About gfx cards. (may be we need another std?)

Prophet (4086)

06-09-2008, 12:22

reading the previous posts http://www.msx.org/forumtopicl8936.html , http://www.msx.org/forumtopicl8923.html
and various GFX9000 or VDU VSU posts, figured me that there is a wish to a good gfx std for msx.

Let's summarize what there is today with IMHO pro and cons:

MSX1 VDP:

PRO:
Almost all pattern mode driven
Simple to program
CONS:
Poor sprite support
Lack of palette
Lack of speed
Lack of scroll register
colourclash limitations
a little limited in screen resolution

MSX2 VDP:

PRO:
Bitmap modes for static images and more colour
Raster interrupts
Blitter
Palette
Vertical scroll register

CONS:
Sprite support is yet to much limited
Lack of horizontal scroll register
pattern modes are not without colour clash

MSX2+ VDP:
PRO:
Same as MSX2, with a little better support for coloured gfx images.
Horizontal scroll register
Extended blitter support
CONS:
There is no much support. Still some architecural limits inherited from previous VDPs for compatibility
Blitter Speed.

V9990:
PRO:
Fast,
DualPlane pattern mode support
Great sprite support
Blitter at decent speed
Good amount of VRAM
Because compatibility was dropped down, does not have to make compromises in terms of performances or architecure.

CONS:
Not compatible with olders VDP
Pratically there is no sw that used it....
There are no unlimited chips available. Less than other vdps.

So there is no ideal gfx standard for msx. Why not to develop a new VDP card to set as the video 2.0 standard for msx?

My idea is to keep the pro and cons of the different VDPs:

- No older vdp compatibility, it's a new standard that is added to the existing machines.
- msx1 compatibility
- no super highres or coloured modes, only a smart couple of pattern modes where each pattern could be used as background or sprite, as in screen 2 the correct arrangement of tiles can make it suitable for bitmapped gfx. 8 bit gfx, palette support
- no speed limitation
- programmable blitter support (not hardwired commands, instead a programmable graphics processor)
- modern compatible monitor output

I think there are great possibilities to realize this on FPGA devices, and this remove the problem of chip availability.

By wolf_

Ambassador_ (10076)

06-09-2008, 12:31

How about these:

- traditional resolutions, 256/320 or 512/640, but also 128/160 modes for demo effects
- non-reactive sprites: just blitting pieces of image together, no native sprites required (think starfields, overlay scores, particle systems etc.)
- reactive sprites: a certain number of fullcolor sprites on screen, no per-line limit, but a screen-limit is ok: 128 or so.. the 'videocard' returns when non-transparent sprite pixels overlay eachother and returns the number of the sprites involved
- native scaling for the blitter

By PingPong

Prophet (4086)

06-09-2008, 12:39

How about these:

@Wolf: you are a little to over my idea of specs level, for an 8 bit computer . However, i think all - msx - generations compatible video system is good, because it's greatly expand the number of people that can try it, thus expanding the possible sw support.

I think that if, a similar video card is available (at a reasonable price) a lot of msx people will be great to buy it.

Time to trash the VDP family in the bin, let's restard in a clean way....

By Edwin

Paragon (1182)

06-09-2008, 13:41

a certain number of fullcolor sprites on screen, no per-line limit, but a screen-limit is ok: 128 or so..

There is a good reason why sprites per row have always been limited. Transparency checks require a lot of processing power and bandwidth. Just increasing the number would have had a huge impact on the required performance.

By PingPong

Prophet (4086)

06-09-2008, 13:44

Maybe hw sprites can be dropped if replaced by a good gfx engine.... today no one uses those hw things.....

By wolf_

Ambassador_ (10076)

06-09-2008, 13:53

Well, that was the idea indeed. You do however need to get a signal when those pieces of screen graphics collide with eachother. That's why I proposed reactive and non-reactive sprites. And whether they're screengraphics or real sprites isn't much relevant, collisions are tho.

By PingPong

Prophet (4086)

06-09-2008, 15:27

@Wolf: May be it's a good thing to know what are the specs that one want. I think a similar card is feasible. In this forum there are people that can manage FPGA quite well, and having feedback from those is a good thing.....

By Pentarou

Champion (486)

06-09-2008, 16:22

By PingPong

Prophet (4086)

07-09-2008, 09:08

There are lots of interesting projects not MSX related but that can be easily adapted, for example this page: http://www.retroleum.co.uk/index.html or all the Russian MSX-Spectrum upgrades.
For a graphic expansion especially this one: http://www.retroleum.co.uk/z80-v4z80-gfxboard.html looks interesting and CHEAP!
It's a good video-board. (56 on a row 16x16 sprites each one @256 colors? wow!)
Maybe some expert guy in vhdl can tell us how much effort is needed to adapt to a msx?

By Edwin

Paragon (1182)

07-09-2008, 12:03

Looks like that board is supplied in schematic files with some Verilog conversions added. That makes it a bit hard to see what's included. I'm not quite the expert, but it looks like it's very much tied to a xilinx component library. Which will make converting it to Altera software for using it on 1cM quite a job.
The guy does have a nice project going on there though!

By PingPong

Prophet (4086)

07-09-2008, 13:29

But the question is another:
Is a similar gfx board doable for msx?
What the estimate cost?
Do we need a state-of-the-art PLD to achieve this goal or it's only needed an entry level PLD?

Unfortunately i cannot give those aswers because of my almost zero-knowledge of VHDL, PLD and similar things.
So i ask anyone that know about this technology.

thx