No offense people, but I'm getting a little tired of everyone FREAKING OUT installing these heatsinks on Socket A processors. I'm also getting tired of people trying to return broken CPUs to my work place with stories like "it got so hot that it shattered" and "it was like that when I got it". It's not brain surgery installing a socket A heatsink. It merely takes CARE and CAUTIOUSNESS and a little bit of common sense.

The following series of photos, I've called "This Can't Be Right". The reason why I named the series this is because I received the board back from the customer with the fan mounted, and right off the bat, without even removing the heatsink from the socket, I knew the CPU would be damaged. I immediately said, "You mean to tell me that they saw how they had this mounted and couldn't come to the conclusion that this can't be right!"

First, I could see quite a bit of ceramic along one edge of the CPU and the heatsink was obviously hiked up in the rear. This was because the fan was put on 180 degrees backwards.

On the other side of the socket, you can see where the heatsink hangs over the CPU. Again, how can you look at this and not say "this can't be right"!

Taking the heatsink off, one can see how the CPU die completely missed the thermal pad. Note how the aluminum is all scratched up from the two surfaces, the die and the heatsink, have been forced together.

This is how the die looked close up. Needless to say, the CPU does not work.

I want to say that I have installed HUNDREDS of socket A heatsinks INCLUDING the Golden Orb and it's cousin the Chrome Orb and have NEVER chipped a die, cracked ceramic or caused any other damage of the sort. We actually TRIED to break a CPU ON PURPOSE by mounting the heatsink wrong and it took one of us putting ALL OF OUR WEIGHT on the CPU to finally crack the ceramic.

Unfortunately, people are cracking, chipping and breaking these things and are not even realizing that they are doing this damage out of pure CARELESSNESS.

Below is a photo submitted by an individual in a forum. I, for one, would think that if I'm having to bend capacitors out of the way of the heatsink that maybe something is wrong, but not our friend. He actually informed us all that this photo was taken AFTER bending the caps back into their original position as possible. You would think that common sense would dictate that if there was interference in the first place, never mind "touching" the capacitors, why even try?

Now, the above "need" to bend caps out of the way has prompted people to install the Orbs 180 degrees backwards, much in the way the fan was installed in the "can't be right" series, in order to accommodate boards not appropriate for the Orb. Unfortunately, doing this places the center of the Orb off the center of the CPU. You can see in the below photo, the mark left on the GOrb created by the edge of the die.

Note how this is similar to the mark left on the "can't be right" heatsink? Guess what the CPU die looked like.

The pink pad is where the die should be. To the trained eye (yeah, right) something was definitely wrong. This CPU, where it was crushed by the heatsink, sort of looks like it has glitter on the side of the die where it had been scraped up by the GOrb. See the below photo.

The above photo (and many more to follow) represents a CPUs that was once functional, but is now completely DESTROYED, not because of faulty engineering, but because of gross user error. This has prompted me to bring to light the rights and wrongs of HSF installation on a Socket A CPU.

First I need to communicate to you that if you install the CPU and heatsink OUTSIDE OF THE CASE things will progress MUCH easier because you can get a good eyeball of whether or not that puppy is really on the die or not. I've actually made it a habit to install my CPU and RAM and do bench tests of the core components BEFORE installing everything in the case long before Socket A's were a twinkle in AMD's eye.

One good reason to install outside of the case was actually submitted by the man who brought us Arctic Silver. Nevin stated that if the CPU is not all of the way in the socket (as shown below. No it's not Socket A, but it was handy) then the heatsink may "seem" to be installed correctly, but the PC may lock up, heat up, etc. for more reasons than I care to list! It's a rare mistake, but it DOES HAPPEN! So make sure that puppy is FLUSH in that socket!!!

In this first photo below, I'm getting the first clip on the socket. Note the clip in between the row of 5 capacitors. It is important that this clip is as much on the socket as possible for doing ANYTHING ELSE.

I then put the CPU down, flush on the die. I DO NOT attempt to clip the other clip on in the same motion as making the heatsink flush with the same hand.

You can see in the below picture that I am holding the heatsink level with my middle finger while applying the clip with my thumb and fore finger. In most cases, I would use both hands; One two hold the heatsink flush to the CPU and one to apply the clip. At the time, no one was available to hold the camera. Another thing to make sure of is that the sink is only touching the die and not any of the resistors on the CPU.

