The IHS (Integrated Heat Spreader) on Intel CPU's is not present primarily to protect the core. It's there to improve thermal dissipation by making better contact with the CPU. The CPU protection is an added bonus. But if it didn't improve the thermal properties, then it wouldn't be there. As someone mentioned earlier in the thread, the vast majority of CPUs are used by OEM's - not end users - and OEM's do not have problems with doing a bare-die attach in a manufacturing situation.
seems tough enoughfor me.guy on Overclockers.com removed the IHS and got 3-5C better temps because of it
but really, how many horror stories of cracked dies do you hear these days? I mean, I heard about em back before, but nowadays I hear almost nothing. not a pip. the core can really take a beating before it dies, you know.
but really, how many horror stories of cracked dies do you hear these days? I mean, I heard about em back before, but nowadays I hear almost nothing. not a pip. the core can really take a beating before it dies, you know.
I've read their article. Leaving aside the fact that they don't have a statistical sample size to conclude much of anything, there is an explanation for it.
When the die is flipped and soldered to the pin-grid array package using the flip-chip C4 process an underfill resin is applied and it wicks through ball array between the package and the part. The resin is then cured and during this process a strain develops between the package, the resin and the die which causes the die to bow out and thus it is no longer a flat surface. The solution to this is to attach the heat sink with a great deal of pressure to reverse the strain from the underfill and form a flat surface for the heatsink to contact the die. This is fine when the die is of fairly small size (Pentium III, Athlon), but on larger dies the amount of pressure required to reverse the strain becomes fairly significant. Intel's solution was to use a custom packaging process which uses a custom material that takes on the shape of the bowed die surface and them forms a flat surface and then encased this in a integrated package.
The overclockers.com guys did not remove the IHS, they merely lapped it down until they reached silicon. This still presented a flat surface to the heat sink and removed a fair amount of interface material in the process. If they had removed the IHS and then slapped a heatsink on it, they would have found that temperatures actually increased due to the die not being planar.
Typing: "C4 underfill strain" into Google turned up several pages on the subject. The highest ranked is this one: http://www.asymtek.com/news/articles/2001_08_imaps_norway.pdf and there is a discussion of the subject and a picture on page 5.
Demon-Xanth
Lifer
- Feb 15, 2000
- 20,551
- 2
- 81
Funny, only a few years ago people were popping the caps on the K6 chips and claiming that the exposed core of the FCPGA was the greatest thing since PGA.
The fragility is only a mere downside, not a flaw. You add a heat spreader, the downside to that is that heat transfer isn't as efficient.
As far as "having a hard time running on stock air", AMD's not having a problem with that, my XP2000+ CPU has been running with a retail heatsink for months now, at a temp of about 130F in a case without tons of fans. Just one in the front and a small one on the side, not high speed. Right on the line with other CPUs in similar cases. If you want to go through and rip AMD apart for not redesigning thier CPU sockets for every little heatsink improvement, lets look at the hassle we've delt with from Intel, since the pentium w/ multipliers was introduced, there was two sockets for the pentium, one for the PPro, a slot for the P2/P3, THREE DIFFERENT socket 370s, a slot for the Xeon, two sockets for the Xeon, and two sockets for the P4. AMD's had one socket for the K5-6, one slot for the Athlon, and one socket for the Athlon. Now, I'm not saying that Intel was wrong for doing this. They just had the ability to push OEMs a bit more. The upside was the ability to add cooling on the Xeons, P2/3s, and P4s when they needed extra cooling and scale back on the Celerons, P3 tulatarians and P3Es. The downside? More confusion, and compatibility questions.
There is not one perfect packaging. If there were they wouldn't be making microBGA, BGA, PGA, PPGA, FCPGA, LGA, PLCC, PQFP, LQFP, and the 100 other packages they do.
The fragility is only a mere downside, not a flaw. You add a heat spreader, the downside to that is that heat transfer isn't as efficient.
As far as "having a hard time running on stock air", AMD's not having a problem with that, my XP2000+ CPU has been running with a retail heatsink for months now, at a temp of about 130F in a case without tons of fans. Just one in the front and a small one on the side, not high speed. Right on the line with other CPUs in similar cases. If you want to go through and rip AMD apart for not redesigning thier CPU sockets for every little heatsink improvement, lets look at the hassle we've delt with from Intel, since the pentium w/ multipliers was introduced, there was two sockets for the pentium, one for the PPro, a slot for the P2/P3, THREE DIFFERENT socket 370s, a slot for the Xeon, two sockets for the Xeon, and two sockets for the P4. AMD's had one socket for the K5-6, one slot for the Athlon, and one socket for the Athlon. Now, I'm not saying that Intel was wrong for doing this. They just had the ability to push OEMs a bit more. The upside was the ability to add cooling on the Xeons, P2/3s, and P4s when they needed extra cooling and scale back on the Celerons, P3 tulatarians and P3Es. The downside? More confusion, and compatibility questions.
There is not one perfect packaging. If there were they wouldn't be making microBGA, BGA, PGA, PPGA, FCPGA, LGA, PLCC, PQFP, LQFP, and the 100 other packages they do.
Thats a slightly misleading response - the IHS will ensure a more uniform contact between cpu and heatsink but a correctly aligned cpu core contacting the heatsink directly will always show better thermal dissipation. The IHS is there for ease of assembly for OEM's imoh.
It's most certainly not a design flaw. I never had a problem with CPU cracking. AMD engineers correctly calculated that the added cost of adding a heat spreader would be lower than the added cost of warranty replacements on cracked CPUs, if any. These decisions are not made haphazardly. There are engineers who make these decisions after doing considerable amount of research on packaging, heat dissipation, power density, cost to benefit analysis, reliablity, etc. If a CPU only costs $20 to manufacture, it's makes no economic sense to add $1-2 heat spreader, because that's 10% of the cost right there. Maybe if the CPU costs $500 to make, the math is different. Also, some engineering resources would have to be diverted to design a efficient heat spreader for the package, and additional step of adding a heat spreader during packaging might have a small impact on final yield if the heat spreader is not properly bonded to the core to provide efficient cooling.
The problem arises with badly designed heatsinks mostly, I had a (chrome) orb cooler - the one with 2 fans - thats sucker could chip a core quite easily as the retention clip was a pain.
There are a bunch of Electrical Engineers at AMD that do this stuff for a living. Probably have PhDs in stuff that most people dont understand the name of. So I dont think it is a design flaw. I am sure they have their reasons.
-doug
-doug
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: 24K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter