CPU clamping force
 

...coming back to haunt you, BOO!


Ok, I recently found out that AMD specs actually call for a clamping force of 12 to 24 lbs, with 18 being typical, but that AMD specifically states that 30 lb is permissible, if one uses the 6 socket tabs (aren't you relieved?).

So I came across this interesting article, a nice chart (p. 3) that describes the thermal property of the TIM joint of a TO-220 IC package (don't ask me why I looked that up, OK?), and graphs it as a function of clamping force.

The conclusion is that 20N (Newtons) is the minimum required force to properly mount a TO-220. If my calc are correct, that's 44 lbs :(

Re: CPU clamping force
 

I am not so sure the board itself would tolerate that.
Question, how would one figure out the stress on each mobo hole with 44lbs being used? Would it be 44lbs on each corner or 44/4 for 11lbs on each corner or....???? :confused:

It aint as bad as you think. You were looking at the "dry" conditions. Check out the chart under "Heatsink compound". Things get much better, even with very light pressure.

Personally, in experimenting with clamping pressure on the WhatBlock? I've found no gains in increasing pressure.

It should be as simple as "divide by 4", assuming that the load is spread perfectly equally.

The core can't take 44 lbs though. I guess my point is that there's a lot of room for improvement in that TIM joint, and that we just can't get to it :(

(Ok now ya'll don't even try to tell me that you've got it clamped to 50 lbs, and that it's perfectly safe, OK? :D )

Not I. Last time I tried that the core crushed. :D I go finger tight with the thumb screws. Which is actually pretty tight. You can pickup the mobo by the water block without it moving. :)

I think that i've given the core even more...poor TbredB

The important thing to remember is to apply an EVEN pressure on the 4 corners. You wouldn't believe how much force these XPs can tolerate.

Was it a thunderbird? If so, I believe you.

Duron 600. The new XP T-Bred's are a lot smaller cores than the original XP's and the old Thunderbird's if you havn't compared them yet. I have an original XP1600+ and a new T-Bred A XP1700+ and the cores are way different.

The original XP is about ~11.5 x 11.7mm
The Tbred 'B' is about ~ 12 x 7 mm

The thunderbird core is more prone to cracking than a new XP (Paly or TBRED) regardless of their size. Seems as if AMD strengthed the core physically :)

Original XP1600+ On left, T-bred XP1700+ on the right.
http://www.customcooledpc.com/CPU/001.jpg
http://www.customcooledpc.com/CPU/002.jpg
http://www.customcooledpc.com/CPU/003.jpg

I am not an expert on core crushing as the only core I ever crushed/cracked was intentional. Which was that Duron 600. Wanted to see what would go first, the core or the mobo. :D

With my old Duron cranking down the Maze2 made a few degree improvement, but that could just have been poor mounting.

suggest googleing before posting

as usual, you are describine your inept testing capability,
and then trying to present such as a truth

do read the technical literature, you might not then seem so ignorant

now now Bill. If the mobo's surface mount thermistor in the socket temperature probe reports no difference, then that settles it. You need to revise your fancy heat tranfer books instead! :)

Yeah actually, the graph with thermal compound on page 4 would indicate that clamping pressure has little effect.

If anything, it confirms your 0.5 C/W measurement for the TIM joint.

grrrr
I have an article devoted to just this,
it recurrs in every other article I have written on wbs,
and this jerkoff says it makes no difference

with such stupidity there is no hope of my being civil

I think it was Pope who said: "pearls before swine"

oh christ, now Null-A Ben is in
aiiiiiiiii

All right... I'll grant you that the graph is for a TO-220 package, and that the graph probably applies with a far relation, given that it's a steel-to-aluminium interface that is used, where we're actually trying to make an association with a silicon die and a copper base, the surface of which is probably radically different...

When can we expect to see this article Bill, or is it already up?

http://www.overclockers.com/articles608/

the ENTIRE issue of mounting variability revolves around 2 aspects:
applied grease thickness, and
clamping force (and uniformity)

I do not run at 90% of the (old 24lbf) AMD max by accident,
and I would add that I know of no data based on the 100lbf limit for P4s and the Hammers

(Ben, did you read that book ? - do so, you WILL like it)

Which one?

Slightly off topic: Your talking reminds me of hannibal lecter :eek:

Unless I missed it, no one has yet pointed out that 20 newtons is 4.496 pounds. At least get your conversions straight. . .

1 N = 0.2248 lbf
1 lbf = 4.4482 N

I just read BillA's article. I can't comment on that, but i did ask about the chipping problem at Ars. It seems that the CPUs before the AthlonXP (Athlon and Duron) had a ceramic substrate with much less "give" than the later CPUs' organic package, and this led to lesser chipping with the later CPUs.

Well, I'm currently assembling parts for my new system, and the one thing I made sure I ordered was an anodized CPU shim for the AMD. If nothing else, the likelihood that I will provide a reasonably even amount of force on all four corners of the core is increased.

As for how tight you should go, I would generally say that the the WB shouldn't move around on the core at all. Once you get it that tight, you should be good to go. At the same time, I wouldn't hang a DD Maze 3 off the CPU socket, especially if it's got the copper top. Instead, I would use the motherboard mounting holes (if available).

Wrong.

A shim will sit below the top of the core, which is somewhere between 0.80 and 0.88 mm, off of the substrate.

It will actually leave a gap, between the block/HSF and itself, but a smaller gap than normal: that's the protection that you actually get from a shim. It does nothing for the applied force.