It's downright vexing how good MSX artists were at designing for 9918 graphics

[+pixelpedant]

 

Looking at some of the finer graphics from MSX1 games (working with the same bitmap graphics as our TI-99/4A), you really get a sense of what's possible with this chip, from those with the artistic ability. 

 

Just for example:

 

 

 

 

Now, to be fair, nobody in North America was creating video game art like this, in the TI-99's heyday.  This is all late eighties Japanese stuff. 

 

But I suppose that's just all the more inspiration to take the TI-99 further today.  So much was possible, that could never have happened in the TI-99's own time. 

 

You can do a whole hell of a lot, with 15 colours. 

 

And after all, Realms of Antiquity, and Dragon's Lair, and others, have done pretty well for us, in recent times:

 

[Tuxon86]

14 minutes ago, pixelpedant said:

 

Looking at some of the finer graphics from MSX1 games (working with the same bitmap graphics as our TI-99/4A), you really get a sense of what's possible with this chip, from those with the artistic ability. 

 

Just for example:

 

 

 

 

Now, to be fair, nobody in North America was creating video game art like this, in the TI-99's heyday.  This is all late eighties Japanese stuff. 

 

But I suppose that's just all the more inspiration to take the TI-99 further today.  So much was possible, that could never have happened in the TI-99's own time. 

 

You can do a whole hell of a lot, with 15 colours. 

 

And after all, Realms of Antiquity, and Dragon's Lair, and others, have done pretty well for us, in recent times:

 

IMHO, the Ti-99 problems wasn't it's graphic chip, it was the CPU and the rest of the architecture. The MSX used the Z80 a more common processor with a vastly bigger talent pool...

[PeteE]

It's amazing what some artists can do with only 15 colors.

[+pixelpedant]

1 minute ago, PeteE said:

It's amazing what some artists can do with only 15 colors.

Oh wow.  That's awesome.  Though it looks like there was no attempt to take attribute clash into consideration, so using them as 9918A graphics in actual practice will in a lot of cases require sprite layering.  Still, a remarkable resource.

[+OLD CS1]

1 hour ago, PeteE said:

It's amazing what some artists can do with only 15 colors.

Taken just for the colors, this is nice, but otherwise not exactly a fair presentation. While he used the 9918's palette he did not honor the color occupation restrictions.

[+pixelpedant]

Yeah, what those MSX1 screen graphics I put up top have in common is they're all extremely clever about working around attribute clash in a way that almost completely hides it, and makes the bitmap mode MSX folks call Screen 2 (and TI folks mostly just call bitmap mode) look like a 4-bits-per-pixel bitmap mode.  From a distance, this just looks like an EGAish bitmap screen graphic:

 

 

But up close, no tile segment ever contains more than two colours:

 

[Tuxon86]

49 minutes ago, pixelpedant said:

Yeah, what those MSX1 screen graphics I put up top have in common is they're all extremely clever about working around attribute clash in a way that almost completely hides it, and makes the bitmap mode MSX folks call Screen 2 (and TI folks mostly just call bitmap mode) look like a 4-bits-per-pixel bitmap mode.  From a distance, this just looks like an EGAish bitmap screen graphic:

 

 

But up close, no tile segment ever contains more than two colours:

 

That was the name of the game back then. 

[HOME AUTOMATION]

3 hours ago, OLD CS1 said:

Taken just for the colors, this is nice, but otherwise not exactly a fair presentation. While he used the 9918's palette he did not honor the color occupation restrictions.

Yeah, that's right...

 

?

[+jedimatt42]

It doesn't hurt that they were professionals.

[wierd_w]

It really *IS* the chunkiness of the TI VPU's paletting that is onerous and vexxing, yes. (EG, an entire tile can contain no more than 2 colors)

 

If I could get around some of the mode restrictions I could do some pretty nice pictures, if I didnt have to deal with that.  Tursi's graphics shuffling methods to interrupt the scanline after 8 pixels being drawn opens a lot of doors, but it also puts a heavy strain on the CPU to constantly scribble on the vram.  Its no good really for a very screen update intensive situation, especially if the 9900 has to be doing actual game processing too. (also, does nothing to address the same restrictions on sprites)

 

What the TI really needed, was a way to bus-master the memory bus. Then a small dedicated CPU could have lived on cartridges, similar to how the NES and SNES did stuff, and a lot of fancy things could have been possible.  Amusingly, the TMS9900 DOES have such a line, it just never gets broken out on either the cartridge slot or the sidecar bus.  (It tells the TMS9900 to just wait as long as it is raised. It is intended for just this kind of thing-- some other device or processor doing stuff with memory.)

