What's more important, CPU speed or memory?

That's the question I'm asking myself at this very moment. Running Puppy Linux 2.14 on the Thin Puppy -- the repurposed Maxspeed MaxTerm thin client -- is a mixed bag at this point in the testing process.

The thin client -- with Puppy on a Compact Flash card -- is built around a Via motherboard with a 1 GHz Via Samuel processor, with a slowish 133 MHz front side bus and equally slowish SDRAM memory. It's spec'd for PC133 memory, but I've got a 128 MB stick of PC100 in there. Previously I've said that Puppy runs OK with 128 MB, but I have rethought that position, and I'm saying now that you need at least 256 MB to be comfortable with Puppy.

One of the features of Puppy is that it's designed to be run from CD-ROM but can be installed to a hard drive, bootable USB flash drive or, in the case of the Thin Puppy, a bootable CF chip plugged into a CF-to-IDE adapter so the motherboard sees it as a standard IDE hard drive. And Puppy is also designed to save to disk only once per computing session -- when you shut down. But with 128 MB, when the processes involved in displaying a complex Web page or streaming video are brought to bear, the CPU starts to strain, the amount of free memory can drop, and Puppy Linux starts accessing the hard drive -- in this case the CF chip -- using a swap file, I presume. That can't be good for the life of the CF chip, nor the computing experience as a whole, since swapping to flash is much, much slower than doing the same operation with a standard hard disk.

And that 133 MHz FSB can't be helping either. While a 1 GHz processor is plenty fast for most pedestrian uses, a quicker FSB with faster memory -- such as today's DDR and DDR2 -- can really make a system work faster.

The dilemma. Puppy is supposedly designed to run well on older systems with less memory. But at 128 MB, you can't even download and install any "major" add-on packages, such as Open Office or Wine -- there just isn't enough memory to do it. And once you get to 512 MB of system memory, Puppy will obey, but you could probably run ANY distribution of Linux out there, like Ubuntu, Mepis, Fedora, OpernSUSE or what have you. If you have the memory but not the CPU speed, the lower-spec'd Xubuntu, MepisLite or Zen Walk distributions might meld well with your hardware.

But on a fast system, Puppy will indeed fly. With less memory, the swapping can really gum things up. Damn Small Linux is designed to work on even lower-spec'd hardware -- down to even 32 MB of RAM, I've heard. It's also designed to be a live CD but can be installed to become a more traditional Debian.

Since it's a thin client, the Thin Puppy isn't awash in peripherals -- or even places to plug them in. I could remove the CF adapter and install a 3.5-inch hard drive, but I doubt that I could hook up a CD-ROM drive, even temporarily, unless both drives could share a single IDE header (as there is only one). I'll have to look into it. Luckily the CF chips can be prepared for Puppy on another computer via a card reader -- the Puppy Universal installer takes care of that. I haven't yet figured out how to make a bootable Damn Small Linux CF chip or USB drive -- and the Thin Puppy won't boot from USB. I wonder if a USB-connected CD-ROM drive would even boot this $75 box. Probably not.

But as it stands now, I'm going to try to stuff the Thin Puppy with more memory at some point and see how she runs. Barring that, I've got to reassemble This Old PC (which has 262 MB of addressable memory) and get that connected to Ethernet to compare and contrast with the Thin Puppy.