lapped CPU for better cooling?

I saw a Core 2 Quad for sale online, with the heatspread
lapped for better cooling. It has been machined right down, so
you can't read any numbers.
It is a Q6600, but the seller claims it overclocks to 3830 MHz
with this butchering. That would be 11.5 x 333 MHz ... hmmm
do you believe it?
How would this be so much better than a thin layer of silver
heatsink paste?

"Orson Cart" ex-privat parts.org wrote:

I saw a Core 2 Quad for sale online, with the heatspread lapped
for better cooling. It has been machined right down, so you
can't read any numbers. It is a Q6600, but the seller claims it
overclocks to 3830 MHz with this butchering. That would be 11.5
x 333 MHz ... hmmm do you believe it? How would this be so much
better than a thin layer of silver heatsink paste?

If it were something else.

On 1 Jan 2012 03:55:39 +0100, "Orson Cart"
wrote:


I saw a Core 2 Quad for sale online, with the heatspread
lapped for better cooling. It has been machined right down, so
you can't read any numbers.
It is a Q6600, but the seller claims it overclocks to 3830 MHz
with this butchering. That would be 11.5 x 333 MHz ... hmmm
do you believe it?
How would this be so much better than a thin layer of silver
heatsink paste?

Lapping would be done to restore surface fatness which might have been
distorted from over heating. Then polishing would be the next process
to increase conductivity. I would stay away from something like that.
The process of polishing would be enough to remove the engraving and
to admit the device was lapped seems to be telling.

Allen Drake wrote:
On 1 Jan 2012 03:55:39 +0100, "Orson Cart"
wrote:

I saw a Core 2 Quad for sale online, with the heatspread
lapped for better cooling. It has been machined right down, so
you can't read any numbers.
It is a Q6600, but the seller claims it overclocks to 3830 MHz
with this butchering. That would be 11.5 x 333 MHz ... hmmm
do you believe it?
How would this be so much better than a thin layer of silver
heatsink paste?

Lapping would be done to restore surface fatness which might have been
distorted from over heating. Then polishing would be the next process
to increase conductivity. I would stay away from something like that.
The process of polishing would be enough to remove the engraving and
to admit the device was lapped seems to be telling.

Here's an article from that time period.

The lid on the Core2 was known to not be flat.

http://www.legitreviews.com/article/402/1/

Some decided to remove the lid entirely, for fun. And
there were a few who succeeded in having a perfectly
good working processor afterwards. The lid is soldered
to the surface of the die with a low temp solder. In that
legitreview article, they used a blow torch, and perhaps
that's just too strong a heat source for the job. A hot
air gun might be enough.

Paul

Orson Cart wrote:

I saw a Core 2 Quad for sale online, with the heatspread
lapped for better cooling. It has been machined right down, so
you can't read any numbers.
It is a Q6600, but the seller claims it overclocks to 3830 MHz
with this butchering. That would be 11.5 x 333 MHz ... hmmm
do you believe it?
How would this be so much better than a thin layer of silver
heatsink paste?

I don't know what you mean by "machined" or "butchered". Lapping is a
polishing procedure to remove cupping or irregularities in the surface.
You are using fine grained sandpaper to slowly grind away any high spots
and typically using plate glass (thick so it doesn't flex) as the
underlying flat surface atop which the sandpaper rests (grit side up).
You start with 400 grit sandpaper and the wetted with water and move
progressively up in fine grit to 600 and up to 1500 (for a really shiny
surface). You want to use a closed-coat sandpaper for polishing metal
and the type that can be wetted (to carry away the dust so the sandpaper
doesn't clog). Open-coat sandpaper is rough so the sawdust from wood
sanding doesn't clog the grit. Just look for 400, 600, 800 (and 1500 if
you're meticulous) wet sandpaper, like what you find for automotive
repair. Don't use your desk or kitchen table for the flat surface on
which you put the sandpaper. Those aren't nearly flat enough. Get some
plate glass from the hardware store. You'll also need lots of patience
as the lapping is very slow grinding. You're polishing the surfaces,
not eroding them away. When lapping, you can see what parts of the
metal plate are getting polished down. When that pattern becomes even
all the way across, you've removed all the high points. Lapping also
smooths the surface of the metal plate so the pores are smaller and
there are no scratches so you have a better mating surface. If the
surface is severely cupped, you'll have to grind away so much of the
material that, for the CPU plate, you make the material too weak (it's
too thin). After lapping for awhile, you can tell how bad is a surface
to determine if you can lap it completely flat and smooth without
weakening it; however, if you're in that situation, you wouldn't want to
use that CPU since it will require a ridiculous amount of thermal paste
to fill in the oversized gaps.

