f



IBM 512k SVGA Chipset

Hello group...

I'm talking of the "famous" IBM SVGA chipset that appears on the 512K SVGA
adapter, Model 40/56/57 planars and the Model 25SX.

Of all these systems, only the SVGA adapter and the 25SX have the 512K of
VRAM included that's needed to make the chipset able to use SVGA modes like
640x480x256 and 800x600x16. I found some VRAMs that would work and have been
able to prove that at least the 57SLC can do SVGA modes when using the VESA
driver that IBM provides for use with DOS. The 40SX I have here only has
solder spots on the planar for extra VRAMs and I do not have enough
soldering skills to put my only set of VRAMs there to see if it works as
well.

While it is no problem to use another video adapter for MCA machines, the
25SX is somewhat stuck with the video it has onboard. Unfortunately, I don't
think IBM ever developed drivers beyond the DOS VESA driver and maybe an
OS/2 driver.

It is obvious that TI made parts of the 512k SVGA chipset for IBM...is there
any possibility (in the hope of finding drivers for, say, maybe Windows
3.1x) that TI ever made their own SVGA chipset?

William The Wonderer


0
William
12/21/2003 2:12:19 AM
comp.sys.ibm.ps2.hardware 10363 articles. 0 followers. Post Follow

15 Replies
9599 Views

Similar Articles

[PageSpeed] 48

Hi William !

First off: most of the "infos" I have on that infamous chipset is based on
talks with IBM tech-people in the late 80s / early 90s and may repeat a lot
rumours and half-through ideas.

Next: Back when the 8556 / 8557 came out IBM still judged PCs as "mainframe
add-ons" - their intention was not focussed on PCs as independent machines but
always with a mainframe in the background, acting as better terminals. The CPU
power of these 386SX was that low, that it were ridiculous to see them in a
complex graphic context like AutoCAD or realtime imaging. Seeing them as
"workstations" would be laughable. The best they could do was stand-alone DOS
(or OS/2) and terminal emulation.

The base idea on that chipset was to extend the capabilities towards Super-VGA,
which came neccessary for displaying non-latin characters like japanese,
chinese and other more calligraphic stuff, which has to use a larger pixel
raster than 8 x 8 to 9 x 12 bit patterns. This consequently requires a larger
VRAM along with a totally different character generator.

The video-section on 8556 and 8557 is largely based on the IBM chipset used for
the 50 / 60, 70 / 80 and 55 / 65 machines. As far as I recall they even used
the same Inmos 25MHz RAMDAC, which is in fact limited to 640 x 480 / 16 colors
@ 70Hz text and 60Hz graphics. The additional 256K VRAM (wasn't it 4 x 41464
SIPPs ?), which can be installed on the 56 / 57 would therefore not
neccessarily extend the screen size, but also the off-screen memory used for
RAM-based character generators or software downloadable fonts.

Additionally the Video BIOS seems to have some more modes.
It *is* capable to do some 800 x 600 SVGA mode (with reduced vertical refresh
because of the 25MHz RAMDAC) as well as a 132 x 25 text mode used for 3270- and
AS400-terminals. There is also something that IBM might have called eXtended
MCGA plus CGA-modes with shared attributes across an 8 x 8 pixel block as used
for PRESTEL modes. The developement on CGA / MCGA / EGA digital RGB branch
however has been stopped for the favour of analogue VGA / SVGA / XGA / SXGA
developements.

The EGA modes have been maintained for text modes and "Terminal Emulation"
modes. Even today you can do a "SCREEN 9" under GWBasic on the most modern
video cards. That SCREEN 9 mode is the HiRes EGA 640 x 480 / 256 colors mode
with shared attributes used by most graphic terminals like the 3278 terminal
for GDDM.

