How to check flatness of CPUs and coolers - INK and OPTICAL INTERFERENCE methods

[BoggledBeagle]

I must say I am slightly disappointed by the response this thread received so far. Out of frustration I upgraded my sorry looking (after my failed attempts to cut off the broken part) 10mm lapping glass to 12mm (15 would be twice as expensive). It is a nice 25x40 cm 3 kilo heavy piece. I also got some small 6mm pieces for in socket examination and a sacrificial CPU.

 

[67Elco]

No need for disappointment...just keep in mind these type of things are only relevant to a minority of users and those that choose to partake I'm sure welcome the input.

 

[ShrimpBrime]

Well if you want to go along those lines, this is how I check for flatness.

Get something that is flat. stand it on edge on the coldplate. Now shine a torch behind it. Now tell me how much light is leaking though between edge of object & coldplate.

After lapping a cpu and or cooler, I don't even bother to "check flatness".

The best way to actually measure flatless would be using a dial indicator to find the high and low spots. This is a procedure I've used many times outside PC hobbiest lapping.

A simple visual indicator works too, like the optical interference or using a flat edge and light.

Bottom line, nobody checks flatness when build a PC in the first place.

We are to assume the stamped copper plates are "flat enough" to apply TIM and a cooler.

I must say I am slightly disappointed by the response this thread received so far. Out of frustration I upgraded my sorry looking (after my failed attempts to cut off the broken part) 10mm lapping glass to 12mm (15 would be twice as expensive). It is a nice 25x40 cm 3 kilo heavy piece. I also got some small 6mm pieces for in socket examination and a sacrificial CPU.

View attachment 331046

That's glass, not a mirror. A flat mirror will give you an exact reflection, you know it's flat.

Glass pane like yours are probably flat enough, but unless you can measure it with your method provided, I'm not yet a believer you're on the right track and sounds like blowing smoke.

Practice what you are preaching please.

 

[BoggledBeagle]

The results are in and I must say it is far worse than I expected.

New 14500 out of the box:



Very nicely flat.


14500 in socket with mounting frame:



Acceptably flat where it matters.


14500 in socket with stock mounting mechanism:



The cpu is significantly bent in both directions, I had to use 5 times more ink to fill the void.


14500 after being removed from the socket:



As you can see the CPU is still slightly bent.


14900K after spending 2 months in the socket with mounting frame:





So the stock mounting mechanism crushes the CPUs far more than I expected. I am disappointed that the mounting frame does not maintain the CPU very flat, but at least it is deformed in the good direction.

I used the same lighting, same CPU placement and same manual settings of the camera, so the pictures are well comparable to each other.

 

[ThrashZone]

Hi,
Well if the cooler is flat all you can do is add extra timm
If you don't like that way just lap the chip.

A lot of water blocks are designed to fit concave chips.
No intel chip I've ever seen is flat unless someone sanded it down voiding the warranty of course hehe

 

[BoggledBeagle]

If you don't like that way just lap the chip.

Even when you see how the CPU gets deformed in the socket? You would need to get the same socket that is on your motherboard, and lap the CPU while being mounted in the socket. Preferably get the same two motherboards, and cut the unnecessary stuff from the motherboard intended for lapping.

This actually might be interesting for experimental type of people who destroyed one motherboard by accident and then got the same replacement motherboard with the same socket and mounting mechanism.

 

[taka]

LGA1700...

From this picture only small parts of the cpu touch the glass, damn that ILM realy crush the IHS and the cpu.

 

[BoggledBeagle]

LGA1700...
View attachment 331104

From this picture only small parts of the cpu touch the glass, damn that ILM realy crush the IHS and the cpu.

Well, and how I wrote I needed to use A LOT of ink to fill the void. In the end I put so huge amount there, that it overflowed a little on the mounting bracket, because I lost patience adding it bit by bit hoping that it will be finally enough.

What is remarkable that you cannot see any gradual falloff, just a sharp edge, then 2mm and it just plummets in dark depths...

 

[Wirko]

sanded it down voiding the warranty of course hehe

Yeah, RMA'ing won't be easy after you've destroyed the serial number and the model number too.

