Why did Atari ditch the 5200?

[atariksi]

 

But SIO port has the COMMAND/MOTOR/etc. which require a PIA would digital joysticks for complete compatibility.

 

If we're only making a few small changes to the 400 to emphasize it as a game console then there is no reason to leave the PIA out. Especially since it is supposed to be both a console first and an entry level A8 second.

 

If Atari had done this I do forsee one unhappy side effect. This machine would have had the highest installed base likely and would have been the development house target. Even less software would have been written for higher memory configurations of the A8 chipset.

 

It could also have had a good effect-- people would have written games for 16K RAM and have better version of same games for computers using 48K RAM. Of course, the 32K ROM of A5200 + 16K RAM = 48K total memory.

 

I think the PIA was purposely being left out to reduce cost of unit.

[opcode]

Kind of off-topic, but here is something I created a couple of years ago, the Atari 1600. Fully Atari 800 compatible but with 16 bits CPU, 4MB of RAM, new faster bus, expansion slots, RTC and more. The Antic/GTIA was going to use a separate bus, so that the CPU clock would be increased freely, with no cycle stealing. It took me a while to figure out everything, but I thought my solution was pretty good for a fairly serious machine.

[Curt Vendel]

Sure - having something like the 7800 would've been nice, but that would require the layout and design of the new chip - the MARIA... Again, we are talking about what would've been the best way for Atari to have gotten the 5200 design out the door...

 

To bring up the 7800 like that, you might as well say - hey Sega should've had the Genesis in 1986 instead of the SMS...

 

 

Gotta stay on topic, stay focused, discuss the 5200 platform specific - we are throwing way too many "other" platforms into the mix, plus now we are skewing into timetravel fantasties of bringing a non-existent console (7800) back in time to the 1981-1982 period.

 

 

 

Curt

 

Granted his post was pretty biassed (and downright trollish), but one thing I think should at least be mentioned was that with the "all new" console proposed, he was talking particularly in terms of video games, not saying Atari Shouldn't continue with their 8-bit computer line. And in this sense I can agree to some point, a cost optimized console aimed specifically as a game console, designed with more recent advances and with Atari 2600 compatibility in mind would have been ideal.

 

At least someone got what I am trying to mean, I am not against the Atari 800, in fact I like the machine a lot, have big collection, etc. I just think it wasn't a good idea to repurpose it as a video game system. And I don't think the problem was the joystick either...

wgungfu was saying the other day that Atari had 80% of the market, and when they went down they took the whole industry with them (which to be honest was something I had never considered, but that makes perfect sense). So I was wondering that if Atari had released the right successor for the 2600 things would have been a lot different.

 

OK, sorry about the misenterpretation then.

 

Something like the 7800 in place of the 5200 really would have been ideal, a much more streamlined, simpler, cheaper, and more compact design that was also (arguably to some degree) technically superior as well. (although something like that may have been a bit different if development had started in '81/82 rather than '83 like the 7800) On top of that, design it around the principal of having backwards compatibility with minimal cost, and preferably taking some advantage of the older hardware. (iirc the 7800 still used TIA and RIOT to read the joysticks/fire buttons, and TIA's sound of course -though they probably have put a POKEY on the board and been done with it rather than fidling around with on-cart audio expansion -and could still had the added benefit of TIA's sound)

 

The other option for a 2600 successor rather than the 7800 configuration with an entirely new video chip and the 2600 hardware just for compatibility, they could have built upon the 2600 hardware itsself, add more sprites, enhanced video modes (higher resolution and/or color), add an Antic-like chip to drive the video rather than monopolizing CPU resourse, add more RAM, use a full 6502 (or Atari's custom 6502C) and either add a couple more sound channels as well or add a POKEY to the board. Have the updated hardware (RIOT+TIA enhanced) integrated into a single chip, just the CPU separate and POKEY if added audio becomes impractical.

