home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
Shareware 1 2 the Maxx
/
sw_1.zip
/
sw_1
/
TEXT
/
S3COM4.ZIP
/
S3COM4.TXT
Wrap
Text File
|
1992-08-28
|
2KB
|
49 lines
TECHNOTE: S3 86C911 GUI Accelerator and COM4 port conflicts
This Technote explains problems arising from a conflict between S3
video cards and COM4 I/O cards. Symptoms on video cards can be
corrupted screen fonts. Symptoms on COM4 can be data transmitted
without the port being used (modems making sounds) or baud rate
changes. Other maifestations may present themselves on both.
History:
Nearly ten years ago the IBM Technical Reference Maunal for the XT
presented conflicting documentation. 1) Pages 1-13 end in a
paragraph that states, "The channel is designed so that 768 I/O
device addresses are available to the I/O channel cards.", meaning
0 to 3FF. 2) Pages 1-22 include an I/O Address Map which lists
four Cluster adapters, all addressed above 3FF.
The first statement led manufacturers to assume that all port
addresses would be between 0 and 3FF and the second statement
immediately started the trend toward using any of the 64k
addresses provided by the Intel 8088.
The base address for COM4 is 2E8, meaning 02E8, not ?2E8. The
first digit is a 0, not ignored. Most Comm boards being
manufactured even today, still assume that there will always be a
0 in front of the I/O address, so they respond to any address that
matches in the last 3 (actually 2.5) digits.
The S3 86C911 uses over a dozen addresses for which a COM4 port
that only cares about addresses less than 400 would respond to.
COM4 boards that do not ignore addresses above 3FF will both listen
to data for the S3 chip (making the COM4 port corrupt) and respond
to requests for data from the S3 chip (destroying data from the S3
chip). Both the S3 86C911 and COM4 can exhibit problems, but the
hardware fault lies solely on the COM4 port, NOT the S3 chip!
As a comment, the realities are, IBM led manufacturers into the bad
practice of using only half the I/O address lines available,
manufacturers still design I/O cards this way even though these
conflicts have been around for years, and S3 assigned addresses in
the 86C911 that conflicted with these cards. Intel provided 64k I/O
locations and IBM did not physically restrict the use of these, and
as we all know in the PC industry, if a resource is available it will
be used.
Stuart Wyatt
Chrisalan Designs, Inc.
73770,735