Intel Skylake CPUs are bending due to cooler mount pressure

[pollaxe]

My Z77 board died on me a few weeks ago which made me bite the bullet and order up an i7-6700k system to replace it. After being messed around by Ebuyer (who I'll never use again) I got my CPU from Scan but as it doesn't come with a cooler, I ended up reusing the only third party spare heatsink I had left.

I think it's a Thermalright AXP-140 (low profile) which is pretty hefty and has a later socket bolt-through fitting system and backplate. This was going to be the rig I used at LANs but now I'm feeling a bit twitchy about it. I'm yet to have any issues but I wonder if I should possibly get something a bit more recent (and perhaps a bit lighter.)

[gupsterg]

Glad I never waited for Skylake and went ahead with Devil's Canyon build .

[Corky34]

Seems odd to blame Intel though when only a single type of mounting mechanism is causing this.

[ik9000]

Well, section moduli for a rectangular substrate is proportional to depth squared for stress and depth cubed for deflection, so halving the substrate depth will lead to four times the stress and eight times the deflection for a given load case. And that's for homogenous material. For anisotropic laminates it gets worse and the thickness and spacing of layers in a given direction comes into play. You gotta hope someone checked this stuff out at the design stage...

[gupsterg]

"Intel confirmed that Skylake uses a thinner substrate but it is rated for the same 50lb. maximum static load as prior generations"

They say that, but the evidence is suggesting otherwise..

I concur , unless my reading skills are not functioning .

Info I used from page 110 of 4th gen CPU datasheet.

[Corky34]

Scythe probably shouldn't have exceed that 50lbs then.

I have to conclude that part of the success of this heatsink comes from applying more than a 60 lb. pressure, but only part.

[ik9000]

I concur , unless my reading skills are not functioning .

Info I used from page 110 of 4th gen CPU datasheet.

interesting that the 6th gen (skylake) datasheets don't have this info anywhere in them! There is no guidance in the 6th gen document (nor any of the referenced documents in 1.7) to state which loadings the CPU and HSF are designed to take. The 5th generation has some data, but nowhere near as comprehensive as the guidance for 4th gen linked above.

5th gen datasheet (section 8.2)
6th gen datasheet (Section 8.2)

So Intel give no data on the design loadings for Skylake!
Broadwell has limited data, but states 15lbf static normal load for no backplate, 25lbf with a backplate
Haswell has the same 25lbf static tensile load, but also gives data on shear, torque, and the pressures the socket itself (i.e. the pins) can withstand.

Given this 25lbf value, I'm not sure where the 50lbs referred to is coming from.

Could the problem be that intel aren't (obviously) sharing the right information with designers? The data in the 4th gen is far more useful to a mount designer.

Large heatsinks won't just impart compression to the motherboard - the offset center of gravity and weight will impose bending moment into the mount location.

[Corky34]

Given this 25lbf value, I'm not sure where the 50lbs referred to is coming from.

At a guess as i can't find any specs direct from Intel, i would say it's because that's what Intel said when asked to comment on the situation.

[ik9000]

At a guess as i can't find any specs direct from Intel, i would say it's because that's what Intel said when asked to comment on the situation.

i get that but this statement:

Intel: said that it has only been made aware of the issue in the last two days and it is therefore investigating what "could be several variables at play". Intel confirmed that Skylake uses a thinner substrate but it is rated for the same 50lb. maximum static load as prior generations.

contradicts intel's own spec sheets for previous generations which were only 25lbf. Something doesn't stack up!

[Corky34]

I could have read things incorrectly but going on the info in the 4th gen CPU datasheet PDF that gupsterg posted a link too, i read that the maximum static load is 135lbf so even that doesn't make sense to me.

The 25lbf that you're referring to appears to be the tensile loading force that they define as "a pulling load applied to the IHS in a direction normal to the IHS surface.", then there's the 70lbf shearing force "a load applied to the IHS in a direction parallel to the IHS top surface", and finally the 35lbf-in torque defined as "a twisting load applied to the IHS in an axis of rotation normal to the IHS top surface"

That's left me totally confused as none of those mention 50lbf

EDIT: They also say "The processor package substrate should not be used as a mechanical reference or load-bearing surface for thermal and mechanical solution" so I'm not even sure using the processor spec sheet is correct, should we be looking at the sockets mechanical loading specs?

[ik9000]

I could have read things incorrectly but going on the info in the 4th gen CPU datasheet PDF that gupsterg posted a link too, i read that the maximum static load is 135lbf so even that doesn't make sense to me.

The 25lbf that you're referring to appears to be the tensile loading force that they define as "a pulling load applied to the IHS in a direction normal to the IHS surface.", then there's the 70lbf shearing force "a load applied to the IHS in a direction parallel to the IHS top surface", and finally the 35lbf-in torque defined as "a twisting load applied to the IHS in an axis of rotation normal to the IHS top surface"

That's left me totally confused as none of those mention 50lbf

EDIT: They also say "The processor package substrate should not be used as a mechanical reference or load-bearing surface for thermal and mechanical solution" so I'm not even sure using the processor spec sheet is correct, should we be looking at the sockets mechanical loading specs?

My reading of it is that first section with the 135lb is the force rating for the contacts on the CPU themselves. The 2nd table is the one regarding the force the heatsink should impart to the CPU. the 25lb normal force is the tensile rating. Torque is twisting of the heatsink about the orthogonal axis ie put the thing flat on the table and twist it like a screwdriver. The shearing force is trying to slide the thing laterally off like sliding a box across the floor.

[Corky34]

The only place i could find a reference to 50lbf is in the technical specs for LGA 1150 (section 4.2 table 5) where they mention "Heatsink static compressive load" maybe that's where the 50lb comes from.

[canopus72]

Just goes to show that bintel cpu's are over-hyped aand over-priced. Bintel skimping on quality control yet again. Damn crooks.

[ik9000]

The only place i could find a reference to 50lbf is in the technical specs for LGA 1150 (section 4.2 table 5) where they mention "Heatsink static compressive load" maybe that's where the 50lb comes from.

that guide is for the socket, not the CPU substrate.

[Corky34]

Yes i know, that's why i mentioned that in the datasheets for the CPU they say "The processor package substrate should not be used as a mechanical reference or load-bearing surface for thermal and mechanical solution"" And then later highlight that the only place i could find the afore mentioned 50lbf was in the datasheet for the socket.

My deductive reasoning would seem to indicate that the 50lb figure quoted in the Tom's Hardware article was in reference to the Heatsink static compressive load on the socket, seeing as the CPU substrate should not be used as a load-bearing surface, no?

[ik9000]

My deductive reasoning would seem to indicate that the 50lb figure quoted in the Tom's Hardware article was in reference to the Heatsink static compressive load on the socket, seeing as the CPU substrate should not be used as a load-bearing surface, no?

You say yes, I say no, she says why? I don't know know know. You say goodbye, and I'll say

Hello.