I have heard that before, but I have not seen the Intel source of that info. Has anyone seen Intel actually give this info? Don't take it personally, but I do not see how Intel would design something like that. There would have to be an air gap or something under the IHS, on top of the CPU, in order for it to flatten out while clamped. The sides would then also have to move outward to some extent to compensate for the flex, and I don't see that happening either.

It may have been a result of the application of the IHS to the chip. Forced in place and glued, the thin center would have bulged slightly, but it may have been necessary to insure a good tim joint under the IHS, then the pad negated the bulge problem without the expense of another machining step to flatten it again. The nickel coating would have to have been done after it was flattened out on the chip also. It was probably a cost vs performance juggle. They kept the cost down and still got reliable performance.
All fun speculation but the spreader is bulged, that's a fact. We know that a thinner tim joint is better also. Kept as stock you are forced to either accept low surface area contact with a thin layer of goop, or use a heavy layer. Something to look for right away when AMD starts putting them on their chips too:)

Hmm... fair enough then. I was just curious... can't afford to actually buy one to see for my self, 'n all.

Nice work N8! So are you writing up an article?;)

There must be some kind of discoloration in your picture: the mirror finish doesn't have a copper color, in fact it looks like the original metal!

Two reasons:

1. I focused on the reflected image, the digital camera I used can be put within a couple mm of the object, and focusing through the depth of the photo gets harder as you get that close.

2. A really good mirror finish is 'almost' colorless due to the reflection of light off the surface - the light reflects the images, not the surface characteristics of the metal.

Here is a copper cold plate with a 0.02 micron finish reflecting the image of the computer monitor behind it. You can see the copper coloring fairly easily on the cut (non-polished edge), but the surface is mostly reflecting the mirror image instead of the metal color of the surface.
http://www.wsu.edu:8080/~i6735189/cp...reflection.jpg

Kewl:cool:

We'll note it as: if you have a mirror finish, then you can't see the copper color, and that it's a reflection of the quality of the lapping.

Many thanks!

Here is another test version at an earlier stage of my block. Top sliced down and a presure chamber(forced channel flow) with directional nozzles incorporated as the focus of attention here:D
Comments appreciated.

My development page

G F,

That block has lots of turbulance I'm sure and also a great deal of surface area. I also like the fact you use a plexy cap to it, eye candy is nice.

One concern however, what does it weigh?

How have your performance tests turned out? Figures?

It's all on the website. Very nice!!!

What was the progress on cross drilling holes into it?

g_f
sorry to hop back into the P4 sanding subject, but several of you might be interested

the IHS is the Marketing Dept.'s name for a CPU protective cap
made of thin copper, nickel plated after forming for corrosion resistance (discoloration), adhesively bonded to the package perimeter with epoxy

-> the protection is afforded by the relatively stiff perimeter which is lower than the central area and serves to limit the application of off-axis (angular) loading of the CPU

the 'IHS' is too thin to contribute significantly to lateral heat dispersion (run the calcs - Waterloo),
and is designed to deflect under the applied load from the spec hsf to a 'flat' plane

-> under the 'IHS' is a PCM insert, there is a 40°F offset due to the IHS plus PCM
(Intel is willing to take this ‘hit’ for the protection afforded by the ‘IHS’)

this info from an ex-Intel thermal engr with whom I have some dealings

now consider what the effect is of sanding 'off' that bump in the center of the IHS
-> instead of a somewhat concentrated load over the CPU area, the load is distributed and born also by the edges that were intended merely to stabilize the hsf
-> the compressive load on the PCM is significantly reduced and . . . .
stretch a bit, read up on what PCMs need to work well, temperature and . . . pressure !!

