Daily Archives: 6/11/2004

CPU Evolution…

Over on linuxdevices.com, there’s interview with Glenn Henry, founder of Centaur Technology, the bright guys behind the C3 CPUs which are marketed by VIA. I’ve got a 1ghz Nehemiah running the webserver that you’re looking at this very moment. I saw this mentioned on Slashdot and on Dan Lyke’s flutterby!, and I thought that I’d comment.

Glenn Henry had an amazingly simple idea: that he could compete in the PC chip market not by making faster chips, but by making cheaper, cooler chips. It’s a good notion. After all, my webserver ran for years on an old 133Mhz P5. Given the relatively low volume, it was more than adequate. I’m typing this message from an 800mhz Celeron, which runs Fedora Core 2 very nicely. It’s far from clear that having a 3.2ghz P4 would significantly enhance my weblog-authoring experience, or the experience of my readers.

What does enhance my weblog-authoring experience is quiet. My webserver runs with fairly aggressive power management. If people don’t access my site for some small number of minutes, disks spin down and the machine goes into an idle mode. Right now, the server is loafing along at a load of 0.02. It’s quiet and cool and energy efficient. I can leave it on all day, something I really wouldn’t want to do with todays power hungry CPUs and GPUs.

The other cool thing is that VIA is making a wide variety of small boards (Mini-ITX and soon Nano-ITX) that are full PC computers, but which run much cooler. Their 1ghz machines loaf along at under 5 watts. They can be passively cooled, which eliminates fans. That makes them desireable for home theater applications, for thin clients and just for machines that run quieter than your average PC. I’m almost more interested in this region of the CPU world than I am in additional speed enhancements (although some hardware enhancements for MPEG-4 encoding/decoding might be useful).

Centaur Technology also recognized an interesting economy of scale: that making a chip which runs 80% as fast as Intel or AMD didn’t cost 80% of what Intel and AMD spent to make their chips. The last tweaks of performance come at a great deal of design cost and die space. By accepting lower performance, they are able to design chips more quickly and with vastly smaller dies and correspondingly smaller costs. This is the kind of “work smarter, not harder” mentality that I wish were more prevalent in business.

The race doesn’t always go to the fastest. Intel and AMD can continue to try to put out muscle cars, but the market is going toward cheaper, more energy efficient technologies. VIA has placed themselves admirably to put x86 chips in all sorts of places that Intel and AMD have overlooked. I think they will do just fine.

Trilobite Fun

TrilobiteI have this old trilobite fossil that I bought from the Discovery Store. I find it kind of neat to have a specimen from 530 million years ago sitting on my desk. I made this picture by plopping the fossil onto the flatbed scanner at work and just scanning it at 300dpi. The original scan was a 2.8megabyte TIFF file, which I processed a bit with GIMP to form the image on the right. Using scanners to make pictures of shallow objects is an idea I’ve talked about before, but this was a nice application.

At home I have a Canon LIDE 20 scanner, which uses an LED light source, but it really doesn’t make very nice pictures (they get fuzzy quickly as the depth falls away from the bed of the scanner), but it was super cheap. It’s okay for doing coins, but pictures of things with deeper focus is not really satisfactory. Try scanning objects on your own scanners and see how they turn out.