Heatsinks are flat and PIII Cu'mines are not... so what "gives"?

[KenG]

I'm a newbie, but have been reading and learning a lot. I apologize if this question has already been answered elsewhere. I DID try to find it.

The PIII coppermine flipchips have a little hump in the middle that sticks up maybe a millimeter or two. I haven't seen one, but the stock heatsink from Intel appears to have a recess in its thermal tape to accomodate this.

Everything I read about attaching heatsinks to cpu's emphasizes maximizing the contact area, minimizing the air bubbles and making the film of goop (whatever is used) as thin as possible.

With the PIII coppermine FC-PGA the flat heatsink sits atop the little square center elevation and looks like a teeter-totter.

I found one test that tried to use the thin copper shim with the hole in the middle that 3DFXCOOL sells, but there were problems (with shorting?) and (electrical, I think) tape was used to fill in the space and make the heatsink stable/balanced (as in the stock Intel setup, it seems). The heatsink tested OK, but it left me wondering what part of the cpu needs to be cooled (i.e., where is the heat, if not evenly distributed?) The copper shim does not cover the whole cpu, and it doesn't really even get all that close to the raised center part. Consequently, there seems to be fairly little contact surface area.

Every heatsink I've seen pictures of is FLAT. So this problem has to be common. Usually, in the pictures the contact part isn't even shown.

Besides the minimal contact between cpu and heatsink, what about the lack of contact of the whole cpu? (The heatsink CONTACT AREA is not as big as the cpu.... not even close.)

Does anybody have any sources of information about this?

 

[KenG]

And another thing: I've read that the chips are somewhat brittle. Doesn't pressing down with the heatsink only touching the middle distort the cpu, either by bending the chip or by crushing the center part? At the very least it seems like it would stress it.

 

[Mday]

the hump is the processor core, the rest is filler... whatever you do , do not damage it. what needs to touch the HS is the blue nub (i call it a nub). the spacer\shim does not have to touch the center at all, it's just filling space.

heatsinks designed with fcpga in mind have a different type of clip (the metal bar) which excerts force over the core itself to balance the heatsink.

the goop is thermal compound, just put in on top of the blue nub.

the core is very brittle. the heating and expansion of the chip is not even. the cache resides at a corner of the core, blue nub, and is what is the part that usually breaks.

though the size of the core does mean a smaller surface area, the fact that the .18 micron processor requires less power, which would be turned to heat, means there is an evening out (not balanced though). the processor does get quite hot, relatively speaking, my celeron 533a @ 800 (which is fcpga) is running about 5 to 7 degrees higher than my 466 @ 582, but then it's summer.

 

[Dexion]

The Blue Core is where all the heat is produced. The rest of the chip are circuits embedded in plastic or silicon that connect to the core, which produces little heat. Of course a Coppershim would work nice so that it would spread the pressure away from the P3 core preventing it from crushing it, but it wouldn't help cooling very much (unless your using Peltiers). The reason heatsinks are so large is the amount of heat released from such a tiny surface, it would require a large enough heatsink(like the Alpha) to dispurse the heat.

 

[Quickfingerz]

You only need to cool that bump.

 

[KenG]

Thanks everybody. Like I said... a newbie.

One more question though. Is it better to leave that gap between the heatsink and the chip (around the core) open and maybe get a little circulation and additional cooling? Or is it better to build it up for stability of the heatsink and maybe risk holding more heat in?

 

[acebathound]

Build it up? As in with heatsink? You don't want to do that. Heatsink's used to fill microscopic gaps that exist between the part of the cpu and heatsink that touch..if the heatsink isn't touching a part, it would be best not to use heatsink to fill an 1/8th inch gap that would then touch the heatsink.

 

[acebathound]

btw.. 3DFXCool.com has some "copper shims" that are supposed to help fill that gap. http://www.3dfxcool.com/ Not sure how well they work..but might be interesting. $9?? hrm..get some pennies and melt 'em down

 

[Mikewarrior2]

I put a copper spacer(1/4inch) that is screwed into my Arctic Circle heatsink(or original orb), and that improves the performance of that heatsink by a lot... it performs roughly 1 C better than my PEP66.

My p3 core is lapped a bit, to flatten it out, but I think the copper really helps in my case because it helps to spread the heat better from the small contact surface(ala alpha's, but mine is a separate copper piece).



Mike