Paul.Birkel@K.GP.CS.CMU.EDU (09/10/88)
I have a Xebex AMNESIA PC/XT expanded memory board which I purchased 2 years ago and installed on a Zenith Z-148. I was never able to get the %@$#! vdisk in expanded memory to work properly, and after awhile I gave up -- assuming that the problem was related to way that Zenith handled memory address decoding above 640k (I had schematics). Now I have moved the card to a true Blue PC-Portable and have encountered the exact same problems. Basically, any file larger than 16k is either garbaged, destroys the FAT, or freezes the machine to the point that a power cycle is required. This is LIM 3.2, and as I recall one "swaps" in 16k chunks ... sounds to me like the software. It's using a page frame located at CC00:0000 and I/O addresses 258/259. So ... Xebec tells me that they no longer support the product, did have several software upgrades, but don't know if they have any left! The release of the driver is 1.24, and I've tried using both MS-DOS 3.1 and PC-DOS 3.0. With MS-DOS I at least get the file "copied" into the VDisk and can run dir on it and attempt to execute the file. With PC-DOS the results of dir are mostly garbage characters, and trying to execute the file results in "file not found". Sometimes the lock-up occurs during the copy with the source drive still active (enabled). With MS-DOS 3.1 (which sorta works) the VDISK is ver. 3.02. My configuration is with three banks of 256k of which 384k are used to backfill, and 384k are assigned to VDISK. Sector size is 128 and directory entries are 64. The other behavior I observe, which may be related, occurs during the loading of the driver when it does a memory test. Page 0 moves along reasonably, but then it either zips through or skips page 1 and the first part of page 2, the remainder of page 2 starting at about 28 or maybe 2A then runs at the "normal" speed. This behavior seems inconsistent with what I would presume would be the "obvious" mapping: first the 1 1/2 banks of backfill and then the 1 1/2 banks of extended memory. I therefore assume that for reasons of their own they apply a different mapping that is advantageous for some other reasons. I haven't tried determing whether it's the slow or fast testing pages that correspond to extended memory. I may be adding another bank of memory soon which should make the sizes of the 2 memory portions asymetrical. My guess is that they're testing only the pages used for extended memory since the POST has already tested the rest. So this may be a "feature" rather than a defect. If anyone has such a board with a later release (>1.24) or a working release of the Xebec driver or some ideas about how to proceed, I would *greatly* appreciate hearing from you. Thanks for your attention, paul Paul A. Birkel Dept. of Computer Science Carnegie-Mellon University Pittsburgh, PA 15213 (412) 268-8893