 

[BoggledBeagle]

ONE VERY SERIOUS WARNING FOR LGA1700 CPU OWNERS:

Choose one mounting mechanism and then use ONLY THIS SOLUTION. The IHS is not the only thing that is deforming. If you will bend the CPU repeteadly in different directions, SOMETHING may break, if not the whole chip, then something in it, or you may get a tear in TIM between chip and IHS.

Honestly when I first bent the 14500 one way and then the other way, I expected that the CPU will not work, but it does. I was expecting to hear a sharp silent snap when I was closing the stock mounting mechanism, but I was doing that slowly and carefully and it did not happen.

I just checked the shape of 14900K out of the socket and it is still the same bulge, I will never force if to change its current shape by different mounting mechanism.

 

[ThrashZone]

Hi,
Pin wise not sure why board manufactures skimp on built in backplate design like x99=2011pins and x299=2066pins has
These new sockets are almost using as many pins now days so what's missing 300 pins got to do with poor cpu board structure design.

It's not like the boards are cheap priced lol

 

[BoggledBeagle]

PGA CPUs do not get deformed in sockets, they are held in the socket by the pins. I cried a little when AMD abandoned them... I really liked them.

 

[ShrimpBrime]

So when are you going to actually take measurements?

How much difference from center to an edge? 32 thousands of an inch or what?

You haven't given any numbers. It's just ink on a piece of glass that we are not sure is actually flat.

Is there no proper testing?

We just assume it's not flat because ink and glass?

 

[BoggledBeagle]

So when are you going to actually take measurements? ... You haven't given any numbers.

I never said I am going to make any measurements, this thread was meant as a presentation of some methods and inspiration for other people. Honestly I think that with the ink you can easily check flatness and also get a very nice visual representation of a flatness of the whole surface.

It's just ink on a piece of glass that we are not sure is actually flat.

Do you want me to get the optical flatness standards to check, if my pieces of float glass are flat? For the ink method they are flat enough, trust me.

Is there no proper testing?

What do you mean? The interference method is an industry standard.

BTW Steve from Gamers nexus mentioned that they got the laser measurement machine so we will get some numbers from them soon.

You do not need a very expensive machine to get a nice picture about (and of) the flatness of a surface, a little piece of glass and a drop of ink will suffice. They are almost free. BTW those 6mm thick small glass pieces I got seem to be perfectly adequate for the job.

 

[ShrimpBrime]

I never said I am going to make any measurements, this thread was meant as a presentation of some methods and inspiration for other people. Honestly I think that with the ink you can easily check flatness and also get a very nice visual representation of a flatness of the whole surface.


Do you want me to get the optical flatness standards to check, if my pieces of float glass are flat? For the ink method they are flat enough, trust me.


What do you mean? The interference method is an industry standard.

BTW Steve from Gamers nexus mentioned they they got the laser measurement machine so we will get some numbers from them soon.

You do not need a very expensive machine to get a nice picture about (and of) the flatness of a surface, a little piece of glass and a drop of ink will suffice. They are almost free. BTW those 6mm thick small glass pieces I got seem to be perfectly adequate for the job.

The lighting and ink is a great idea used times before you know. This isn't new, what you are doing.

Everything involving a science needs measurements.

Look at the picture, got some lines with the light and ink.... tells us nothing at all.

It's OK, I should not be getting on your case. It's obvious you're learning. My apologies, I'll try not to be a dick from here out.

 

[BoggledBeagle]

BTW to visualize larger gaps, you may want to use less dark ink, for example blue or green.

I have the bent 14500 with the stock ILM in the computer now, before I remove it I may try to use less dark ink and also slit some strips of paper under the glass to see, how many will fit in the middle. That it about all I can do.

 

[ShrimpBrime]

BTW to visualize larger gaps, you may want to use less dark ink, for example blue or green.

I have the bent 14500 with the stock ILM in the computer now, before I remove it I may try to use less dark ink and also slit some strips of paper under the glass to see, how many will fit in the middle. That it about all I can do.