 

It would then be possible to get the same "8 pixel per scanline" trick, without the 9900 being the one doing the interruption/scribbling on the memory. (It would be the secondary processor doing it.) Some really nice looking titles would be doable that way.

 

I suppose it might be possible even without the bus hold line with creative use of bank switching, assuming the switching can happen fast enough and reliable enough-- but out of bank memory manipulation would still require a secondary processor. Maybe the strangecart could do something like this? It would DEFINITELY be fast enough.

 

 

[sometimes99er]

11 hours ago, pixelpedant said:

 

 

When you keep "black" your base / background color, things can work out quite nice.
 
The TI-99/4A base console had no access to machine code, so only a few got into the more interesting possibilities (of the 9918A) - and then the audience was again limited - the cheapest upgrade to go there was the MMM (Mini Memory Module).
 
?
 

[sometimes99er]

Also remember that the ZX Spectrum, which actually have less graphic power than the 9918A, had tons of beautiful graphics. Bottomline. The stock TI-99/4A was too limited (no access to machine code).
 
?
 

[Willsy]

6 minutes ago, sometimes99er said:

Also remember that the ZX Spectrum, which actually have less graphic power than the 9918A, had tons of beautiful graphics. Bottomline. The stock TI-99/4A was too limited (no access to machine code).
 
?
 

That's right - no access to machine code, and no access to bitmap - of course, by storing TI BASIC programs in VRAM there was no possibility for bitmap. The whole 4A architecture is just crazy!

 

The ZX Spectrum only has a bitmap screen. There are no text modes, and no sprites. Consequently, ZX Spectrum BASIC has PLOT, LINE, DRAW and CIRCLE commands built right into the language - so bitmap was accessible to everyone. The 8x8 pixel colour clash is quite a limitation though. Still... some beautiful graphics were created. 

[ti99iuc]

@sometimes99er
ah yes, you you reminded me of this collection of images I found on a disk

https://www.ti99iuc.it/web/index.php?pageid=database_cerca&archivioid=394

 

[sometimes99er]

15 minutes ago, Willsy said:

That's right - no access to machine code, and no access to bitmap - of course, by storing TI BASIC programs in VRAM there was no possibility for bitmap. The whole 4A architecture is just crazy!

Yes. Too many talented people simply moved on - to other more capable home computers.
 
?
 

Edited February 9, 2022 by sometimes99er

[sometimes99er]

6 minutes ago, ti99iuc said:

@sometimes99er
ah yes, you you reminded me of this collection of images I found on a disk

https://www.ti99iuc.it/web/index.php?pageid=database_cerca&archivioid=394

 

Very nice. You should state, apart from XB, if 32K Expansion Memory is required.
 
?
 
Edit. Oh you did - in the titles. What is LM routines ? ?
 

Edited February 9, 2022 by sometimes99er

[sometimes99er]

16 minutes ago, Willsy said:

That's right - no access to machine code, and no access to bitmap - of course, by storing TI BASIC programs in VRAM there was no possibility for bitmap. The whole 4A architecture is just crazy!

 

The ZX Spectrum only has a bitmap screen. There are no text modes, and no sprites. Consequently, ZX Spectrum BASIC has PLOT, LINE, DRAW and CIRCLE commands built right into the language - so bitmap was accessible to everyone. The 8x8 pixel colour clash is quite a limitation though. Still... some beautiful graphics were created. 

And the stock ZX Spectrum had access to machine code. Buy a book and you're flying ...
 
?
 

[oddemann]

15 minutes ago, sometimes99er said:

Yes. Too many talented people simply move on - to other more capable home computers.
 
?
 

How hard/easy would it be to take the TI and fix that?

I guess that a LOT of stuff would change. Basic would not be Basic and so on. I guess there would have to be a change of all the software. Would it be interesting to brainstorm this, only in theory, to imagine what COULD be. I am very curious about what would have been possible to do, at that time to make it a better computer. Maybe a interpreter could be made to translate everting into the "imaginary new TI" hardware.

I guess there could be many different ver. of a "dream team TI". Dependent on what you would like to do. How difficult would it to make a tread and imagine the "right" TI. I would have loved to see what the smart ones would imagine. Not to make it, but more as a "what could have been". I would love to see what the different minds would imagine up.

I am so impresses by the knowledge I see here, be it software or hardware. I think that this is one of the strong point of this group, A LOT OF Einstein's live her ? hehehehe I get to be a little smarter every day by being "infected" by your greatness hehehe

Edited February 9, 2022 by oddemann
... Bugs daaaa!

[ti99iuc]

9 minutes ago, sometimes99er said:

Very nice. You should state, apart from XB, if 32K Expansion Memory is required
 

ok, I'll try to do more attention on it. Often I use classic99 to test software and doing screenshots and I haven't found a fast way to disactivate/activate the 32K on it.
Have you some suggest?

