Once upon a time on usenet Paul wrote:
~misfit~ wrote:
Once upon a time on usenet Paul wrote:
[snip]
So, we need an excuse for why it would get hot...

1) Heatsink isn't sitting flat on the CPU.

Or the TIM underneath the lid is defective - highly unlikely on
the LGA775 ones, as they used low temperature solder and the lid
s soldered to the silicon die. It's one of the best TIMs there
is (in terms of not being defective). Modern processors, like the
very latest Haswell, have switched back to "cookie dough" style
internal thermal interface material. Some flavors of cookie
dough, are not uniform from one section to another. Or even have
voids in them.

Intel switched from the low temperature solder (so they claim),
because the low temperature solder is a "conflict mineral".
I think they stopped using it, to boost the profit margin :-)
Solder is more expensive than the cookie dough they use.

This is really interesting thanks Paul. (I stopped reading tech /
hardware news / sites etc a while back as my financial situation
went even more downhill and reading about the latest stuff made me
miss not having even close to 'cutting edge' tech any more. Hi, my
name's Shaun and I'm a hardware junkie....)

I've got a socket 775 E7300 that's always had one core running
around 15º C hotter than the other. I spent quite some time messing
with TIM, bought an overkill cooler and even lapped the IHS and it
made no difference. In the end I put it down to bad TIM between IHS
and die. I tried to remove the IHS (like I used to with Tualatins
when I was O/Cing them) but it didn't move with as much force as I
cared to exert trying. Now I know why.


Sorry for the late reply. For some reason I've not checked this newsgroup
for ages.

One of the enthusiast sites showed how the users
there remove them. This is an example.

http://www.overclock.net/t/305443/ih...-and-the-facts

"The solder Intel uses roughly melts at 80-90c so it could take a
while."
If that were true, boiling water would remove it :-)

Hmmm. From what little research I've done 'In52' solder (Indium 52% Tin 48%)
would probably be best for attaching an IHS to a silicon die if you were
going to use solder. It has a melting point of 118º C and has a good
'wettability' of glass, quartz and ceramics. Also I think 118º C is above
'THERMTRIP'? Even though the IHS is glued on and isn't about to fall off you
wouldn't want the solder 'sagging' due to gravity in a tower case, migrating
away from one side of the die.

The solder with the lowest melting point is 'Cerrolow 117' which melts at
47º C. There are a lot of solders in between those two;
http://en.wikipedia.org/wiki/Solder#Solder_alloys

http://www.legitreviews.com/intel-ih...775-cpus_402/2

"the extra solder on the core was scraped off using a plastic
credit card" "The system would power on, but sadly it wouldn’t
post."

Heh!

Just a small failure rate problem. Good if you start
with a bucket full of processors.

But if you have the money for a bucketful of processors why not just buy a
faster one?

I like the lapping job they did on the lid in the second article.
They claim it only takes 20 minutes to do that, but if I was
doing it, I'd need to start with the belt sander, to be
able to finish off with 2500 grit in 20 minutes :-)

I first lapped the IHS on my E4300 a few years back, my last real overclock
to speak of. I managed to get it Prime95 stable with a 50% overclock for a
year. It took quite some time. I didn't clock it but I'd say a couple of
hours (not counting the breaks to rest the tired arm). It was massively
dished with high corners and lapping it resulted in about 5º C temperature
reduction under full load overclocked, to around 58º. Before that it was
over 60º and I wasn't comfortable with it.

After a year I bought a faster 45nm CPU (the E7300 mentioned above). Since
about then I've had no need to overclock as CPU performance isn't the
bottleneck it once was and powerful CPUs are affordable too. However I still
ended up laping the E7300 due to the 15º differential between the two cores.

These days it's I/O that holds things up. Newer CPUs can do all that I'll
ever need (and much much more) so my days of overclocking are behind me.

Cheers,
--
Shaun.

"Humans will have advanced a long, long, way when religious belief has a
cozy little classification in the DSM."
David Melville (in r.a.s.f1)