micky wrote:
In alt.comp.os.windows-10, on Wed, 09 Sep 2020 06:07:13 -0400, micky
wrote:

Old computer, chapter 37.

I wanted to fill you in to how it's going, and ask some questions at the
end here.

I didn't do this earlier because things were piled on my Dell Optiplex
755 (iirc) including the laptop I am using, plugged in to various
things. .

But I've noticed that the computer will repeatedly run for 24 hours,
including playing web radio and video, but some specific actions,
opening Settings for example, makes the computer crash, and I'm thinking
if each time it loads windows into the same parts of the RAM, maybe the
exact same parts, whatever code is loaded to the bad part is what makes
it crash, so it crashes from the same set of actions of mine. ????

I tried Prime and it crashed within 5 or 10 seconds.

So I finally started testing the memory by removing parts of it.

It has 4 2gig sticks in 4 slots.

I removed two of them

Unfortunately I removed sticks 3 and 4 and it wouldn't start.

I'd forgotten what I knew 5 or 10 or 15 years ago.

So I put back the one that had been in a white stick holder and removed
the other one that was in a black stick holder, leaving sticks 1 and 3,
both white.

The computer crashed the first two time before fully loading windows,
probably beforfe the PIN.

So I took out the 2 sticks and put the other 2, that had been in 2 and
4, into instead 1 and 3. Again crashed quickly the first two times. .
So neither pair of sticks seems good.

Maybe I have two bad sticks, one per pair, and I will test one stick at
a time if it will run that way, but I figured I didn't have 2 bad sticks
and I put aside testing for that,
-----

Because the computer crashes often even before I can put in the PIN and
if not that, often before windows fully loads, I've figured it can't
overheat that quickly if the thermal cement is at all good, so that
means it's not the thermal paste. Fair conclusion????
-----

So then I went to replace the fan, something I tried years ago but I
couldn't get the old fan out. I removed 2 screws holding on a cover
and I gradually lifted up the cover, only to find it was attached to the
CPU cooling fins. So I don't know if the thermal paste was in bad
shape before or if I ruined it.

I have a little syringe of Artic Silver 5 thermal paste. It's 5 or 10
or 15 years old, stored in a cool basement. I squeezed a bit out and
it looks and seems the same as I remember it, grey, even colored, nice
looking, like toothpaste but thicker. Too old???????

"The correct amount of thermal paste, which is roughly the size of a pea
or a grain of rice." I think a pea is twice as big as a grain of rice!!
unless they mean a cooked grain of rice but a pea is still. 50% bigger.
?????

The online directions say not to let it squeeze out beyond the edge of
the CPU. But because of the heat sink, I can't see if that has
happened. What to do?

This is only for discussion. My plan hasn't change and that is to do it
the way Paul suggested and two earlier webpages had said that the dot
was the best way (and one or two said the credit card was a bad way).

But while I was googling, I found the Artic Silver page. Surely they
should know. They have different instructions for different processors
and I spent 2 days trying to find out what I had in my Dell Optiplexx
755.

I couldn't find out but I read the page anyhow.

Even if I knew if I had an i5 or i7, I woudln't know what generation I
have.(although further down I figured it out.) Do people keep track of
their generation?

http://www.arcticsilver.com/intel_ap...n_method.html#

What 's most ilnteresting is that the dot method is the he 2nd least
often recommended, maybe tied with surface spread. Least often is the
horizontal line, and most often, that is for the most differenct
processors is vertical line.

Aha, I did find my processor, Core 2 Duo, which is only listed under
Previous Generations and since it's not a laptop (all of which as far as
I noticed have Surface Spread recommeded), it's got Vertical Line
recmmended.

They have 8 pages about Vertical Line. I'm sure that 7 pages are the
same but still, that 1 page makes the difference. So why doesn't Artic
Silver sagree with the others.

(Even for vertical line, they suggest the credit card for "tinting" the
heatsink and metal cap!!!)

So I"m reading about Vertical Line,
http://www.arcticsilver.com/pdf/appm..._line_v1.1.pdf
don't know what the others say, but picture QP4 is related to the paste
going over the edge.

"Applying Thermal Compound:With the triangle mark on the substrate
pointing down and to the left," In the picture it's very little.

It also says "Once the heatsink is properly mounted, grasp the heatsink
and very gently twist it slightly clockwise and counterclockwise one
time each if possible (Just one or two degrees or so if possible). "

but an unrelate set of instructions said not to twist it. Air bubbles.
And it must be hard to twist only 1 or 2 degrees

" Please note that some heatsinks cannot be twisted once mounted"

Mine can't so I no longer have to think about this.

The break-in period for Arctic Silver 5 is 200 hours. !!!!! Can be
400+ hours, but doesn't say when.

Here are the instructions for the center dot. Like the other 3 methods,
at the start of the pdf, they list what cpus should use it, and I think
they are all old.