Of course, lapping just one surface is of dubious value since one great
mating surface pushed against another rough surface isn't going to
benefit you much. You want BOTH surfaces to be lapped to polish and
flatten them. Metal to metal contact is far superior for heat
conduction than metal to air to metal. If you're going to lap the
heatsink then you should also lap the CPU's thermal plate. The heatsink
is think so there is no concern about grinding away too much material
even for the vapor tube type heatsinks. The thermal plate on the CPU is
much thinner but still more than thick enough to lap it smooth and flat
without worrying about losing structural integrity for the pressure
applied against it by the heatsink.

You didn't provide any pictures or a URL link to the lapped CPU for
anyone else to see what what this butchering you claimed.

Air is the worse thermal conductor. Remember that it is used in the
insulation in your walls and attic to keep *in* the heat. The
microscopic pores in the mating metal surfaces of the heatsink and CPU
plate have air in them hence the need for thermal paste. But do you
think thermal paste is as good a conductor as metal? I don't care if
it's silver, gold, or magic chalk mixed in with the grease. It's still
not the same as metal to metal contact between the heatsink and CPU
plate. You don't use paste to be the thermal conductor. You use it to
replace the *air* that is a worse conductor than paste. Any paste, even
toothpaste, is a better thermal conductor than air. But you want to use
as little thermal paste as possible. Remember that you're using it to
fill in the microscopic pores in the mating metallic surfaces, NOT as
the conductor between those mating metal surfaces. You do NOT want the
paste between what would've been metal to metal contact. You do NOT
want to goop on the thermal paste like you apply toothpaste to your
toothbrush. You dab a tiny amount and then spread with some mylar or
credit card so you end up with a *translucent* layer of thermal paste.
Even if its white thermal paste, you should be able to see through it.

Lapping the mating metal surfaces means there is more metal to metal
contact which is the best conduction medium. It means less of the
less-conductive thermal paste is filling up the air pores. Heat kills
so you do NOT want anymore than necessary a layer of less-conductive
thermal paste between the mating metal surfaces. While I lap my
heatsink and CPU plate to reduce temperature because the heat is more
easily carried away with a more thermally conductive mating, some folks
use that faster thermal conduction to overclock their CPUs but then, of
course, they're undoing the lapping by upping heat production so they
need a better heatsink. Lapping alone without overclocking will
probably lower the CPU temperature by 5-10 degrees which improves the
MTBF. Overclocking will raise the temperature but the lapping helps
transfer the heat faster.

You want metal-to-metal contact as much as possible for the best thermal
conduction. You don't want to use thermal paste except to fill in the
microscopic pores and any cupping or mismatch in the mating of the metal
surfaces. Lapping reduces the roughness, cupping, or other mismatches
(as long as both surfaces are lapped similarly so they match up better)
so you use LESS thermal paste. The best case is you don't need any
thermal paste and the metal surfaces have perfect contact. Alas, there
hasn't been much demand for heatsinks that are soldered onto CPUs or
where the CPU's plate is actually part of the heatsink.

