Making vintage computers

Not terribly. You would want to re-design the CPU using a single Xilinx FPGA instead of the two Altera parts (18 years is a long long time). The XKL-1 board is nowhere near state of the art. You would need to guarantee the bug-for-bug compatibility of the floating-point instruction sets. Memory sizes/prices have moved way beyond what was possible in 1993, when the XMG-1 board was designed. We could only offer 32MW per board, and the 512MW board design died when the engineer moved on to other pastures. At this point a 1GW board would make the most sense--just fill up the entire address space. The XNI-1 4-port Ethernet card is only capable of 10baseT and 10base5/10base2 (via AUI connector and a Cabletron box). Create a 100baseT or 1000baseT board instead, with or without multiple ports. FASTWIDE differential SCSI-2 ("HVD")? If you stay SCSI, go for LVD U320. Better, introduce SATA to the 36bit world. You also have the problem that the TOPS-20 filesystem maxes out at ~24GB (as in, the on-disk data structures simply do not have any way to handle more data than that), so you're going to have to do some monitor work--either introduce the concept of partitions, a la Unix (because those are only somewhat less bletcherous than DOS partitions), or devise an entirely new filesystem for TOPS-20 over which you can lay the original filesystem semantics to preserve your software investment. Don't get me wrong. The Toad-1 was a great idea for its time, and I'm really sorry that we missed our market window, but that time is past. The best way to get superior PDP-10 cycles is to run the KLH10 emulator (a commercial class program, as opposed to the toy provided by SimH). On X86 processors, the speed of the emulated CPU is very nearly proportional to each 100MHz of the base clock speed; I haven't seen the benchmarks for X64 or PowerPC, but they are probably in the same ballpark. The TOPS-20 monitor has not been tweaked for larger filesystem sizes, but you can emulate more disks than most sites ever had (8 per channel, 6 channels--or go without Ethernet and get 8 channels). The current maintainer of KLH10 is not interested in taking TOPS-20 beyond 27 bits of addressing (unlike the Toad-1's full 30-bit address space), but others might be persuaded to take that on. Rich Alderson Vintage Computing Server Engineer Vulcan, Inc. 505 5th Avenue S, Suite 900 Seattle, WA 98104 mailto:RichA at vulcan.com (206) 342-2239 (206) 465-2916 cell