Wed, 17 May 2006

AMD experiment

I thought I should mark this blog’s arrival at the ripe old age of two years by writing something. The last couple of weeks have been filled with computer hardware dramas—ranging from the ongoing flakiness of my Powerbook (now revealed as largely the result of a dead disk) to the sudden catatonia of my main workstation.

I’ve been thinking about moving to a 64-bit platform for a while, not so much because I need the performance but because it seems like a good idea to expose my own software to any possible bugs that might be revealed by the change. And since FreeBSD, my preferred operating system, now treats the AMD 64-bit platform as a tier 1 (i.e., fully supported) platform, it seems like a good chance to try out some new gear in the form of an AMD Athlon 64 X2 Dual Core 3800+ CPU.

Somewhat to my surprise, after a simple and very fast system install, I ran into trouble with ported applications—normally one of the strengths of FreeBSD. For a large number of things I wanted to install, my attempts were met by messages chiding me for trying to build something that was i386-only on an AMD platform. This list includes big things like OpenOffice (apparently because it depends on Java, which is also i386-only) and small things like the Darcs revision control system. That was all pretty irritating.

There are claims that one can run i386 binaries on the FreeBSD AMD64 platform, but my early attempts to do that with pretty trivial programs have not met with any success so far. So I don’t hold out much hope for monsters like OpenOffice.

Since the software issues looked like being slow to resolve, I thought I should at least get as much as possible of what I had already installed going. And that led to an attempt to start X11. What a mess that turned out to be.

I stupidly thought I could use the Xorg server’s -configure option to create an initial xorg.conf file. It created the file OK. But any attempt to run it with that config file resulted in a machine that was so badly locked up that it even ignored the reset button and required a power cycle before it returned to service. And this all happened so fast that it didn’t write anything in any log files, so there was no clue about a solution.

Having proved that this behaviour was completely repeatable, I then tried the old text-mode xorgconfig program. I gave the obvious answers to all the questions, created the xorg.conf file, started X and all was fine. Why this was the only way to get there is beyond me.

In the next few days, I expect to get most of the software I use all the time running on the Athlon box and to start using it for my work. I’d be surprised if that all turns out to be plain sailing, and expect to revisit this topic over the next few weeks.

Posted at: 14:07

Path: /technology   |  permalink