XP does. Yes, the Mhz is higher, but P4 is not a cool chip by any stretch of imagination. Clock-for-clock, P4 is only slightly cooler to Athlon XP
Apparently reading isn't one of your strong suits when you've already made up your mind. I provided a link above to Intel's website describing their packaging. They call it a sink, not a spreader. If you actually believe that one decent connection will always outperform 2 better connections than there really isn't much else to debate as you'll believe what you want regardless.
Actually, Randum is right.
The P4 operates at a lower temp than athlon XP.
Yes, you can take an athlon XP and say that it dissipates 70watts of heat. Then you can take a P4 and say that it dissipates 75watts of heat but the p4 will run cooler. Why? Because it has a larger die. The more surface area the die has, the more heat it dissipates more quickly. You could have a processor with 50 billion transistors that threw off 1000watts but if it had a large enough surface area to cover the earth, it wouldn't be warm to the touch.
The P4s larger surface area on the die allows it to run cooler than the athlon XP. The 100watt, 80watt figures are only useful when comparing processors of equal die size etc.
The P4 operates at a lower temp than athlon XP.
Yes, you can take an athlon XP and say that it dissipates 70watts of heat. Then you can take a P4 and say that it dissipates 75watts of heat but the p4 will run cooler. Why? Because it has a larger die. The more surface area the die has, the more heat it dissipates more quickly. You could have a processor with 50 billion transistors that threw off 1000watts but if it had a large enough surface area to cover the earth, it wouldn't be warm to the touch.
The P4s larger surface area on the die allows it to run cooler than the athlon XP. The 100watt, 80watt figures are only useful when comparing processors of equal die size etc.
Yes. I believe that one decent connection will outperform one great connection and one decent connection. That's physics for you.
Maybe you're the one that should practice your reading. This is from that link you posted:
"The OOI then has an Integrated Heat Spreader (IHS) that helps heatsink dissipation to a properly attached fan heatsink. The OOI is used by the Pentium 4 processor"
They use the term IHS to mean Integrated Heat Sink and Integrated Heat Spreader in the same document so obviously Intel uses BOTH terms. Most commonly, it's knows as a heat spreader.
I was talking about the absolute heat-dissipation, not heat-dissipation/square mm.
And if the difference in die-size does cause such drastic difference, it's more than negated by the fact that P4 needs to runs alot faster than XP, and that causes the heat-dissipation to increase substantially.
And what is the difference is size between P4 and Barton?
Barton: 101 square mm
P4: 131 sqaure mm (if I remember correctly)
Is that difference enough to compensate for the higher heat-dissipation? And since the CPU's use heat-sinks, the absolute number of area they use to dissipate heat is more up to the heatsink than to the CPU IMO.
Actually, yes. It has 30% more die size. That means that as long as the total heat dissipation is less than 30% higher than the athlon XP (which it is) the die will run cooler all else being equal.
I haven't looked for myself, but... Are you sure about that? A P4-2.26ghz is only slightly cooler than a T-bred "2800+"?
I agree with Wigz. Clock for clock, the P4 is MUCH cooler than any Athlon. An AThlon and P4 of the same speed or speed rating: the P4 will dissapate more heat into the case but the die itself will run cooler.
I'm still pondering those pins. But thought I'd stop in to make a short comment on this "P4 IHS" discussion which hijacked the topic somehow.
Yes, more mating surfaces lead to higher resistance. That is a correct statement by itself. However there are instances where more mating surfaces are preferable. Example - two years ago 99% of us had aluminum only heatsinks. With flip chip packaging the contact areas decreased at the same time thermal output increased. What came along? Heatsinks with copper inserts. And temps went down even as mating surfaces went up.
In the same way the IHS does aid thermal transfer in many, but not all, situations. By spreading the heat to a size much larger than the core itself the process is helped when: A) using an aluminum only heatsink B) using reliable, if not the very most efficient TIM C) when the sink is not well mated to the contact area, be it direct core or IHS.
Yes, if you take off the IHS and go core to copper with high efficiency TIM temps will likely be lower. That is not the OEM situation, nor the situation for most home builders. While the IHS can actually increase temp in some instances what it ensures is that it takes a very bad mating surface between it and the sink to lead to overheating issues. There are also instances where it may not reduce the overall core temp, but prevents a hotspot over a portion of the core due to poor mating between it and sink.
It's like antilock brakes, in theory you can stop faster without them. In fact you can stop faster without them if conditions are right. In practice most cars stop shorter with them.
