Intel Skylake CPUs are bending due to cooler mount pressure

[Corky34]

You've lost me there, not sure what the Beatles & Adele have to do with a processor package substrate.

[gupsterg]

I thought when I looked at the 4th gen datasheet originally to evaluate if the ThermalRight Archon SB-E X2 was ok to use (mounting has adjustable 40lbf-70lbf) I had cross ref'd correct information. Through the discussion here I'm now not sure at all .

Each table is dealing with a different aspect, ie page 110 table 58 of 4th gen datasheet is just the CPU package, ref'ing note 3 of table. The CPU package is defined as including substrate, see text page 109.

Next my thoughts on text Intel places above table 58 :-

The processor package substrate should not be used as a mechanical reference or load bearing surface for thermal and mechanical solution.

is:-

a) to state IHS must be the load bearing surface for thermal and mechanical solution.
b) creates confusion to the whole "data" as the IHS must transfer load to TIM/die/substrate and ultimately the substrate bears the load plus transfer to socket/mobo PCB. I do not know the technical points of this "load transfer" element and look forward to a post by someone explaining.

Table 59 on page 110 of 4th gen datasheet is just dealing with load bearing aspect of IHS relative to substrate. Best example of the shear for me is delid, delid tool by der8auer. This is where IMO a compressive load should be detailed and isn't.

Page 25 table 5 of LGA1150 socket application guide is only dealing with socket, ref'ing note 4. The ILM / heatsink form part of data table but they are not using a CPU package in this data IMO. So this data then creates further confusion.

To me it explains 70lbf is minimum ILM will exert to seat CPU package. The maximum value the ILM can exert I think is worked out by deducting a predefined value of heatsink with a 50lbf compressive load deducted from total socket can take (185-50=135).

Then you have this text above table 5:-

The socket body should not be used as a mechanical reference or load-bearing surface for thermal solutions.

and I feel & me to understand please what can be a load bearing surface!!!

Next above table 5 we have text:-

The socket will be tested against the conditions listed in Thermal Solution Quality and Reliability Requirements Chapter of the Processor Thermal Mechanical Design Guidelines (see Related Documents section) with heatsink and the ILM attached, under the loading conditions outlined in this section.

I then go to page 30 of that guide and it's clear as mud .

The Intel reference thermal solution will be evaluated to the boundary conditions provided in Volume 1 of the processor Datasheet (see Related Documents section).

Now the 2 tables (CPU Package/Socket) use a 500g heatsink, the ref heatsink is not 500g!!!

My i5 4690K ref HSF unit is 216g inc fan/fastners/TIM.

I now believe Intel do not supply information at all regarding what the substrate can take, regardless of gen. Gen 2,3 & 4 are all identical on data, Gen 1 does differ.

Page 38 Gen 1 CPU package
Page 28 Socket 1366

[Corky34]

Glad I'm not the only one confused by the whole thing.

[Strawb77]

my AMD cpu connects via pins, so no issues of this nature there (well someone had to say it)

[ik9000]

my AMD cpu connects via pins, so no issues of this nature there (well someone had to say it)

BS that has nothing to do with it.

[Strawb77]

i was making the point that the contacts are not going to be squashed by excess pressure, as shown in the picture above (the one with the big red arrow)

[HelldiverS]

I've not been following this. if you are only using the stock/intel HSF there should be no problem. I was thinking of building a basic i3 rig.

[gupsterg]

I've not been following this. if you are only using the stock/intel HSF there should be no problem. I was thinking of building a basic i3 rig.

No issue in your case.

i was making the point that the contacts are not going to be squashed by excess pressure

It's not the contacts which is the issue, the issue being discussed is the substrate becoming mishaped from excessive pressure.

Glad I'm not the only one confused by the whole thing.

If I disregard the text:-

The processor package substrate should not be used as a mechanical reference or load bearing surface for thermal and mechanical solution.

Then basically Intel are stating the CPU package (inc substrate) can take 135lbf as static compressible load in table 58 page 110 of 4th gen datasheet.

Then I cross ref what the ILM would be exerting ie min 70lbf to seat CPU package to socket and your left with 65lbf for the heatsink to exert, but then what confuses me is where Intel state the ILM can exert max 135lbf, then technically the heatsink can only exert 0lbf as the CPU package would not take anymore.

What would be ideal is a defination that the ILM exerts say 70lbf +/- 10lbf, to me for ILM to be manufactured with a min 70lbf & max 135lbf would be too big a variation.

I also believe what the ILM exerts after a processor is seated changes, when I sold my original i5 4690K I noticed on removal where the ILM had made contact with the IHS small indents either side.

