Mounting Considerations
 

Here we discuss Mounting Considerations, which may include, among other things, details of the very challenging TIM joint.

How about mounting pressure and tim joint application. Depending on how anal you want to be maybe all the testers should use something more consistant then a paste. Also mounting pressure can have a significant effect on your C/W value as you can gain or loose a 1C with just a few pounds of pressure. Or am I wrong?

your definately right; there are a lot of things to consider; the lapping of the waterblock, what thermal paste should be used for repeatability (will AS3's setting affect results?) and mounting pressure flat on the die.

Any of these can drastically affect temps.

We need to do like BillA on this issue, NO LAPPING should be performed by the WBTA ever! This is up to the mfgr's/DIY's to do and the blocks should be tested as they are received. I think the mounting force is easy to overcome, each bench will have the same springs and use the same amount of turns once the nuts contact the springs, like 12 full turns for example. The TIM application is another beast altogether but, maybe a specific technique can be 'standardized' and closely followed obviously all need to use the same paste.

Can anyone recomend a place to get springs? My local sears has plenty but they are all to large and or they are the ones that are already compacted, IE designed for pulling not pushing.

Can’t we just use a torque wrench to match an agreed pressure? E.g 16 psi :confused:

MsMaster ( www.mcmaster.com ) has some good info for springs, and is an excellent source.

How reproducible would a torque wrench be?:shrug:

I used torque wrenches in car engines where the torque on some bolts need to be the same .E.g. the bolts on a given engine need to be at 150 PSI. How do you know that the 6 or 12 bolts are at 150 PSI? You use a torque wrench. I have seen two types; one with a dial and the other where you set the torque and the wrench slips when that torque is reached. This would also eliminate the need for things such as “thread count” that can be highly unreliable. The type of spring would also not matter, as long as the spring can provide the PSI desired. E.g. you wont use a spring capable of providing only 10 PSI when 16 PSI is needed. So you could have a 25-PSI half compression spring and I could have a 30-PSI half compression spring it would not matter using a torque wrench because the setting would only allow – lets say 16 PSI.

Torque wrench won't give you a pressure/PSI reading. Strangely enough, torque wrenches are for torque, measured in foot-pounds or newton-cm or whatever. To determine a thrust (pounds, not PSI) you'd need to know stud radius, thread pitch, and friction (both static and dynamic). Accuracy will suck and isn't worth the effort.

A better method would to be directly measure the springs. Load the spring with X pounds of compression and measure the resultant length with a micrometer. When mounting a block, compress the spring to that predetermined length.

An off the wall thought I've had that might make the pressure / mounting / TIM question less of an issue would be to change the design of the CPU die and fixture. This might also be a good way to cut down on secondary path losses.

As I understand it, all the current simulators attempt to more or less reproduce a CPU sitting in a socket, and mount the WB like it was sitting on a real CPU and socket.
This is good for 'looking like the real world' but seems to make life much more complicated.

Instead, what would happen if one made the simulator so that it consisted of a hollow outer shell that more or less matched the outer dimensions of a CPU and socket on the outside, and the dimensions of the heater die on the inside.

This shell would be made out of a rigid material with good thermal insulation properties (Phenolic?) It's outside surface would be designed to provide mounting for a variety of WB designs.

The WB would be solidly clamped across the face of the shell, using enough pressure to ensure that it will NOT move. The heater die would then be slipped into the center of the shell and pushed against the WB's face by pressure applied from behind it.

Advantages:

1. The shell provides a constant position for the WB face, and since it basically is sealing around the heat die, would insulate against secondary losses.

2. Because the surface the die is pushing against is always in the same relationship to the die, repeatable pressure should be easy to accomplish

3. Because the die and the shell are seperate parts, it should be possible to accomodate new form factors for processors or mounting types just by changing the shell, and not having to replace the more difficult to make and expensive heat die.

4. Concerns about torque loads etc. produced by plumbing connections are minimized.

5. Simpler mounting, could be made to work with multiple sizes / types of mounting hardware.

Disadvantages:

1. Slightly less match to 'real world'

2. Might have difficulty with blocks the same size or smaller than the heat die (I believe these are quite rare however)

I think #2 is rare enough to not be a problem, and the advantages in consistency and repeatability are enough to justify ignoring #1

Gooserider

I had to print that to read it, and I still don't get it (sorry!). Can you clarify?