BTW, looked over at a couple articles overclockers.com. The position taken in them seems to be the IHS is not about crush prevention only, if at all. Sure we all can find articles supporting our own premise, it's a big internet after all. My position is simply the IHS is a thermal safegaurding device, and yes, protects cores from crushing as well. All in all a very nice addition. Which if you don't like for your application you can chop off.
--Mc
Yes, more mating surfaces lead to higher resistance. That is a correct statement by itself. However there are instances where more mating surfaces are preferable. Example - two years ago 99% of us had aluminum only heatsinks. With flip chip packaging the contact areas decreased at the same time thermal output increased. What came along? Heatsinks with copper inserts. And temps went down even as mating surfaces went up.
In the same way the IHS does aid thermal transfer in many, but not all, situations. By spreading the heat to a size much larger than the core itself the process is helped when: A) using an aluminum only heatsink B) using reliable, if not the very most efficient TIM C) when the sink is not well mated to the contact area, be it direct core or IHS.
Yes, if you take off the IHS and go core to copper with high efficiency TIM temps will likely be lower. That is not the OEM situation, nor the situation for most home builders. While the IHS can actually increase temp in some instances what it ensures is that it takes a very bad mating surface between it and the sink to lead to overheating issues. There are also instances where it may not reduce the overall core temp, but prevents a hotspot over a portion of the core due to poor mating between it and sink.
It's like antilock brakes, in theory you can stop faster without them. In fact you can stop faster without them if conditions are right. In practice most cars stop shorter with them.
BTW, looked over at a couple articles overclockers.com. The position taken in them seems to be the IHS is not about crush prevention only, if at all. Sure we all can find articles supporting our own premise, it's a big internet after all. My position is simply the IHS is a thermal safegaurding device, and yes, protects cores from crushing as well. All in all a very nice addition. Which if you don't like for your application you can chop off.
--Mc
So it's nothing more than a way of removing the extra mating surface from the heatsink and putting it right on the processor. Sounds like a reliability thing to me. Of course, the best solution is all copper to have only one mating surface.
Of course, the IHS doesn't just offer 2 mating surfaces, it offers 3. One to the CPU -silicon to copper-one to the top of the IHS-copper to nickel- and then one to the heatsink. Nickel to aluminum.
The best solution is using an all copper heatsink=1 mating surface
the second best would be an aluminum heatsink with a copper insert=2 mating surfaces.
Do you mean the amount of good die (the correct meaning of yield) or the bin splits (speeds)?
A larger die doesn't really have an effect on speed. But it definitely has an effect on yields. Since you get less die per wafer, you are going to need to manufacture more wafers to produce the same amount of die.
Excuse me? That is just about the silliest thing that I have had anyone reply to me in a comment in a long time. You are right, Para, I should go read up on packaging from Overclockers.Com rather than believe the data from the company who switched over their main product line from a cheap direct die attach system to a more costly IHS system. Data that I received during a seminar that I was taking on future packaging trends by the package development team at Intel.
The IHS works because the bump underfill material causes strain on the die that causes it to bow out in the middle thus presenting a fairly small overall surface area to the heatsink. Initially direct die attach techniques attempted to maximize surface area by applying the heatsink with a high amount of pressure to the die. This works when the die is small, but on larger dies the warpage of the die is such that the die makes very little contact over the entire surface area of the die. The IHS uses several thermally matched materials that are directly bonded to the warped die and then attached to the overall IHS surface to create a flat surface for the heatsink. Encapsulation technqiues such as the IHS also helps reduce metal fatigue issues and other FC reliability issues.
The lapping techniques by the guys over at overclockers.com avoid this problem by actually grinding down the IHS to create a level surface. Of course, their results show better cooling - they lapped off quite a bit of metal and created a mirror-like surface. For a real comparison, they'd need to remove the entire IHS structure. But this experiment is pointless because there is a lot of data in the journals on the subject.
For references, look at any of the IEEE Transaction on Components, Hybrids, and Manufacturing Technology in the years 1990-1992. There are, as I recall, quite a few papers on the subject at the time. If you want specific articles, I can log on to work and look up several, but I'm on my Intel sabbatical and I'd rather not unless you are going continue on that I've been brainwashed and that I'm talking BS. For less useful information faster, check google for "flip chip warp" "bump underfill warp" and "flip chip encapsulation".