Below is a design I really like to offer the end user. The clip has a ratchet. If you feel that the tension is insanely tight, you can ratchet the clip up to the next setting. Yes ladies, it's true. That's a wedding band. I'm a married man.

I love this HSF. It is an EXACT RIP OFF of the Foxconn heatsink, but with a ratcheting spring clip and a larger ball bearing fan. If you elect me president there will be one of these on every CPU and a turkey in every oven...

How good of a rip off is it? To the left is the Dynatron HSF unit with the ratchet clip. To the right is the Foxconn that was supplied to me by AMD. SCARY isn't it?

For the most part, removal IS the reverse of installation. Be sure to only flex the clip and NOT apply pressure to the heatsink pushing the sink down into the CPU. This is unnecessary. Simply push the clip towards the board, and then flex it AWAY from the socket. Below, I'm pushing the hinged clip of my Foxconn-copy down and then hinging it outward away from the socket's tab.

In the photo below, I'm using a newer Dynatron HSF who's clip does not feature a hinge. Fortunately, it offers a thumb tab large enough to easily squeeze inward to clear the tab on the socket.

On some heatsinks, like the Volcano II shown below, the tab is NOT large enough to allow easy leverage and the clip is NOT hinged. These clips typically have a means of grabbing onto the clip with a small pair of pliers or a tiny screwdriver. You can see on this Volcano that this means is in the form of a tiny hook that points up away from the clip.

Don't be ashamed if you have to bust out the pliers. The well seasoned tech knows when to stop and try another approach. The guys with tenacity tend to be the ones that cause the most damage.

Alphas are funny little heatsinks. It looks like a heatsink that is easy to install. Well, it is easy to install, if you install it Stan's way.

My coworker Stan informed me that the Alpha is easy as pie to install if you put the hinged clip on the socket FIRST as shown below.

After you do this, you steady the heatsink with one hand, and push down on the BACK SIDE CLIP next! He's right! It's a cake walk!

One heatsink that a co-worker of mine has turned me on to, as a solution for those boneheads that muscle their units a bit too much and end up breaking the mounting tabs off of their socket, is the Swiftech. It does not mount on the tabs of the socket, but rather the four holes that most Socket A boards come with.

The Orb has gotten a lot of flack. In my opinion, the flack is coming from people that don't have the patience required to NOT break a CPU in half. What a shame. The first thing to do is to flex the hell out of the clip.

No need to buy fancy hand exercise devices. Just put a orb heatsink in each hand and repeatedly flex the clips to build up your hand muscles. You never know when you may need strong hand muscles.

The ability to READ is a must when assembling your PC

Just like with the other heatsinks, you have to make sure that the 1st clip you put on is absolutely flush with the socket. The Orb is unique in that it has three holes that mate with the three tabs on the 462 socket.

Actually, the worst part of the Orb is the diagram on the side of the box. The image seems to imply that you simply ROCK the heatsink into position. Wonder how you took that chunk out of the corner of the die? I don't.

Regardless of the drawing on the box, make sure the heatsink is flush before even THINKING of trying the other side of the clip. Don't even THINK about it. Think about something else. An Audi TT, Comdex babes in bikinis, a fishing weekend in Alaska... whatever! Just make sure the heatsink is FLUSH!

Fortunately, the NEWER Thermaltake Orb (the silver one often called the Chrome Orb, or ChOde for short, even though it is not chrome and has been in seen in BOTH gold and chrome colors) has a couple slight improvement. Slight. For one, it's silver because AMD recommends a non colored heatsink surface. Secondly, the newer orb has little plastic spacers on either side of the die contact to prevent the heatsink from "rocking" on the CPU.

Apparently, Confucius say that round eye American devil break many CPU with superior cooling solution, thus AMD pulled the Thermaltake Orb from it's list. This prompted Thermaltake to make these improvements PER AMD. But apparently, these improvements simply weren't enough. For a brief time, Thermaltake actually had the above HSF on their website AND this below one:

This one (called the DUO462) has a SQUARE BASE instead of a round one so now the rubber pads on the top of the CPU come in contact with something more than air. Also, they took the two little chrome tabs off of either side of the clip (seen best in the next two photos) so this version of the Orb can be used on the Biostar motherboards. That's good. I never knew what those tabs were there for in the first place and often broke them off.

After you have the heatsink flush AND ARE HOLDING IT FLUSH you can work on the other side of the clip by pushing it down onto the socket. DO NOT MOVE THE HEATSINK.