The short 512K SVGA Adapter (90EE) is a different story.
In the early 90s IBM looked for some cheap VGA card to substitute the
(relatively) expensive XGA-2 adapter for *servers*, where the primary purpose
is supervision of the machine rather than real *work* with it in Hi-Res. It was
just to supply a base video, where a XGA-2 were a waste of potential. They had
a contract with TI for some DSPs in multimedia already (the MWave for instance
is based on TI-DSPs as well as many Thinkpad internal chipsets) and TI offered
them a rather cheap - and inexpensive - chipset and combined it with a cheap
clock oscillator and an Inmos RAMDAC. That chipset was already pretty much
outdated at that time but IBM decided it would suffice for that low end
purpose.

Driver support was given under DOS and OS/2 only for base functions like
selection of the vertical refresh and few different modes only. Not even the
Win 3.x support has been finalized. Technically the adapter could do better
than VGA, but its video BIOS is largely undocumented and intentionally crippled
down to a few functions.

I am not sure if the chipset might have been used in Texas Instruments
portables as well or within the early IBM ThinkPad series (720 or 750). I
vaguely recall a 256-colour-support driver for the color Thinkpads, which would
require a 512K VRAM for 640 x 480 as well. But I don't know if that wasn't for
the odd WD chipset used on early Thinkpads as well.

While we're on it: Alternatively the Cirrus Logic SVGA (917B) was available
too.
Based on a CL reference design with the CL-5428 (ISA/MCA/PCI) chipset similar
to that used on the PS/Value Point Series 1 (CL-5422 on the 6384-Cxx / CL-5428
on PS/VP-Si 425SX and 433SX machines) there were two different cards available:
the IBM SVGA CL28 Adapter that Tatsuo described on his site (917A ? I think) -
which was the already-existing japanese version AFAIK and the infamous 917B
with the falsely mounted capacitor which was a later / alternative design.
Physically limited to 1MB VRAM and handicapped with an 80MHz integrated RAMDAC
the adapter never really took off. It ends at 1024 x 768 / 256 colors @ 70Hz
vertical refresh.

The chipset itself can handle 2MB VRAM - but the Video BIOS IBM used does not
make advantage of that, not even detects the additional RAM IIRC.

The very same CL-5428 chipset can be found on most Reply upgrade boards. Here
the sockets for the two 256K x 16 bit chips are present and the video BIOS
appears to be fully functional, detecting and supporting 2MB VRAM, which
allowes to go up to 1024 x 768 / 64K colors. There it is "only" a driver
question whether you get to see the higher modes or not.

Does that help somehow ?



Very friendly greetings from Peter in Germany
http://www.mcamafia.de/mcapage0/mcaindex.htm
0
peterwendt
12/21/2003 1:07:20 PM
> The base idea on that chipset was to extend the capabilities towards
Super-VGA,
> which came neccessary for displaying non-latin characters like japanese,
> chinese and other more calligraphic stuff, which has to use a larger pixel
> raster than 8 x 8 to 9 x 12 bit patterns. This consequently requires a
larger
> VRAM along with a totally different character generator.

Basicly, there is a difference between displaying text in graphics modes and
alphanumeric modes. Video Int 10h/Func 11h provides the character generator
interface -- for graphics modes, one can point the interrupt vector to a
user-defined glyph table residing elsewhere (!) in memory, whereas for
alphanumeric modes the glyph tables are loaded in the character definition
tables in the video RAM. Plain VGA can have up to 8 of them (EGA up to 4),
and both EGA and VGA can work with 9-byte characters (for a total of 512
characters).

As for the size of a BIOS resident glyph matrix (a bit pattern), it is a
function of the line resolution: it varies from 8x8 (200 lines, CGA) to 9x16
(400 lines, VGA), these are the default character sizes for alphanumeric
mode and appear somewhat compressed (elongated) when used in graphics modes.

Basicly, it would be difficult to work with alphanumeric character
definition tables for Asian characters (VGA, 512 chars with 9-byte codes),
since once has to take into account the readability as well, hence a
graphics mode provides the greater freedom of choice, also because one has
to consider how the chars will be entered, it is most often a combination of
several keystrokes for a single composite char.