Have you seen the pictures of Bill's first heat die? It's a copper slug that includes a heater, and protrudes out of a phenolic resin wall, to simulate a CPU die.

That's essentially the route that I'm going. There's nothing simple about it though: if I decide to use another die size, I have to replace the slug, which makes up the die, and includes the heater.

Maybe a diagram of your idea would help?

Ben-
Seems to me that he is talking about instead of the inconsistancies of various mounting mechnisims for the block side mount the cooler/waterblock to a fixed static position and set the pressure for the die/heatslug to 'rise' to the block...

Seems like it would equalize the mounting varibilities, but would create a large problem of making an insulated and movable heat die.

Don't feel bad, I didn't do a great job of describing what I had in mind. The picture in my head is pretty clear, but it's hard to turn it into words.
Will try. Essentially there are two parts to my proposed setup:

1. An outer sleeve that simulates the socket & CPU, and provides a mounting surface for the WB under test. It should be made of non thermally conductive material, and be insulated to the greatest extent feasible. The WB is mounted to this sleeve in a manner that is as rigid as possible, giving a constant location for the base's die contact area.

2. A heated 'die core' that fits snugly inside the sleeve, This would have the surface which contacts the WB, along with the heating and temp monitoring components. It would have a consistently repeatable mechanism for applying a constant pressure against the test WB's base (Possibly a cross bar holding one end of a spring, with the other pushing on the die core.)
There would be no insulation on the core, but it would be necessary to insulate the back side of the hole that the core fits into.
Yes I have, it certainly works, but seems to me like something excessively complicated.
Agreed, and it seemed to cause no end of grief in terms of repeatability in mounting (Witness the stated need for doing multiple mounts to ensure repeatable results)

I'm suggesting a design that sacrifices some slight level of 'real world' duplication in an effort to get a more repeatable mount that would reduce or eliminate the need for multiple iterations of the same tests
Note that my design would also require a new slug IF you change the DIE size. However if you could keep the same die size, it should be possible to just change the sleeve, and keep the die.

Some of the earlier discussion seemed to be suggesting standardizing on the die size. If this is done, then it would work well with my proposal, as the die, which I see as the most critical part of the WB setup stays the same regardless of the sleeve that is used to match different form factor CPU's.
[/quote] Maybe a diagram of your idea would help?[/quote]
Difficult as I don't have an easy way to attach a drawing, but will try some ASCII art...
Note that the slug is shown spaced away from the WB for clarity, normally the two would be in contact with a known pressure. The slug would be a snug but sliding fit inside the sleeve (note, enough allowance must be made for thermal expansion differential to ensure this is the case at all operating temps.)

I think you have it Yoshana.
I'm not sure it would be that big of a problem -

1. The top is the WB contact area, no insulation needed.

2. The sides would be insulated by the sleeve material. so the block itself doesn't need insulation.

3. The bottom would need some insulation, but it shouldn't be that hard to make an insulated cover for the bottom.

I don't think it would matter that there was a small amount of airspace along the sides and bottom of the slug. It might take a few moments longer to reach equilibrium, but once the airspace temp reaches the same as the block, then you shouldn't have any further secondary losses other than what is allowed to filter past the sleeve insulation.

Gooserider

You know, that's not a bad idea. Thanks for the clarification!

What I like about it is that it allows the slug to sit up against the block with a certain flexibility.

What I don't like is that I was considering using a vacuum to insulate the slug, but I might be able to work with it.

Hmm... wierdness - I replied a couple minutes ago, and it dissappeared - below is a recreation...
You're welcome.
Agreed, not sure if this is a good thing or not, but I don't see it as a major issue either way.
I don't see a problem - The WB would seal the top of the chamber, the sleeve makes the sides, and it should be simple to extend the sleeve far enough to put a sealing plate over the back end, then evacuate the entire space.

However as designed I don't know how much help the vacuum would be. Since the sleeve and the slug would be in close contact, there would still be alot of thermal conduction losses.
A minor redesign could probably handle much of this though. Obviously the sleeve would need to be made out of poor thermal conducting materials. Second get rid of most of the material touching the slug, leaving just enough to work as a guide. (Possibly a honeycomb pattern of some sort?) It should be possible to minimze the losses that way.

Gooserider

Hmmm... maybe I could use an o-ring here... I'll have to give it some consideration.