Better Than Thermal Grease?

As chips get hotter and hotter, there will be more difficulty on the part of
the manufacturers to keep the chips cool. What that means is the speed
rating will occur at the highest speed that is stable at the temperatures
that they can achieve with the cheapest air-cooling solution they can come
up with.

It’s all $$$, and they’re not going to throw $50 copper/aluminum 3
fan heatsinks with cool-air ducting to an intake on the case any time soon.

So, for the overclocker wanting to get the most from the chips, a good
water-cooled system can make such a difference that heat is ALMOST removed
from the equation, and the other limiting factors of the chip come into
play. I know, to really get there, you need sub-zero temperatures, but just
room temperature water cooling will be vastly superior to the mass-produced
solutions from the manufacturers.

That said, I think as the dies shrink we’ll see greater and greater return
from our water-cooled systems.

On a side note, I’ve found that sometimes the difference between success and
failure comes down to the connection between the heat transfer plate and the
CPU. Something that I’ve started doing with my kids computers is using JB
Weld to attach the CPU to very large heatsinks.

This is PERMANENT, NEVER GO BACK type of thing, but by doing the following I’ll tell you what results
I’ve had.

First, I lap the CPU and heatsink until they’re mirror smooth and
a perfect fit. If I can smear a tiny bit of water on the CPU, press it to
the heatsink and then lift straight up and pick up the heatsink then it’s
done correctly.

Next, I use a hair-dryer to heat up the CPU, heatsink, and
a 35 pound steel weight.

I mix up some JB Weld, and put a big dollop on the
center of the CPU. The heated CPU (about 160-170f) heats up the JB Weld as
I’m putting the hot heatsink on top of it, followed by the 35 pound weight.

The JB Weld will act as a lubricant between the CPU and heatsink and the
heatsink will start gliding off to one side or the other, so you have to
maintain the positioning until the heatsink metal touches the CPU and the
weight keeps it from sliding around.

The heat causes the JB Weld to become very, very thin. The majority of it is pushed out of the joint and there
is close to a metal-metal contact between the CPU and the heatsink. Any
small gaps are taken up by a TINY film of JB Weld (good heat conductivity
anyway) and the 35 pound weight will keep forcing JB Weld out until whatever
layer left is MUCH MUCH thinner than any film of heatsink grease would ever
be.

Now let this cure for 18 hours or so, and then take the combination of
heatsink and CPU and find a way to tape around the perimeter of the CPU to
seal between the CPU and the heatsink and then fill that up with JB Weld and
let it cure. That way, the slug itself is attached to the heatsink AND the
surrounding chip surface is connected through the JB Weld, so you get as
much heat transfer as possible.

Doing this, I’ve successfully gotten a 300 celeron that wouldn’t do 450
air-cooled (I even tried C-clamps to get the best
contact, but no go) to 464 with the exact same heatsink, but attached with JB Weld
instead of clips and heatsink grease.

I’ve also used this on a 366 that wouldn’t do 550
aircooled. I used a large pIII heatsink with the 366 celeron in a slocket,
and no go no matter what I tried. When I attached that same heatsink with
JB Weld and no other changes, 567 no problem.

I now use this routinely for video card heatsinks and m/b chipset
heatsinks. I offer the following cautions: