This machine used to be my main server, but is presently disused.
It has gotten fairly seriously "hacked on" and has been used to supply spare parts for other systems, so that any correlation between the configuration mentioned below and reality may be purely coincidental. It's liable to either disappear or to spontaneously turn into a PCI bus machine. But there are some useful notes here nonetheless.
This chip is pin-compatible with Intel 80486es, and internally represents a "fast clone" of the 486. (The similarity may go further than mere "similarity;" AMD's early chips were identical to Intel's as AMD was a "second source" for the products.) The claim I hear is that this is about equivalent to an Intel P5 running at about 75MHz. I'd tend to agree.
As it only supports 500KBit/s FDC data transfers, this prevents my tape drive from using it with the high-capacity QIC 3020 format.
I attempted to "upgrade" to a SIG VLB/EIDE/Multi-IO card with the faster FDC. (Picked it up at Computer City; model #CN4205.) Unfortunately, it did not happily co-exist with Linux for some inexplicable reason.
More information on this can be referenced at Linux Floppy Tapes.
My attempts to get the TR-3 tape drive working are now pretty stagnant as I now have an array of other backup devices, including SCSI tape drives and a CD burner, with corresponding improvements in performance and usefulness.
See also FTAPE Page for more details.
With all the "plug-and-play" configuration entertainment that ensues...
The Viper uses the Weitek "P9000" chipset which is unfortunately not supported by SVGALIB. To ease X configuration, I (when I, rather seldom, run X on it) run the Accelerated-X X server.
This machine is very much a hodge-podge of components acquired from various places Old IDE disk drives, CD-ROM, and video card were raided from an old system, case and monitor came from an Addison-based surplus store called CompuSurplus, newer IDE drive bought at some "outlet" computer store. It's all been in pieces and all been dysfunctional at various points in time.
Once legacy files get dumped off, this machine ought to become a "test bed" for OS experimentation; I'd like to try out some combination of:
Wolfe and Dantzig created a scheme known as Dantzig-Wolfe Decomposition that splits large linear programs into a set of subproblems (that are themselves LPs) that are connected together using a "master problem." These pieces are all smaller and more easily solved than the original LP.