0
Nihavent
12/21/2003 2:54:00 PM
Hi William,
> ...[the] chipset that appears on the ...Model 40/56/57 planars and 
> the Model 25SX...
     I see Peter has already more specifically corrected your 56/57
statement to know you mean the 8556/8557 planars. The 9556/9557 models
have XGA-2 as planar video (I say it not for you, but other viewers).
The 40SX planar is also shared with the 35SX, so include that model in
the list (there was also a high-end planar for the 35/40 much like
that of the Model 33 "E", with planar XGA-2).
                  David
                  David@gilanet.com
0
David
12/21/2003 8:20:54 PM
Hi !

>Basicly, it would be difficult to work with alphanumeric character
>definition tables for Asian characters (VGA, 512 chars with 9-byte codes),
>since once has to take into account the readability as well, hence a
>graphics mode provides the greater freedom of choice, also because one has
>to consider how the chars will be entered, it is most often a combination of
>several keystrokes for a single composite char.

Absolutely true.
As said in the beginning: a lot of my remarks is based on rumours and legends
passed on the campfires of IBM-freaked IT-staff over the years.

Fact is OTOH that the generic 256K VGA / EGA was insufficient for most enhanced
display stuff and even medium-resolution graphics required larger VRAM.
Remember having Win 3.1 running in VGA mode with 16 colours in 640 x 480 ?
A pain in the ass to say the least.

Most of the PS/55 domestic japanese machines of that time had capabilities
above that already - I hope there are people out to confirm that - since even
the simplified font set was almost unreadable with such a low screen resolution
in graphics, not to speak of text modes with the fixed and limited raster size.

But the reason *why* the 8556 and 8557 had a base 256K VGA expandable to 512K
is mainly guesswork. The chipset itself is mainly plain VGA and -after my
knowledge- there was no official video upgrade available.
At least I haven't ever seen one.
I myself stuffed in some RAMs I had at hand and tested around with a 57 to see
if I could get into some higher resolution modes. The effort was ... well ...
wasted. It seemed as if the modes were missing in the Video BIOS. The ABIOS
didn't give many hints and I didn't have any non-european references at hand to
compare if e.g. japanese versions had a different BIOS.

A co-worker played around with an 8557 M57 Ultimedia under OS/2, fully equipped
with the AVC stuff and he claimed that he'd managed to get the unit going in
256 colors @ 640 x 480 - a mode called "256-colour VGA" somewhere in the
drivers.
Practically that is the first SVGA mode already and that required the
additional VRAM already.

Whatever. When the Premium Line 56 / 57 came out the onboard 1MB XGA-2 made
things pretty much easier. I loved those boxes - still do and have some of them
around. I don't have am 8556 / 8557 anymore ... or only in the storage
probably.



Very friendly greetings from Peter in Germany
http://www.mcamafia.de/mcapage0/mcaindex.htm
0
peterwendt
12/21/2003 11:19:47 PM
Hi David !

>(there was also a high-end planar for the 35/40 much like
>that of the Model 33 "E", with planar XGA-2).

To be precise: that *is* the 9533 PS/2e planar.

The machines are therefore labelled "9535" and "9540", since they have a
display adapter above VGA and belong to the Premium Line - hence 95xx.
I have seen one once. Martin from the Netherlands brought the one he owns
along, when he visited me.



Very friendly greetings from Peter in Germany
http://www.mcamafia.de/mcapage0/mcaindex.htm
0
peterwendt
12/22/2003 9:08:26 AM
Hi Peter,
>> (there was also a high-end planar for the 35/40 much 
>> like that of the Model 33 "E", with planar XGA-2). 
> To be precise: that *is* the 9533 PS/2e planar. The machines are therefore 
> labelled "9535" and "9540", since they have a display adapter above VGA 
> and belong to the Premium Line - hence 95xx. I have seen one once. Martin 
> from the Netherlands brought the one he owns along, when he visited me.
     I meant to say 8535/8540 (as I am sure William did with the
8556/8557) when talking about the 35SX & 40SX. Since I am getting a
couple more 9533s I have thought about doing a conversion to a 8535 or
8540 case. The riser might be the only worry (but I have a number of
clone ISA risers I could graft into the top of the 33 riser). My study
showed the 25SX riser (two normal ISA slots high & one on the reverse
side) to look like it would work on the 9533/9535/9540 planar.
                    David
                    David@gilanet.com
0
David
12/22/2003 5:25:15 PM
> > (VGA, 512 chars with 9-byte codes),

9-bit codes, of course. "9-bytes" is a too good invention.