"the middle dot method is applicable with any thermal compound
used with the following CPU:
Pentium® 4 Or Legacy Single Core with Metal Cap
Celeron® D or Legacy Single Core with Metal Cap
Xeon® Series: Legacy Core with Metal Cap"

http://www.arcticsilver.com/pdf/appm...e_dot_v1.1.pdf

I'll have to say their method is imaginative.

You can tint a heatsink with a credit card, as the heatsink
is milled flat. No attempt is made to shape a heatsink as such.
Tinting the heatsink is when no material of any kind was
applied, and you want to make sure that any material that
goes in voids, is a ceramic particle (boron nitride etc).

The Core2 Duo uses a compressive cap which is soldered in place.
If you turn the CPU upside down, slice around the outside edge
so there is no additional material to hold it in place,
then heat the cpu until the low-temperature solder between
the lid and the silicon die releases, you would hear a
"pop" sound and the lid falls on the floor. The CPU being
upside-down and suspended in your work vice so the lid can
fall away when heated. The "pop" could imply a bit of
stress being released. The lid could have fallen away
almost silently if no stresses were being relieved.

I've never seen any information on how Intel puts the
lid on, so can't conclude much from what happens there.

Notice that the cap is not perfectly flat. The cap
is symmetric, in the sense that it's square, and
as far as I know, the pre-stress pattern is a circle
within that square. That would apply to a single die
Core2 duo at least.

A Core2 quad consisted of two Core2 duo placed inside
the same CPU package. The two chips share an FSB as
if they were on a dual socket Pentium motherboard
of long ago. They take turns being bus master or whatever.
One processor listens to what the other processor is doing,
and "cache coherency protocol" ensures both processors
know where cache lines are stored, which cache line
needs to be evicted on a particular operation and so on.
Sharing an FSB results in some performance loss (cache
coherency costs you, and when done on the FSB like that,
is not free).

Now, perhaps the pre-stress pattern is different on a
quad, but I don't own any quads to know that.

The photos in Google aren't conclusive - you can see there
is a reflection off the "edges" implying some sort of
slight curvature. The dot will still work. The dot
should spread out in some pattern when you squash it.

https://5.imimg.com/data5/VC/UA/MY-1...or-500x500.png

Heat transfer works best, if the entire heat-spreader
surface is engaged. Placing a line along the axis
of the two dies of a quad, that ensures the area above
the two dies has paste, but it does not guarantee the
spreading pattern. You still have to look at the
squash pattern and figure out for yourself, how
to compensate for the pattern you see. If any
direction does *not* have wide spreading, that
implies more material must be placed along that
axis to get it to spread.

If the surface is curved, then the areas where the
paste is thicker won't cool as well. The paste will
be thin where most of the mechanical pressure is
resting.

The purpose of paste is to exclude air voids, *not* make Oreo cookies.
For example, if you use gallium liquid metal as a TIM, the
layer can be very thin since that one involves
no suspension of solids in a carrier.

Areas where the paste is thicker, won't contribute
as much to the cooling process. They will have a
slightly higher thermal resistance. But it's better
to have that material present, than to have an air
gap. Putting a curved CPU together dry, would be
a disaster (small single-point-of-contact).

But you still want to encourage good cooling over the
entire surface.

If the lid is symmetric about the central point, then
a dot gives a good pattern with respect to that detail.

If the lid isn't symmetric about the central axis,
then use your best judgment. Doing the squash test is
meant to "inform you" about what the best pattern might be.
If you saw an ellipse, you'd say to yourself "golly,
what does this mean???".

Newer processors vary in terms of internal TIM solution.
Some processors use cookie dough inside, and the lid
on those can be dead-flat because the lid does not
need to be prestressed with cookie dough. Cookie dough
has voids, and both companies have made processors where
the material choice inside (Poppin Fresh) kinda sucks.

Intel switched away from soldering because the low temperature
solder was a "conflict mineral", perhaps some part mined
in the Congo or something. But once enthusiasts started
complaining about the usage of Poppin Fresh, they've alternated
from time to time with solder or dough. And each requires
a different solution for the lid, as perhaps prestressing
is required as some part of designing for mechanical
stresses on top of the CPU. Intel are geniuses at
engineering for stress (unlike some heatsink companies
where the arms holding the cooler, snap on them).

Haswell might have been a departure point for lid fastening,
since the FIVR power regulator was under the lid with
the CPU. After Haswell, they got rid of FIVR, so it
was a short term dalliance. Your CPU comes before Haswell,
has just silicon underneath, and is most likely to have
a soldered lid. The solder melts at a lower temperature
than tin/lead. The lid may have a curvature, and not be
as mirror-flat as a modern CPU where Poppin Fresh was
used instead of solder.

I got my practices partially by reading those docs. I
didn't ignore them. They've re-written them a couple
times.

Paul