Sciency stuff

21Sep/10179

P-III autopsy

0000047

For teaching a course I needed to take a closer look at a CPU. I asked around and got my hands on an old P-III Coppermine that was about to get thrown out. I’ll start with a disclaimer: I know virtually nothing about CPUs, so if I claim something to be true, it probably isn’t.

The first challenge is to get the actual silicon processor chip off of the plastic bonding board. In the picture below, the blue thing you see is  the back side of the processor chip. When the processor is finished, it is turned upside down and bonded to the green circuit board. This allows the metal pads on the silicon chip and the pads on the circuit board to join, creating a connection (this is one of those claims...). I believe that the CPU at that stage is heated up in order to melt the joints and thereby solder them together.

Click images for full-size. Especially the scanning electron microscope images below could be interesting to view in all its splendor.

The blue part in the middle is the actual Si chip. I needed to remove it in order to further inspect the CPU.

Back side of the circuit board containing the CPU. Each pin you see (give or take) should be connected to a pad on the silicon ship.

I figured I should be able to remove the chip by heating it. I first tried using a heat gun, but that just made some bad smelling fumes. I instead turned to brute force and used a power-saw to cut out the part containing the actual chip. Using pliers I managed to get a few pieces off of the board (cracking the chip in the process, but I was planning to do that anyway) and got the rest off by using a scalpel.

A saw comes in handy sometimes...

Below you can see the result. On the bottom side of the piece that came off you can see all the connector pads that were previously connected to the pins on the backside of the circuit board.

A piece of the processor chip came off.

This is the piece that came off. It's been flipped so that the side you see was originally facing towards the circuit board. Each little dot contains a metal pad that connects the interior of the chip to the leads on the board.

Now the interesting part begins. I looked at the piece above in an optical microscope. The picture below shows an enhanced version of the little dots in the above picture.

This is a piece of the processor. The side you're seeing was once facing towards the plastic circuit board. Each little hole is a metal pad connecting the interior of the chip to a macroscopic lead on the circuit board.

Looking closer, we start seeing some structure inside the holes.

A processor contains many layers of metal leads in order to connect the transistors at the surface of the silicon chip into useful units. The metal layers are clearly visible through the small holes in the chip.

Furthermore, by changing the focus of the microscope, we can see multiple layers within each hole

Focus is on upper layer.

Focus is on middle layer.

Focus is on bottom layer.

In a CPU there are multiple layers of sandwiched metal leads going down to the transistors at the bottom (at the surface of the silicon wafer). I believe what we're seeing is simply those different layers.

Since optical microscopy doesn't show very much detail, I decided to load the chip into a scanning electron microscope (SEM).

What I did was to cleave the chip into smaller pieces. This way I can peek from the side of the chip and get some cross-sectional images. Below is a series of images that shows a zoom-in on the surface of the Si chip. For some reason I had a lot of trouble getting a good focus. I'm not sure what kind of Si is used for these CPUs, but if it's a non-conducting silicon, the electron beam in the SEM can charge the material making it difficult to focus. Another possibility is that all the plastic encapsulation material caused charging effects.

You're looking at the processor chip from the side. In the top part of the image you see the metal pads that were once connected to leads on the circuit board when the chip was bonded face down.

We still don't see a lot. The light stuff between the metal pads is probably some kind of polymer used to fill up the space.

We start to get more detail at the surface of the silicon chip (bottom part of the image. The Si starts somewhere at the bottom of the large metal blob). The texture in the polymer filler (next to the metal blob) could be due to something being mixed into the polymer to increase its thermal conductivity.

We start seeing something below the metal blob. The vertical lines that are barely visible would be the multiple layers of metal leads. To the right you see the same structure but from another angle since the cut changes direction.

At this point they are clearly visible. I count about six layers of metal leads visible in the image.

The feature size of the lowest metal layer is around 200-250nm. Since the P-III started out at 250nm process but developed into 180nm (according to wikipedia), the transistor layer must be fairly close to the lowest visible metal layer in the image.

Just a nice overview.

This is not a cross-sectional image but is taken from the top (along the sample normal for those of you who speak science). I accidentally chipped the processor and this is how it looks. Several of the metal layers are visible and as we go down in the image, we go down through the metal layers. I would guess that the bright spots you see are vias connecting leads lying in different layers.

I believe that in order to see the actual transistors at the bottom (at the surface of the silicon wafer) I will need to remove all the layers. I'll let you know if I figure out a good way of doing this.