Lastly, think about it this way. Every single high-end vendor uses a form of IHS on their server and workstation flip-chip products. PA-RISC, Alpha, Power4, Itanium - all of them use an IHS type system. Most of these parts are on the extreme end of their power dissipation envelope. And yet they all use an IHS-type encapsulation. Why? Are they worried that someone is going to take the heatsink off of the CPU on $20k server and then incorrectly attach it and then they'll get customer returns? Not likely. It improves heat transfer efficiency.
paralaguazer: I mean it in the context that the bigger die won't bring lower temps due to the IHS
That is true if the in-plane thermal conductivity is significantly higher than the through plane thermal conductivity.
Under such boundary conditions, the same size IHS will present the same thermal profile to the external heatsink regardless of die size (so long as die size < IHS size).
after i removed the IHS, my idle temps dropped 10 degrees C and my load temps dropped 6 degrees. i think that is pretty substantial.
Interesting stuff, pm. So heatspreaders have evolved, not all are created equal to start with. Complicates things when comparing old info like "I took the heatspreader off my K6-450 and..." beyond the obvious.
So I went looking again to see if I was correct in remembering reading about people removing their heatspreaders from P4's and getting lower temps. Seems it did reduce temps in some cases, but very little. Problem with many reports is people also lapped their sinks and switched from the thermal pad to AS3 or made other such changes at the same time which basically invalidates their results. Including the links in case anyone else wants to read about real world experiences in heatspreader removal and possible improvements to temps along with dead chips. Googled, didn't find anything so ended up going back to overclockers to find anyone who'd done this.
Link 1 - someone who did it, wouldn't report new temps
Link 2 - general rant sounding familiar
Link 3 - someone removed theirs, got a 1 degree drop
Link 4 - Several reports of 2-3 degree drop, some on older cores
Seems that yes, you can drop temps by removing the IHS. If you also lap your sink and use AS3 or similar as most if not all these people have done. But all this proves is that the IHS is not much if any hinderance even when great care is taken (and still question validity as mentioned above). Not seeing temps going up either though which should be taken into consideration. Seems to be of minimal effect in carefully prepared and monitored heatsink installations, and even the removals aren't reporting higher overclock speeds which I assume was usually the goal. And I still say in a less than perfect sink/TIM installation it can't help but improve temps by spreading heat across a non-lapped or otherwise imperfect sink base. I've never yet seen a heatsink come out of the box with a surface which will give you anywhere near the uniform contact as the IHS to core contact is. My Alpha 8045 is pretty nice, my FOP-32 was horrid. So I can definately see why it'd help cooling, all while keeping your core nice and safe.
Actually glad to have had a reason to go looking now, I admit I had something of a false impression of the effect of the IHS. Hope my next chip has one, iNTEL or AMD.
Now, back to me pondering those pins!
So I went looking again to see if I was correct in remembering reading about people removing their heatspreaders from P4's and getting lower temps. Seems it did reduce temps in some cases, but very little. Problem with many reports is people also lapped their sinks and switched from the thermal pad to AS3 or made other such changes at the same time which basically invalidates their results. Including the links in case anyone else wants to read about real world experiences in heatspreader removal and possible improvements to temps along with dead chips. Googled, didn't find anything so ended up going back to overclockers to find anyone who'd done this.
Link 1 - someone who did it, wouldn't report new temps
Link 2 - general rant sounding familiar
Link 3 - someone removed theirs, got a 1 degree drop
Link 4 - Several reports of 2-3 degree drop, some on older cores
Seems that yes, you can drop temps by removing the IHS. If you also lap your sink and use AS3 or similar as most if not all these people have done. But all this proves is that the IHS is not much if any hinderance even when great care is taken (and still question validity as mentioned above). Not seeing temps going up either though which should be taken into consideration. Seems to be of minimal effect in carefully prepared and monitored heatsink installations, and even the removals aren't reporting higher overclock speeds which I assume was usually the goal. And I still say in a less than perfect sink/TIM installation it can't help but improve temps by spreading heat across a non-lapped or otherwise imperfect sink base. I've never yet seen a heatsink come out of the box with a surface which will give you anywhere near the uniform contact as the IHS to core contact is. My Alpha 8045 is pretty nice, my FOP-32 was horrid. So I can definately see why it'd help cooling, all while keeping your core nice and safe.
Actually glad to have had a reason to go looking now, I admit I had something of a false impression of the effect of the IHS. Hope my next chip has one, iNTEL or AMD.
Now, back to me pondering those pins!
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