> But the reason *why* the 8556 and 8557 had a base 256K VGA expandable to
512K
> is mainly guesswork. The chipset itself is mainly plain VGA and -after my
> knowledge- there was no official video upgrade available.

Aspiring to be SVGA, I guess. There were controllers at the time which did
800x600/16 colors with 512K VRAM, and the SVGA was not yet established.
Siemens had a card called ERGO VGA, basicly Tseng ET2000, IIRC, and it had
its own drivers for Win 3.x to supply the "non-standard" VGA modes (even a
1024x768/16 colors) which later become de-facto SVGA standards. Those 512K
or 1M video controllers required custom drivers, and the custom video modes
on the plain VGA have been too an experiment (at least how to fry a monitor
when you get it wrong). But remember, all this coincided with Win 3.x and
people were paying for higher video resolutions and more video RAM. May be
the latter was the true reason, to market and sell it better.




0
Nihavent
12/22/2003 8:13:51 PM
Hi David !

>Since I am getting a
>couple more 9533s I have thought about doing a conversion to a 8535 or
>8540 case. The riser might be the only worry (but I have a number of
>clone ISA risers I could graft into the top of the 33 riser).

Ooof !

If you have a -say- 8535 already you have a matching riser as well I think
..
AFAIK there were no differences between them. Clone risers tend to have "silly
surprises" buried in them - where "Reserved" pins are tied to GND or + levels
for instance - or have totally different pinout at all.

I must have a 35SX or 35SLC (no-media model) in the Hamburg office ... if it
hasn't been disposed in the meantime.

Time to compare risers with a multimeter. ;-)



Very friendly greetings from Peter in Germany
http://www.mcamafia.de/mcapage0/mcaindex.htm
0
peterwendt
12/22/2003 11:21:44 PM
Hi!

Many thanks for typing that up...it does answer some of the questions I had
and provides an interesting history on the chipset.

However, I'm still wondering if IBM cared to write drivers for any operating
systems beyond OS/2 and DOS?

I'd love to find some drivers that could get a 25SX (where the 512K video
subsystem is the final word in video, unless perhaps I get out the soldering
iron and try to burn the house down) displaying 640x480x256 under, say
Windows 3.x. . I know that the 256 color display can be done on the
25SX--Fractint and the IBM VESA DOS driver were able to hit a nice
640x480x256 with no problems.
I even got an out of sync 800x600x16 display that way.

William


0
William
12/27/2003 2:29:21 AM
Note: I am _NOT_ a windoze 3.xx expert!

I thought, however, there was a "generic vesa" driver for windows 3.11 (the
"windows for workgroups" version)  That way, you loaded the VESA TSR
program for your video card then ran windows and voila.

I could easily be wrong on this one, but I seem to remember using this
"trick" for non-natively supported SVGA cards.



"William R. Walsh" wrote:

> Hi!
>
> Many thanks for typing that up...it does answer some of the questions I had
> and provides an interesting history on the chipset.
>
> However, I'm still wondering if IBM cared to write drivers for any operating
> systems beyond OS/2 and DOS?
>
> I'd love to find some drivers that could get a 25SX (where the 512K video
> subsystem is the final word in video, unless perhaps I get out the soldering
> iron and try to burn the house down) displaying 640x480x256 under, say
> Windows 3.x. . I know that the 256 color display can be done on the
> 25SX--Fractint and the IBM VESA DOS driver were able to hit a nice
> 640x480x256 with no problems.
> I even got an out of sync 800x600x16 display that way.
>
> William

0
Everyone
12/27/2003 4:25:07 PM
Note: I am _NOT_ a windoze 3.xx expert!

I thought, however, there was a "generic vesa" driver for windows 3.11 (the
"windows for workgroups" version)  That way, you loaded the VESA TSR
program for your video card then ran windows and voila.

