This post provides a photographic chronicle of the steps I took to replace the CPU fan in Sofie, my Fujitsu Lifebook s2110. I've had the computer for around 3 years, and it's never been able to maintain a reasonable operating temperature. As will become clear below, it seems the problem lies in some shoddy application of thermal grease by the manufacturer.

For this little project, I followed the excellent guide found here. For this reason, I don't provide my own detailed instructions.

The beginning, including the computer, the fan, and some tools.

First, removed all of the components on the underside of the system, including the memory, hard drive, DVD-ROM, and battery. Then removed the hinge cover and a load of screws. After the hinge cover was removed, the keyboard comes right out.

First damage: I couldn't figure out how to remove the small ribbon that connects the motherboard to the indicator panel. In the process of figuring it out, I broke the connector. The small piece of plastic lying on the motherboard is the piece I broke off. For future reference, the dark brown "tab" lifts up to release the ribbon cable.

Second, removed the plastic cover, heat shield, and wireless networking card. Be careful with the network card---the antenna does not detach and is weaved through some other components.

And what did I find? Thermal grease unevenly applied across the CPU heat sink, a piece of small thermal tape applied off-center on the Northbridge, and a very dirty fan.

A picture of the radiator through which the CPU fan (removed) is supposed to blow.

With the copper heat sinks removed, discovered that the thermal grease wasn't applied any better on the underside (note: the Southbridge and CPU have already been cleaned off in this image, but I didn't find them in any better condition than the Northbridge).

I then applied a new batch of thermal grease above (shown) and below the copper heat sinks. Everything I've ever read says you should apply a thin and even layer, being careful not to get any of the grease on any other components (the grease contains silver and is conductive).

Reassembling the machine.

Completely reassembled system, doesn't look too bad. Unfortunately, the system wouldn't start.

I decided to remove the hinge cover to verify that the panel ribbon was correctly connected. But because I had stripped a screw, preventing me from easily removing the panel, rage set in and I ripped the hinge cover off by force. Obviously, it broke. In my fit, I also knocked off two keys. Luckily, the keys were undamaged and I was able to reattach them with just a bit of delicate work (once I had relaxed).

A closer look at the damaged components.

The location of the screw that I stripped (it was accessed from the back).

As it turns out, the system's refusal to boot had nothing to do with my inability to reassemble the system correctly. The battery had died while sitting on the desk and, because I neglected to plug the system in, I didn't make this discovery until the next day. I have managed to get the system back into working order, though with the hinge cover broken and a screw missing (because I stripped it), the keyboard bounces a bit while in use. At least with the new fan installed and a new batch of thermal grease correctly applied, the system does run much cooler and quieter. That's something at least. Moral of the story: Plug a computer in before trying to start it.

Below is a plot of spam being identified daily by my crm114 filter or finding its way into my spam mailbox (this plot does not include spam being delivered to Devon or other people with valid accounts on my systems). I began tracking my spam at the end of November 2007 because I was noticing a significant upturn in spam and was looking for ways to 1) monitor the spam hitting my system and 2) measure the effect any addition anti-spam measures have on reducing spam. The vertical line plotted at early January 2008 marks the day I began using PF greylisting -- my day of (temporary?) spam liberation.

This plot is created on-demand using a simple shell and R-project script of my design. It is not run daily through a cron job because R is located on my workstation and I feel no need to install it on Crichton. The blue line shows spam coming to my system from systems other than Tim's. The red line shows spam coming from his server. The dotted lines are 5-day moving averages.

This plot will be updated as often as I feel necessary.

Significant dates (marked with vertical lines on the graph):

• 2008.01.04: Greylisting enabled on sentinelchicken.net mail server.
• 2008.01.22: sentinelchicken.org is shutdown as Tim moves.
• 2008.02.02: sentinelchicken.org is back as Tim settles in to his new place.
• 2008.02.15: Comcast thinks it's a good idea to begin blocking traffic to Tim's port 25. He is forced to use an alternative port on my system to relay his mail. Tim begins to look for a different ISP.
• 2008.05.14: Changed spamd settings to -G 15:4:864 from 25:4:864 to see if that makes any difference to spam rates. The goal is to speed up initial delivery times, while not increasing spam acceptance.
• 2008.08: My new server reinstall has made these statistics incomparable to previous periods. I have discontinued producing these charts. Hopefully, something else will take their place soon.