I’m sure everyone will report lower CPU temps after sanding,
because we see that which we wish to see
- no one here read the article about the DIYers who know more than Intel ?

mercy, tough crowd here

I read about the intended deflection. I know I know, can't know more than the designers who made the darn thing. I do not believe that it goes to a flat plane. The paste residue was always heavy around the center which means it always was never achieving a flat plane (perhaps because it was not designed to be used with paste?) Clear marks made by the edges also, to the point of marking the base so definate load was at the edges, not just a balancing act. Have not noticed a detrimental effect. The pressure of the PCM, now that you brought it to my attention, bothers me. I think I will sand the edges down now that you have enlightened me:) Did that engineer mention anything about how many times it could be mounted before the phase change material looses function? And also is the 40F offset including the PCM of the stock HSF also (it's used there too)?

no, the 40°F was for the IHS plus its PCM
I too seriously doubt the entire surface deflects to a uniformly flat plane, but it doesn''t need to
a greater load in the center is benefical

PCMs deform on the first (several ?) cycle(s)
then the goal is to keep them undisturbed, a good app for an IHS PCM

bear in mind that no quantity of heat is being convected far from the CPU, because it can't get far from the CPU with the thin IHS

view the top of the IHS as a flexible copper membrane

and yet again g_f, what is the actual measurement capability ?
listen to your own words, eh ?

I think you can tell by this pic that the edges were in contact with the center with just a little sanding on the left. Even after further sanding the same amount was removed frof the edges as in the middle. The pic on the right shows the copper exposed in the center and the edges but the plating is still on the belly and that is how I left it (good thing too). Some creative sanding by me will remove some more from the edges alone to get some central presure for the way it was designed to be. Thanks for the info BillA.

The weight is not a factor, it is a lot lighter than the Spir@l block that I have.

This got me thinkin:eek:

As you can see from the lapping pic my edges were already even with the center so no more deflection was ever going to happen. This may be from this deformation of the PCM? The photo you can see is a different pattern than N8's, his is all copper exposed at the center with a little sanding. Perhaps he did his a lot sooner with less mountings than I had mine? If this is the case then all I did was thin out the IHS, expose the copper over the core and get rid of the neat etched in lettering which was hogging up some AS3.
Man I love to speculate but I have the photo to show where my thoughts are coming from and the photo from N8.:)

one would have to speculate about the P4 tolerance stack, I've no info like that

PCMs 'overheat' and then take their 'set'' on the first cycle, or several depending
in any case I would think the higher the center, the better;
for thats just where the pressure is desired

wish I could lap like N8

Although should ,possibly ,not be completely disregarded :
"green Diamonds" and "green squares" on "Christmas Tree" http://www.jr001b4751.pwp.blueyonder.co.uk/Catharbp.jpg or on http://www.jr001b4751.pwp.blueyonder.co.uk/Cathar14.jpg

Les
in your garland of many colors there is a assumption which I'm pretty sure is not valid
i.e. that the conductance across the TIM joint is uniform across the face of the IHS

this CANNOT be true
both in the 'unsanded' state as well as the 'sanded' one
- due to the IHS's flexibility and lack of support away from the CPU area, and away from the IHS's edges
- and note also that once sanded there is a thickness variation to deal with as well

this could all be modeled I guess, but not by me Thanks

Yes it is the "Uniform Flux" case.
However to illustrate that the spreading should not be automatically completely ignored the graphs may serve.
Waterloo do both "Parabolic" and "Equivalent Isothermal" for circular heat sources sources but not "rectangles with end cooling". I have no intention of pursuing.

no argument that there is some contribution;
but when 'large amounts' of residual goop are reported, then one has to wonder

BTW, I suspect the circular parabolic might not be so bad,
but what to use for 'r' returns us to the same point

It hit me in the face. What I observed (large goop around the core) is not a bad thing. What do I mean? The IHS is thin and pliable. So? the portion above the core is stiff because of the core underneath. So? The thicker application of goop is forced away from the center because there is no give there but as it collects around the core from the mounting presure, the unsupported region around the core will give way. If you follow then a thicker application will result in a thicker collection of goop around the core area and bend the IHS under the mounting presure and as a result apply more force to the PCM material under the IHS.
Am I deluded or did that make any sense?

5 x 5 (loud and clear)
thats how it works

- but the area away from the die is not doing much at all
get used to it, many more IHSs on the horizion

The area away from the die is not transfering any significant amount of heat, agreed, but it is needed to apply presure to the core. Lay a piece of seran wrap over a solid surface, then grap the edges and pull down. The seran wrap stays at the same level over the solid surface but force is still applied downward even if the surface does not move.

Now a thinner material over the core should be a good thing no? And what about the copper exposed instead of the etched plating?

Valid points. I don't (didn't) see the IHS flattening out in my case before I decided to lap it.