I could easily be wrong on this one, but I seem to remember using this
"trick" for non-natively supported SVGA cards.



"William R. Walsh" wrote:

> Hi!
>
> Many thanks for typing that up...it does answer some of the questions I had
> and provides an interesting history on the chipset.
>
> However, I'm still wondering if IBM cared to write drivers for any operating
> systems beyond OS/2 and DOS?
>
> I'd love to find some drivers that could get a 25SX (where the 512K video
> subsystem is the final word in video, unless perhaps I get out the soldering
> iron and try to burn the house down) displaying 640x480x256 under, say
> Windows 3.x. . I know that the 256 color display can be done on the
> 25SX--Fractint and the IBM VESA DOS driver were able to hit a nice
> 640x480x256 with no problems.
> I even got an out of sync 800x600x16 display that way.
>
> William

0
Everyone
12/27/2003 4:32:18 PM
Note: I am _NOT_ a windoze 3.xx expert!

I thought, however, there was a "generic vesa" driver for windows 3.11 (the
"windows for workgroups" version)  That way, you loaded the VESA TSR
program for your video card then ran windows and voila.

I could easily be wrong on this one, but I seem to remember using this
"trick" for non-natively supported SVGA cards.




"William R. Walsh" wrote:

> Hi!
>
> Many thanks for typing that up...it does answer some of the questions I had
> and provides an interesting history on the chipset.
>
> However, I'm still wondering if IBM cared to write drivers for any operating
> systems beyond OS/2 and DOS?
>
> I'd love to find some drivers that could get a 25SX (where the 512K video
> subsystem is the final word in video, unless perhaps I get out the soldering
> iron and try to burn the house down) displaying 640x480x256 under, say
> Windows 3.x. . I know that the 256 color display can be done on the
> 25SX--Fractint and the IBM VESA DOS driver were able to hit a nice
> 640x480x256 with no problems.
> I even got an out of sync 800x600x16 display that way.
>
> William

0
Everyone
12/27/2003 4:32:42 PM
Hi!

> Note: I am _NOT_ a windoze 3.xx expert!

I think you're on the right track however. I heard rumors ages ago of such a
"generic VESA" driver for Win 3.1x, but I can't think who might have made
one and Googling on it hasn't seemed to help.

I'll keep looking...anyone with any hints, thoughts, files, etc...please
tell me!

William


0
William
12/28/2003 5:54:27 PM
William R. Walsh wrote:
> Hi!
> 
> 
>>Note: I am _NOT_ a windoze 3.xx expert!
> 
> 
> I think you're on the right track however. I heard rumors ages ago of such a
> "generic VESA" driver for Win 3.1x, but I can't think who might have made
> one and Googling on it hasn't seemed to help.
> 
> I'll keep looking...anyone with any hints, thoughts, files, etc...please
> tell me!
> 

If I remember right, there was a 800x600 16 colour driver for most SVGA 
Chipsets in Windows 3.1. I used it with an ISA Trident TVGA9000 without 
VESA TSR.
Worked with ATI 28800 and ET4000 too.

No generic VESA driver for 256 colour modes.

---
Uli

0
Ulrich
12/29/2003 12:08:49 AM
"William R. Walsh" <newsgroups1@idontwantjunqueemail.walshcomptech.com>
wrote in news:nNEHb.685702$Fm2.590554@attbi_s04: 

> Hi!
>
>> Note: I am _NOT_ a windoze 3.xx expert!
>
> I think you're on the right track however. I heard rumors ages ago of
> such a "generic VESA" driver for Win 3.1x, but I can't think who might
> have made one and Googling on it hasn't seemed to help.
>
> I'll keep looking...anyone with any hints, thoughts, files,
> etc...please tell me!
>
> William
>
>

Hi

www.simtel.net still have a large list of vesa drivers
use vesa in the search engine.

univbe was the most popular a shareware program used to come bundled
with early 3d games shooters such as doom and duke nukem etc.... 


Cheers......
0
Cheers
1/3/2004 1:09:31 PM
Reply: