I purchased a Genesi Pegasos 2 motherboard with a 1GHz 7447 processor from Swedish dealer GGS-data in March 2004. The system was to be used for the GMP team.
Unfortunately, it has been an extremely sour experience thus far.
Pegasos 2 is a MicroATX motherboard that comes packed with features. It is actually smaller than most x86 MicroATX motherboards. The processor is connected on a SEC contact, similar to the how Pentium 2 and early Athlons where mounted.
The firmware (ver 1.1) is very buggy. I've seen buggy firmware before, but this one takes the cake. To make matters worse, the manuals--both the printed ones that come with the board and the "updated" versions on the web, are full of errors and typos. A certain non-reproducibility within the firmware, paired with incorrect documentation is a frustrating combination.
When I bought this hardware, the web sites of the Pegasos vendors claimed that MorphOS, several GNU/Linux distributions, and OpenBSD were all working on the system.
With my first motherboard, not even MorphOS would install properly.
After the first motherboard had managed to get into some firmware state causing it to hang early, GGS-data replaced the motherboard. GGS-data's analysis of the cause of the problems with installing MorphOS is that I didn't use a pristine IDE hard drive. The partition program assumes disks are clean, and can apparently make the system permanently non-bootable when a non-pristine drive is used, possibly due to inconsistencies in firmware variables. (Note that the Format function in this program is a dummy; they only implemented the button and not any actual formatting code. There are other remedies, a program called HDwrite is mentioned in the troubleshooting section. But since this isn't mentioned in the step-by-step instructions, and since my first motherboard was completely locked, the troubleshooting section was of little help to me. And who would expect a partition program to fail its main task--to partition and initialize a drive?)
With the replacement motherboard, and by using HDwrite before the hardware had locked, I could get MorphOS onto the non-pristine disk. MorphOS is a toy environment that I don't intend to run, but the installation guide for GNU/Linux suggests that MorphOS should be installed first.
The Genesi web sites as well as the web site of GGS-data promised at the time I purchased my board that OpenBSD and Linux were ready and working. Now the web sites only promises working Linux. OpenBSD dropped development after Dale Rahm did not get the agreed payment from Genesi for his work, and after show-stopper problems with the firmware.
The truth is that GNU/Linux doesn't really work yet. It is being ported, but it is not stable at this point. Also, installation is not reliable due to bugs in the firmware, bugs in the installation program, and due to the fact that the bits you download with the installation program is in rapid flux.
The attitude of the Linux developers working for Genesi is that everything is ready. "Do you have problems? You're the first one to complain, you've surely made something wrong! But, try this new firmware and that new kernel, they are sure to cure any problems! Still problems? Try this newer kernel and flash your firmware again! Install from scratch! Everything *is* solid and has always been!"
I have urged GGS-data to stop telling people that they ship the systems with Linux installation CDs. They don't ship any such CDs of course, since such CDs don't exists since the installation CDs aren't ready since Linux isn't ready.
If you are willing to flash your firmware with a version of yesterday, if you think a kernel older than 24 hours shouldn't be expected to work, and if you think running shells scripts without frequent spurious EAGAIN errors is a luxary, then by all means get yourself a Pegasos 2. If you want a useful system, you should wait for confirmed stable operation. There is hope for these systems, but they are not there yet, whatever the vendor web sites claim.
Update 2004-12-21: In the end, we kept the system, to avoid a legal fight with GGS-data. Unfortunately, the system is unstable and fails the daily GMP builds every night.
Update 2005-07-19: Last spring, we found that the little CPU fan had stopped, due to a failed bearing. GGS-data replaced the processor module and the memory module to see if the system would get stable. Alas, it is still as unstable as ever.
Torbjörn Granlund
Long-term GNU volunteer
PS. If you dear GGS-data owner and you dear Genesi folks deny this story, I'd be more than willing to publish our mail conversation here...