🚀 go-pugleaf

In article <15af3b53@p1.f36.n245.z2.fidonet.org>, Thomas_Hornschuh@softstream.fido.de (Thomas Hornschuh) writes:
|> PP> The downgrading of the VGA card to 8 bit has to do with the mere
|> PP> detection of an 8 bit MDA on the bus. It does not matter whether you use
|> PP> reverse video or not. On the other hand, I have found a program that
|> PP> lets you force an ET4000-based S-VGA to 16 bit even when the MDA is
|> PP> there (of course, the MDA is unuseable in that mode, but it beats
|> PP> openning the cover to swap the MDA in and out depending on whether you
|> PP> are going to fire up the debugger or not).

Any details on that program?

|> It depends on the MDA/VGA/mainboard combo if it becomes unusable. My MDA work 
|> fine, even with the ET4000 in 16Bit mode. The VGA must decide quickly, if a 
|> memory access in the segment AXXX/BXXX is for the MDA or itself. In case of an 
|> MDA it must not activate the /MEMCS16 signal on the bus. The mainboard need 
|> this information very early in the bus cycle. If the VGA is faster than the 
|> mainboards "time out" for the signal, the MDA is accessed correctly with 8Bit. 
|> My combination is a OPTIMA Mega ET4000 card, some noname MDA a mainboard with 
|> 386/40Mhz and OPTI chipset (TMC PAT38PX). 
|> 
|> 
|> PP> I also had a discussion with a person on Compu$erve who seemed fairly
|> PP> knowledgeable about how a S-VGA video buffer is accessed where he
|> PP> mentionned 16 bit read/writes are fairly rare anyway. It turns out that
|> PP> the downgrading to 8 bits is actually felt only when executing code from
|> PP> the VGA card's BIOS which is something that happens under full DOS, but
|> PP> not in Windows.

I vaguely remember that companies use "16-bit vga" to mean different things:
16-bit ROM access, 16-bit video RAM access, 16 days to the week, etc.

|> ET4000 cards supports 16Bit video buffer access in all graphic modes. The 
|> effect of limiting the access to 8Bit, depends on the amount of 16Bit accesses 
|> in the video drivers code. Most plane oriented 16 color drivers conceptually 
|> cannot use 16Bit very often. But the 256 color drivers can use 16Bit access for 
|> many of the BitBlt operations. The amount of speed degradation is  proportinal 
|> to the quality of the driver implementation. Good drivers will suffer most from 
|> 8Bit access.

I just attempted to put a MDA in my system with a SpeedStar-Plus (ET4000) board
running 1024x780x256.  The slowdown was very noticeable, especially on full
screen repaints.  I had a program which checked video performance, and it told
me I had 24 wait states (on a 486/33) accesing video RAM.  With the board out
the wait states went down to 9.  The driver came from WndSpeed.

The slowdown was bad enough that I decided to keep the MDA out of the system.
(since I do V&V I never need to debug anyway :-)

jim mullens
mullens@jamsun.ic.ornl.gov (128.219.64.31)
voice: 615-574-5564