Also, keep in mind the SURFACE MOUNTED CAPACITORS AND RESISTORS ON THE MOTHERBOARD. The clips of the heatsink can easily SCRAPE these off rendering the board USELESS and possibly frying the CPU too!

Below is another angle shot of C481 and C480 and the LACK of capacitors on those points. On the other side of the socket there are similarly mounted resistors.

Some fried CPUs have been blamed on, not overheating but, the heatsink touching the resistors on the top of the CPU. Even some websites have "reported" this being a potential problem.

I'm sorry. If you have that heatsink on so crooked that it "might" touch a resistor, then it's not touching the CPU enough to offer adequate cooling in the first place and it's going to fry anyway.

Below you an almost make out where the resistors don't even come close to touching the heat sink. Don't even think about using this lame excuse when requesting an RMA number. Not that it CAN'T happen, but because to do so, you would REALLY have to be a screw up.

For those of you who have to have one of these Socket A CPUs right now, but feel that they may not be ready for the challenges of installing a heatsink you may be tempted to use a shim like the one shown below:

Unfortunately, using a shim made of a thermally conductive material, like copper, pulls heat back down from the heatsink and into the ceramic part of the CPU. This isn't a big deal with a non-overclocked CPU since the CPU's temperature shouldn't be near it's threshhold to start with, but when overclocking, the added temperature poses a problem. My take on it? If you're not competent enough to install a heatsink without screwing something up, you're likely not competent enough to overclock. Take offense to that statement? Good.

ONE MORE THING I'd like to run by you before I sign off for the night... I went to an AMD "get together" where they reviewed future product and told us something that FLOORED me. They stated that AMD absolutely DOES NOT recommend thermal grease and would rather one uses phase change thermal compound. Apparently the grease doesn't hold up under the heat and heatsink tension. The expansion and contraction of the two surfaces during on/off cycles causes a "pumping action" that pumps the thermal compound out from on top of the die). SUCK! So much for telling people that they can just run down to Radio Shack to get a tube of thermal grease! I've been telling people they can do this FOR YEARS!

Chromerics (shown below) is on the top of the AMD list of thermal interface materials.

I have to admit that Arctic Silver does have a much higher temperature tolerance than most greases I have used and personally I have never seen AS "break down" under prolonged use on most applications, unlike Radio Shack grease. Despite this AMD still hasn't approved it for use and after I had been using it for several months, I can see why. I had removed a heatsink that had been running with Arctic Silver for several months and never had the heatsink off and saw what could be described as a "blast pattern". All of the Artic Silver had "pumped out" from the where the die was, leaving only an outline of the die and a dry section in the middle. I mentioned this to an AMD tech and he told me that the viscosity of the compound is one reason why it is not recommended by AMD.

The "pump out effect" or "blast pattern".

Sorry it's a crappy picture, but to get the shot I needed I had to use only natural lighting.

What's the other reason? Electrical conductivitiy. Arctic Silver is barely electrilcally conductive at all, but the fact remains that it does have some electrical conductivity. Mind you, of course, that if you only apply a small amount of compound to the die of the CPU, then electrical conductivity should not be an issue because the compound should not come in contact with the resistors on top of the CPU. For lab use (because thermal pads are too expnsive to replace over and over again), AMD recommends one particular brand of paste and I'm a dumb ass and didn't write it down. But the paste they recommend has NO electrical conductivity, but still has the viscosity issues that currently all greases have.

But alas! There is hope! Arctic Silver II is out and it has a HIGHER VISCOSITY than regular Arctic Silver! And so far, this higher viscosity seems to be enough to prevent the pumping phenomenom! Certainly AS raised the viscosity of their product for a reason. I had bought a tube not four months ago and have yet to experience an increase of temperature! I removed the heatsink and found no pumping out.

Explain That

In closing, I would like to voice my opinion that I think the Socket A AMD is a HORRIBLE DESIGN. Word is, at AMD, theat they are working on installing a heat spreader. I can understand why they couldn't make the CPU with a heat spreader in the past, because the Thunderbird cores just got way too hot too fast. But the Palominos run so much cooler, AMD can spare the added thermal resistance of a heat plate. To give you an idea, I ran a 1.2 GHz Palomino next to a 1.2 GHz T-Bird next to each other on the same type of motherboard and the heatsink on the T-Bird was warm to the touch throughout. The Palomino's heatsink, even after an hour of Final Reality in software rendering mode, was not even warm at the base! Kick ass!

Re-Printed From SLCentral