[wierd_w]

3 hours ago, oddemann said:

How hard/easy would it be to take the TI and fix that?

I guess that a LOT of stuff would change. Basic would not be Basic and so on. I guess there would have to be a change of all the software. Would it be interesting to brainstorm this, only in theory, to imagine what COULD be. I am very curious about what would have been possible to do, at that time to make it a better computer. Maybe a interpreter could be made to translate everting into the "imaginary new TI" hardware.

I guess there could be many different ver. of a "dream team TI". Dependent on what you would like to do. How difficult would it to make a tread and imagine the "right" TI. I would have loved to see what the smart ones would imagine. Not to make it, but more as a "what could have been". I would love to see what the different minds would imagine up.

I am so impresses by the knowledge I see here, be it software or hardware. I think that this is one of the strong point of this group, A LOT OF Einstein's live her ? hehehehe I get to be a little smarter every day by being "infected" by your greatness hehehe

 

Option to load binary data (machine code) from the internal basic, without restriction. (Similar to commodore basic allowed)

Raw control of the VDP from basic (such as with POKE)
Ability to bus-master the memory/data bus from both the cartridge and sideport (*)

Sound OUT mapped in the cartridge port (+)

16bits wide data bus

VDP ram addressable, but separate from, main RAM

A scratchpad that was big enough for multiple stacks.

 

The TI would have really come alive with those changes.

 

 

* This would have allowed the cartridge port to have ancillary processors sitting on it, for instance.  Need a sidescroll in hardware? Implement the bit-shift magic in silicon, and have the cartridge port manipulate VDP memory using an internal CPU/engine for that.  Want a math coprocessor? Memory map IO that in the cartridge port, and give it a service routine.  See for instance, all the amazing things on the NES with mapper chips, and on the SNES with the likes of the SuperFX chip.)

 

+ Custom DSPs could have lived on a cartridge this way. A simple bit of exposing some wires would have allowed commercial game makers to supply whatever hardware (that TI could have sold to them even!) to generate additional sounds they wanted, and route it out the sound out line.

 

 

This arrangement would make purchasing a specific game in cartridge form, (assuming it has a useful coprocessor on it), also be an investment for other kinds of software loaded from disk, as the disk based software could also leverage that processor, even if you did not start the game in the cartridge- It would just need to be present.  Likewise with DSPs being present.   Since VDP memory and main memory would NOT be the same thing, a coprocessor to manage the VDP's memory with bulk data writes (or better yet, being able to define the address range the VDP looks for, and pointing it at RAM owned by the cartridge so that bank switching can happen quickly, with the coprocessor being able to scribble on non-mapped pages) would have allowed some pretty fancy tricks with the VDP.  Faux Bitmap mode, for instance.  (the coprocessor reads the flat bitmap you have in main memory, and does the unmapped data manipulation and paging functions to get the VDP to generate that image, all for you-- without involving the main CPU, leaving it free to be used for the game, etc.)

 

The base machine would have stayed pretty similar in spec and complexity as the current one, but would have had much greater potential for fancy add-ons that would have changed the game considerably.

Edited February 9, 2022 by wierd_w

[+mizapf]

5 hours ago, Willsy said:

That's right - no access to machine code, and no access to bitmap - of course, by storing TI BASIC programs in VRAM there was no possibility for bitmap. The whole 4A architecture is just crazy!

 

The ZX Spectrum only has a bitmap screen. There are no text modes, and no sprites. Consequently, ZX Spectrum BASIC has PLOT, LINE, DRAW and CIRCLE commands built right into the language - so bitmap was accessible to everyone. The 8x8 pixel colour clash is quite a limitation though. Still... some beautiful graphics were created. 

 

I think the TI-99/4A really smells like a bag of management decisions.

[GDMike]

I mean, unless TI had something planned that was going to address their memory and maybe video upgrade issues. But I'm not sure what that would look like.

Edited February 9, 2022 by GDMike

[Casey]

The 99/8 kind of addressed it right?  It had CPU RAM separate from VDP RAM.  TI Extended BASIC II had access to machine language.  I'm not sure sticking with the 9918 (9118) was the best choice when they could have gone with a better VDP by then.

[+mizapf]

Due to my MAME involvement, I learnt about the 99/8 in great detail, and got to the conclusion that it was probably best not to try to bring it to market. Virtually everything in the 99/8 was (later) better solved by the Geneve.

[Tuxon86]

The only way that Ti could've been a serious contender is if they would've went with another CPU/architecture. If they hadn't declared war on Jack Tramiel, they could've used a CMOS 65xx or they could've gone the Z80 or Motorola 68xx route. But know, Ti was greedy... They wanted total control but they surevaluate their potential to compete in the micro market.