lane@cs.dal.ca (John Wright/Dr. Pat Lane) (08/01/89)
OK, I understand about ROM shadowing (our 386 uses Chips & Technologies
chipset and AMI BIOS and can shadow both system BIOS ROM and video BIOS):
How the 384K from 640K to 1Meg (which becomes extended RAM at 100000 to
15FFFF in a normal 286 AT) is, in a 386 with the shadowing feature, mapped
mapped to A0000 to FFFFF and is used to hold copies of the system and video
BIOS's so they will execute faster (in RAM vs. ROM). In many/most 386's
turning off the shadowing feature does not return the 384K of RAM to the
extended RAM address space and renders the block of RAM useless.
(Am I right so far?)
What I'm curious about is the fact that the shadowing (on our machine)
affects only a 64K block at F0000 to FFFFF and two 16K blocks in the
C0000 to CFFFF range. That's only 96K. What about the remaining 288K?
Wasted silicon?
Also, what about video RAM. Even when the ROM is shadowed, is the video
buffer not still kept on RAM chips on the video adapter? Does this not
limit the speed of video processing, both because of the speed of the
on-board RAM and because of having to go through the bus (especially if
it's an 8-bit board - or does this even make any difference)? Would it
not be sensible to shadow this address space as well (esp. since the
unused system RAM is presumably just sitting there)? How about the TSR
program or device driver versions of video BIOS that can be loaded into
RAM (the precious under-640K variety) - what RAM do they use for a video
buffer?
While on the subject, I could use some pointers on how to juggle my MS-DOS
4.01 setup - ie. drivers HIMEM.SYS, EMM386.SYS, SMARTDRV.SYS, FASTOPEN.EXE
and the old BUFFERS=. For instance, with all those caching systems
(SMARTDRV is a regular disk cache, FASTOPEN caches directories, there's
even a 32K RAM cache) do I need any BUFFERS=? In this 386/25MHz with a
(noisey) 133Meg Priam drive and WD1003V-MM2 (or something like that)
controller, I find the hard disk sluggish. Also, I have less than 512K
of RAM available for running pgms even with all TSRs removed.
This was a problem when installing Windows/286. The Install pgm wants
512K min. If you pull everything out, install Windows, and put everything
back, Windows says "You've chnaged memory configuration; re-install for
best performance".
Is DOS 4 IBM's way of encouraging us to try OS/2? :-)
Many thanks for any enlightenment and assistance rendered?
John Wright ////////////////// Phone: 902-424-3805 or 902-424-6527
Post: c/o Dr Pat Lane, Biology Dept, Dalhousie U, Halifax N.S., CANADA B3H-4H8
Cdn/Eannet:lane@cs.dal.cdn Uucp:lane@dalcs.uucp or {uunet watmath}!dalcs!lane
Arpa:lane%dalcs.uucp@uunet.uu.net Internet:lane@cs.dal.ca
--
John Wright ////////////////// Phone: 902-424-3805 or 902-424-6527
Post: c/o Dr Pat Lane, Biology Dept, Dalhousie U, Halifax N.S., CANADA B3H-4H8
Cdn/Eannet:lane@cs.dal.cdn Uucp:lane@dalcs.uucp or {uunet watmath}!dalcs!lane
Arpa:lane%dalcs.uucp@uunet.uu.net Internet:lane@cs.dal.caJohn.Wright/Dr..Pat.Lane@mamab.FIDONET.ORG (John Wright/Dr. Pat Lane) (08/05/89)
-- Fidonet: John Wright/Dr. Pat Lane via 1:363/9 Internet: John.Wright/Dr..Pat.Lane@mamab.FIDONET.ORG Usenet: ...!peora!rtmvax!libcmp!mamab!John.Wright/Dr..Pat.Lane
james@raid.dell.com (James Van Artsdalen) (08/06/89)
In <1989Aug1.111017.1094@cs.dal.ca>, lane@cs.dal.ca (John Wright/Dr. Pat Lane) wrote: > Even when the ROM is shadowed, is the video buffer not still kept on > RAM chips on the video adapter? Yes. The video buffer is not shadowed, only the option ROM. The idea is that a large number of fetches are done to execute the code from the option ROM to print each character. > Does this not limit the speed of video processing, both because of > the speed of the on-board RAM and because of having to go through the > bus (especially if it's an 8-bit board - or does this even make any > difference)? Yes it limits performance. But if you use shadowing and never actually write to the video RAM, you'll never see changes on the CRT. One could cache the video buffer I suppose, but this is difficult as one would need to flush the video buffer fairly often (each time a different video page is chosen for example, or in [EV]GA graphics modes). 8 bit boards are in principle slower, but 16 bit boards are harder to make work. I don't know how much slower the 8 bit boards are in real life. -- James R. Van Artsdalen james@raid.dell.com "Live Free or Die" DCC Corporation 9505 Arboretum Blvd Austin TX 78759 512-338-8789