Some final words on the Ultimate Linux Boxes, including a contender for the Ultimate Notebook.
It's too bad we didn't get enough notebook computers to choose an Ultimate Notebook this year, particularly because so many people have problems installing and using Linux on many notebook computers. If it is any consolation to you, I've had great success with my ABS Mayhem G4 A78 notebook (www.abs.com). It has been replaced by the G4 Revolution since I bought it (and the price has dropped significantly, naturally). The key to the G4, or any other notebook you consider, is to pick one with Linux-supported hardware. Most notebooks come with either an NVIDIA display or ATI. The G4 A78 has an NVIDIA display driver, and I've always had good luck with NVIDIA, so that was a driving factor in choosing this notebook. It also has an Intel M-PCI PRO Wireless Chip, which works out of the box even with distributions like Kubuntu. The only manual work I had to do to make it work was add a line to my configuration file to specify my WEP security code. All in all, I installed and configured Kubuntu on this notebook faster than I was able to get the pre-installed Windows XP to work properly. The Windows wireless driver couldn't connect to my network unless I advertised the network name (SSID). It didn't matter that I typed in the SSID manually. Linux had no such trouble.
The bottom line is that the Mayhem G4 A78 is a great performer and very Linux-friendly. The problem is that the latest Mayhem notebooks have slightly different hardware. ABS is going with the Intel WM3B2915ABGNAX Mini PCI Wireless Adapter now (can they add any more letters to that item number?). I'm not at all confident that most Linux distributions will work with this chip out of the box, but then I haven't tried it.
Here are some postscripts about the Do-It-Yourself Ultimate Linux Box. We offered some advice on how you can get a little wiggle room on price, but you might benefit from some common-sense tips on what not to do. One of the biggest mistakes you can make is to opt to go with a dual-card NVIDIA SLI configuration without getting any real benefit for the price. For example, don't compromise big on the price of your processor in order to invest in a dual-card NVIDIA SLI configuration. If you don't have enough CPU power, you won't get what you want from the display cards. Likewise, there's no point in doing SLI if you're going to connect it to a monitor that can do only a 1024x768 resolution. With a monitor like that, you probably won't see any improvement in performance or quality compared to a single card.
Oh, and here's something you need to know. You can't use dual monitors and SLI mode at the same time. You can switch between dual monitors and SLI mode without making any hardware configuration changes, but you can't have your SLI cake and eat your dual monitors too (whatever that means).
Here's a do-as-we-say-not-as-we-did tip. If you are a do-it-yourself type, you probably have several computers and play the hand-me-down game. When you upgrade your own video card, you hand down your existing card to the computer your kids use (or vice versa, if they get the premium stuff, first).
If you are going to hand down your CPU, be very careful when removing the CPU from one motherboard to transfer it to another. We transferred a CPU several times in order to test the various motherboards, and there was one time when it stuck to and popped out while removing the heat sink.
Here is where I'll personally take the blame and switch from the editorial we to I. I bent a few pins while prying the CPU loose from the heat sink. Try as I may, all my efforts to get the pins back into position just made things worse. I was never able to re-insert that CPU into a socket. Dual-core AMD64 CPUs don't come cheap, so that was a painful lesson to learn, more painful than leaving a piece of my thumb in the CPU fan on the Aberdeen server.
I guess the lesson here is that if you are as clumsy as I am, get someone else to do the tricky work for you.
Finally, we made mention of a number of benchmarks that weren't fit for publishing because they didn't provide useful information. In one case, we're referring to a benchmark called 3DMark 06. The numbers didn't add anything to what we'd already found, and we couldn't publish the results anyway, because our copy of the benchmark is not for commercial use. But if you dual-boot Windows, I recommend you have a look at this benchmark for yourself. You can download a copy from www.guru3d.com. If your hardware is good enough to render the tests well, you'll be blown away by the test scenes.