[Strawb77]

the picture definitely shows damage to the contacts- can you see what the big red arrow is pointing to?

[gupsterg]

the picture definitely shows damage to the contacts- can you see what the big red arrow is pointing to?

Agree it is showing damage , but IMO due to substrate misshape from excessive pressure.



If the CPU package was Pin Grid Array and substrate misshape you'd could get damaged pins and/or incorrect seating of CPU.

SO regardless of PGA / LGA if substrate misshape there would be an issue.

I'd just like to add I'm not an Intel fan.

I have in the past had AMD CPUs, AMD K6, K6-II, K6-III, Athlon XP (Barton), Athlon 64, Athlon Opteron 144 (Athlon 64 x2 equivalent).

[Strawb77]

ok gupsterg, now we`re getting somwhere.

i have seen this sort of damage before when i have been fixing computers for people.
in a great many instances it has been because the cooler was attached to the mobo directly, and had flexed due to ie; being moved and such (lan parties etc.)l. the coolers are generally held in by those plastic pins with an insert, which you push and twist to lock.
in comparison, amd (and i`m not saying all) have a back-plate behind the mobo, to which the cooler is fixed. because of the design the mount screws generally can only be screwed in so far, and the mechanical force is transferred to that rather than the mobo.also as far as the cooler is concerned, it rests on a mount round the cpu socket which also stops it moving around.
any weight is held by these two plates rather than the board itself.
as i say, i am generalising in both cases, and there are different means employed by both intel and amd.
i`m not an amd fanboi, i`m just saying what i have seen, and the leaves in the intel slots are vunerable to damage.
because amd uses pins rather than leaves, compression will not harm cpu or board, wheras as we can see in the photo, any slack can cause the actual cpu to `nutate`, and do damage.
i must also say that i have never once found the cpu damaged in an amd machine.
the long and short of it is; if the subsrate is thinner then that means that it is not clamped as well (as standard) it can move about, unless the `slack` (for want of a better word) is taken up somehow.

that`s all, just saying.

[gupsterg]

My apologies if at all I came across as suggesting your a "fanboi", I never meant that at all .

My intention of stating my past AMD CPUs was to show I have no brand loyalty with say Intel .

I replied to your post as wanted to further discussion, ik9000 reply to you, to me seemed may switch you off from giving input.

we can see in the photo, any slack can cause the actual cpu to `nutate`, and do damage.
i must also say that i have never once found the cpu damaged in an amd machine.
the long and short of it is; if the subsrate is thinner then that means that it is not clamped as well (as standard) it can move about, unless the `slack` (for want of a better word) is taken up somehow.

Yes, I understand this and can think this can be a reason for the issue.

SO then Independent Loading Mechanism (ILM) must have been changed to allow correct clamping of Skylake, if it wasn't then regardless of what static compressive load the substrate can take it will get damage and socket.

in comparison, amd (and i`m not saying all) have a back-plate behind the mobo, to which the cooler is fixed. because of the design the mount screws generally can only be screwed in so far, and the mechanical force is transferred to that rather than the mobo.also as far as the cooler is concerned, it rests on a mount round the cpu socket which also stops it moving around.

The backplate on Intel is only the socket.



The cooler does not fasten to backplate.

Even my LGA775 (Q6600 rig) has same design, the ThermalRight Ultra eXtreme backplate does though end up meeting the socket backplate IIRC.

On the LGA1150 (i5 4690K rig) I can't recall backplate of Archon making contact with the socket backplate, but will check on next removal.

[sybrows]

So aftermarket chavy badly designed bling heatinks are causing problems. Who would have thought it.

I concur on this..... also put something on it that cools it without it looking like a blingtastic is always good, usability and stability is more important

Brow

[gupsterg]

In other MSI news, it has been quick to act in the face of the news that some Skylake processors are bending under cooler mount pressure - as it has launched CPU GUARD 1151.

Amazing how MSI are using the image going around on the web, I wonder when someone will do more tests and release info.

I'd been searching info regarding does ILM apply min static force highlighted in Intel PDFs and so far found only this PDF to state that, Link:- http://www.intel.com/content/dam/www...lies-guide.pdf

Extract page 24:-

The ILM is designed to achieve the minimum Socket Static Pre-Load Compressive load specification. The thermal solution (heatsink) should apply additional load. The combination of an ILM and HS will be used to achieve the load targets

Does anyone think this would be the same for other sockets? currently I'm assuming it would.

This would then follow what I thought in post 40:-

Then I cross ref what the ILM would be exerting ie min 70lbf to seat CPU package to socket and your left with 65lbf for the heatsink to exert