You have a bent 14500?
How is this bent!!??

ONE VERY SERIOUS WARNING FOR LGA1700 CPU OWNERS:

or you may get a tear in TIM between chip and IHS.

THIS I can tell you that all these chips are soldered, there is no paste between the IHS plate and "the chip/s"

As far as I've encountered removing IHS plates from modern processors, Athlon 200GE and Athlon 220GE did not have solder. The IHS plates where pasted.

The only time I can think solder would crack is when running LN2 or LH. And even then, there would be no way to know except for when you suddenly have 1 or 2 cores 15c hotter than the rest. Which Inhave experienced with a couple FX processors some years back.

I've used Dry Ice cooling on my 13700K and the IHS plate is fine, core temps all seem normal. And that's v-core of 1.65v and up.

 

[BoggledBeagle]

The Banana and the banana maker left the socket and The Bulge returned!

So, after 24h in socket with the stock mounting mechanism, you can see the amount of ink needed to fill the void:



The result in socket with blue ink:



After being removed from the socket:



I remind that the photography conditions and camera settings have not changed. You can thus compare the out of the socket state after spending just few minutes in the socket with 24h is the socket:



Finally, you can compare The Banana with The Bulge:



Remember, The Banana spent just one day in the socket and The Bulge 2 months with one day relaxation out of the socket.


This concludes my testing.

I encourage others to examine CPU and cooler flatness, research the best ways and materials to mate them and let the heat flow as easy as possible.

 

[67Elco]

I'm wondering why anyone with Intel cpu's are not using frames instead of the factory retention brackets.

 

[BoggledBeagle]

I'm wondering why anyone with Intel cpu's are not using frames instead of the factory retention brackets.

Well, the normal mounting mechanism works, the CPUs not only bend easily but also are very easy to cool, so even with a "little bit larger" amount of thermal paste the temps are good (provided that you limit the insane power draw some CPUs have out of the box).

People wanting to squeeze the maximum from their CPUs welcome the cooling improvement with the mounting frame, and technically inclined people like me simply like the frame as a better way of mounting the CPU.

If you do not care that the CPU is "a little curved" and you are not chasing each degree of temperature, you can simply ignore this problem.

 

[sneekypeet]

While I apprecite the work and effort here, one thing is being left out of these tests.

Proper mounting pressure.

While I am sure you are pressing hard in your glass tests, CPU coolers exert a ton of force onto the IHS, and could likely change what you are seeing in your way of testing.

 

[BoggledBeagle]

While I apprecite the work and effort here, one thing is being left out of these tests. Proper mounting pressure.

Well, I already wrote this in the first post:

The overall result - the thickness of the gap between the CPU and cooler base and resulting thermal resistance of this thermal interface depends on many factors:

• shape of the mating surfaces,
• flexibility of the mating bodies,
• contact force,
• mounting of the CPU and cooler,
• TIM,
• etc.

I also wrote about how it is important to be able to dial in the correct contact force and how some cooler mounting mechanisms are not very good for that.

 

[67Elco]

Has anyone ever actually tested the force that the average cpu cooler exerts on the chips? I've always wondered why the factory retention brackets were so stout when the cpu coolers provided all the real force needed.

 

[BoggledBeagle]

I've always wondered why the factory retention brackets were so stout

This bruteforcing has a lot to do with manufacturing cost. They could make a nice gentle sophicticated mounting mechanism that would not bulge your motherboard, crush the CPUs, etc, but it would be much more complicated and expensive to make. And most of the time these things are one time use only. They just need to be cheap and to work and nobody cares about parts deformed by brute forces, because nobody will see them.

 

[Noci]

This bruteforcing has a lot to do with manufacturing cost. They could make a nice gentle sophicticated mounting mechanism that would not bulge your motherboard, crush the CPUs, etc, but it would be much more complicated and expensive to make. And most of the time these things are one time use only. They just need to be cheap and to work and nobody cares about parts deformed by brute forces, because nobody will see them.

that's for example why they had a pre-set torque wrench with the Threadripper. That chip has an even bigger surface so more prone to warping (elastic deformation).