You want to use as LITTLE thermal paste as possible. Lapping polishes
the surfaces and gets them to mate better so you can use LESS thermal
paste. Lapping gives you more metal-to-metal contact between the mating
surfaces for better heat transfer.

http://www.youtube.com/results?searc...g+cpu+heatsink

Lapping just of the heatsink has little value if you don't also lap the
CPU plate. You want BOTH surfaces to mate together as much as possible.
Shiny against rough is only marginally better than rough against rough
and in both cases you end up using more thermal paste. You want to use
LESS thermal paste! Heatsinks are notorious for being rough and cupped
(not flat) so you get the most benefit from lapping heatsinks will
little worry about damaging it. CPU thermal plates tend to be flatter
and more polished so you could just use it as-is. If you're meticulous
and trying for that extra 1-2 degree temperature drop then lap the CPU
plate, too; however, be careful not to grind too much as the CPU plate
isn't that thick. Since the CPU plate is usually much better than the
heatsink regarding polish and flatness, it shouldn't take much lapping
of the CPU plate to get a flat mirror finish.

And, yes, now that you know what is lapping, it will remove the ink
markings atop the CPU plate because, after all, you are polishing that
metal and the ink is sticking up on the surface of that metal.

On 1/01/2012 1:55 PM, Orson Cart wrote:
I saw a Core 2 Quad for sale online, with the heatspread
lapped for better cooling. It has been machined right down, so
you can't read any numbers.
It is a Q6600, but the seller claims it overclocks to 3830 MHz
with this butchering. That would be 11.5 x 333 MHz ... hmmm
do you believe it?

You would need some heavy cooling but it can be achieved. The Q6600 was
an early CPU in the range so had a lot more headroom for overclocking
than the later higher clocked CPUs. The Q6600s were a popular Quadcore.
I even have one here on my main workhorse. I do not overclock it as it
is fine for my needs at stock clocks. Some of my friends pump them up to
speeds around 3 GHz using stock cooling.

How would this be so much better than a thin layer of silver
heatsink paste?

The whole idea of a heatsink is to transfer the heat to another medium
as quickly as possible, then dissipate it. Metal to metal contact will
be much more efficient than any form of heatsink paste sandwich.
Technically, if your lapping is perfect, you don't need the heatsink
paste at all.
Note that lapping takes place when the CPU is cold therefore any
distortion of heatsink or plate as the CPU warms up will render all bets
off. The components should be thermally stable at normal and slightly
overheated temperatures but there are no guarantees.

--

Krypsis

"Krypsis" wrote in message ...
On 1/01/2012 1:55 PM, Orson Cart wrote:
I saw a Core 2 Quad for sale online, with the heatspread
lapped for better cooling. It has been machined right down, so
you can't read any numbers.
It is a Q6600, but the seller claims it overclocks to 3830 MHz
with this butchering. That would be 11.5 x 333 MHz ... hmmm
do you believe it?

You would need some heavy cooling but it can be achieved. The Q6600 was an early CPU in the range so had a lot more
headroom for overclocking than the later higher clocked CPUs. The Q6600s were a popular Quadcore. I even have one here
on my main workhorse. I do not overclock it as it is fine for my needs at stock clocks. Some of my friends pump them
up to speeds around 3 GHz using stock cooling.

How would this be so much better than a thin layer of silver
heatsink paste?

The whole idea of a heatsink is to transfer the heat to another medium as quickly as possible, then dissipate it.
Metal to metal contact will be much more efficient than any form of heatsink paste sandwich. Technically, if your
lapping is perfect, you don't need the heatsink paste at all.
Note that lapping takes place when the CPU is cold therefore any distortion of heatsink or plate as the CPU warms up
will render all bets off. The components should be thermally stable at normal and slightly overheated temperatures but
there are no guarantees.


If the surfaces are smooth and flat enough to have a friction fit, then with heatsink clamped down on it, there really
shouldn't be any warpage of the CPU case since the heat would be transferred (almost) immediately to the base of the
heatsink (assuming the transfer coefficient of the heatsink base is equal to or better than the CPU case material). Of
course, in the real world, hardly any of those conditions will exist, most likely :-)
--
SC Tom