I started out with my 1.8a and the stock Intel heatsink. I cleaned off the TIM material on the stock Intel heatsink and used AS3. I believe one of the reasons for Intel making the 'peaked' IHS is as Bill mentions, to lessen off-axis loading, and also because they intended that their heatsink and TIM (thicker than paste) would be used. Even with the TIM removed from the stock Intel heatsink, I had to use a LOT of pressure on the stock attachment arms (pretty idiot proof design, BTW) to lock them in place. My mobo did NOT look good bent like that once the pressure was on the heatsink. I have never seen a mobo bend to that degree. I reseated it several times, but it was the same each time. I said, "screw it, must be what they intended", and fired it up. I ran it for about 4 days or so with the stock heatsink, as overclocked as I could get it to run RC5. The thing ran pretty hot, 55-70 C. Remember, this was with an extreme amount of pressure on the CPU/heatsink.

When I started testing the CPU with air cooled water and then phase change cooled water, I took the stock Intel heatsink and lapped it. It was very high in the center as well. It took an incredible amount of lapping to flatten it out. Between the rounded IHS and the rounded stock Intel heatsink, there was no way to side load the damn installation since they were basically two rounded surfaces. The AS3 interface pattern showed that they were only touching at the center, no flattening had occurred, at least not in respect to the whole IHS. I re-installed the stock Intel heatsink after lapping it, and the mobo bending was not as bad, although it still looked more bent than I liked. I ran it for several days again, and temps were better, but not by much. Could be just the fluctuation in room temps.

When my MCX4000 arrived, I decided to check the cpu itself for flatness, since the MCX4000 had a very flat base and the mounting method would suit a flattened IHS better. After seeing how un-flat the P4 was, I decided to lap it. This was not a hasty decision, and I thought through some of the reasoning that has already been brought up by others in this post. Seeing as how hot I had run the CPU and the amount of pressure that had been on it, I didn't see how the IHS could flatten out any further, if it was even supposed to. I re-tested with the lapped Intel heatsink, and temps were still about the same. This would support Bill's statement that the IHS does not substantially affect lateral heat dispersion, which makes sense anyway, since the core is fairly small compared to the thickness and dimensions of the IHS. Having both the IHS and the heatsink peaked at the center of the heatsource is a good design consideration 'for the masses.'

Since we are not 'the masses', but enthusiasts seeking better cooling, we modify and consider design advances to enable better cooling. Since I knew I was going to be installing a flat heatsink with mounting through the 4 mobo holes, AND I knew I had 'run-in' the cpu with a very high amount of pressure and heat, AND I have years of professional lapping and polishing experience, I felt comfortable lapping down the IHS, since I felt it was at it's final resting place. I did not do this solely for cooling reasons, but also for the amount of bending that my mobo was under. I did not like it bent like it was, and the only thing left was to flatten the IHS, which was very high in the center still.

Now with the MCX4000 and a quiet 80 mm fan, the P4 is running a full 10C cooler than with the stock Intel heatsink while under full load. Did the IHS lapping do anything? Possibly a degree to two better cooling, but in my case, my mobo is not under a huge amount of stress anymore. Maybe I just got one of the thicker outliers to the variation of PCM thickness. I don't know how much variation Intel has in this application of the PCM.

As far as recommending lapping the IHS for cooling purposes? I don't think it really does that much to help cooling, and I wouldn't recommend doing it unless you have some additional considerations or reasons. In this respect, I guess I was a good guinea pig.

N8

given any thought to removing the IHS ?
(I'd love to get a 'handle' on that 40°F number)

Yea, I have thought about it. It will have to wait until I can get a C1 stepping for under $200. Maybe in a month or two. That way I can experiment on this 1.8a.

I am in the middle of renovating a house to move into right now, so I don't have any spare time until mid- December or so. I have to get as much done to this house before it starts to get too cold or snow too much. Managed to get our first house through a closed bid auction/bank repo. Needs a lot of work though.

ah, the landed gentry now, eh ?
congrats

PM me or something when you do it, am interested - but not always here

Here is another view of these first and second stages. You can see the nozzle and gasket arrangement. The central gasket is placed at the proper depth with a shelf milled into the second stage.

Further additions to my development page.
Notice the very center of the bp and also the central nozzle extrusion.

He is a profesional of lapping, look at his web pages.