Not as advanced as the 7800, but probably enough to at least compete on par with the ColecoVision or Atari computers, and at much lower cost than the 5200 or such. In fact, I think one of the early ideas GCC considered was a more basic upgrade to TIA with more sprites, but they decided to go for much more with MARIA based on their "SPRING" computer idea. (then again that was in 1983, not 81)

In the end, enhancing TIA to CTIA/GTIA+ANTIC capabilities might have been more work than worth it, and keeping a stock TIA+RIOT for compatibility and a new video chip could be a better route all arround. (and maybe still design an ASIC with all 3 built in to save cost)

 

I don't know much about the SARAH design, but it might have been something along one of those lines, apears to have used 2700ish joysticks in the mock-up (but with 2 buttons like the 7800), but it fell through (due to odd CPU architecture or some such), hence the 5200 as the quick alternative. (with curt's proposal being an ideal incarnation of that rather than the mess the 5200 turned out as)

[carmel_andrews]

Perhaps Atari, in their infinite wisdom should have reversed the earlier decision and do a PAL version, I am sure it would have knocked the socks of whatever non atarti console available in PAL land that was available

 

There again, i think AA member mellis intimated that Tramiel breifly ressurected the 5200 (only to get shot of unsold inventory) perhaps if tramiel was hoping to make some money (for his ST) he should have thought about doing a (PAL)European version...seeming as though tramiel was more into pushing atari in europe then it's home market

[Chickybaby]

Just wanted to post a friendly official reminder that it's been a good thread for the most part so far and disagreeing with someone is fine but personal attacks aren't. Thanks and have a good one!

 

- It's been a technically interesting thread and I hope we can keep it that way!

[Curt Vendel]

This is the way it should be Chicky.... having good conversation, putting a lot of idea's, views and possibilities on the table and keeping things on topic and not allowing things to break down into any petty or personal.

 

If we can't come and have a great conversation on such things, then there is no point of forums.

 

 

 

Curt

 

Just wanted to post a friendly official reminder that it's been a good thread for the most part so far and disagreeing with someone is fine but personal attacks aren't. Thanks and have a good one!

 

- It's been a technically interesting thread and I hope we can keep it that way!

[Retro Rogue]

This is the way it should be Chicky.... having good conversation, putting a lot of idea's, views and possibilities on the table and keeping things on topic and not allowing things to break down into any petty or personal.

 

If we can't come and have a great conversation on such things, then there is no point of forums.

 

 

 

Curt

 

 

*que Obama speech applause*

[carmel_andrews]

Oops....slippy finger syndrome...meant to type atari and briefly (must have been late)

 

Also I don't see anyone arguing chicky or any bad mouthing....perhaps your seeing something i am not (or perhaps i am missing something)

 

 

 

 

 

Just wanted to post a friendly official reminder that it's been a good thread for the most part so far and disagreeing with someone is fine but personal attacks aren't. Thanks and have a good one!

 

- It's been a technically interesting thread and I hope we can keep it that way!

[Curt Vendel]

Actually - *que Tea Party speech applause*...

 

 

This is the way it should be Chicky.... having good conversation, putting a lot of idea's, views and possibilities on the table and keeping things on topic and not allowing things to break down into any petty or personal.

 

If we can't come and have a great conversation on such things, then there is no point of forums.

 

 

 

Curt

 

 

*que Obama speech applause*

[Curt Vendel]

Hi Carmel,

 

Nope, that's the point - Chicky was just pointing out that its nice to see such a long running thread without it degrading...

 

I enjoy the idea's of an alternate timeline and alternate solution - the key is everyone has to keep from pushing fantasies of technology that doesn't exist, or radically going off course. We have to look at the situation and see the big picture of resources, time, budget, approval and how Atari could've possibly have made a different version of the 5200 that would'nt have given them a black eye, and used up so much time and resources to design, made it easier to code for it rapidly, and deliver different controllers and a different pack-in game.

 

 

Curt

 

 

Oops....slippy finger syndrome...meant to type atari and briefly (must have been late)

 

Also I don't see anyone arguing chicky or any bad mouthing....perhaps your seeing something i am not (or perhaps i am missing something)

 

 

 

 

 

Just wanted to post a friendly official reminder that it's been a good thread for the most part so far and disagreeing with someone is fine but personal attacks aren't. Thanks and have a good one!

 

- It's been a technically interesting thread and I hope we can keep it that way!

[Retro Rogue]

*que teabagger sounds*

 

Actually - *que Tea Party speech applause*...

 

 

This is the way it should be Chicky.... having good conversation, putting a lot of idea's, views and possibilities on the table and keeping things on topic and not allowing things to break down into any petty or personal.

 

If we can't come and have a great conversation on such things, then there is no point of forums.

 

 

 

Curt

 

 

*que Obama speech applause*

Edited November 4, 2009 by wgungfu

[Curt Vendel]

You have a seriously dirty mind!!! :D

 

 

 

 

Curt

 

 

*que teabagger sounds*

 

Actually - *que Tea Party speech applause*...

 

 

This is the way it should be Chicky.... having good conversation, putting a lot of idea's, views and possibilities on the table and keeping things on topic and not allowing things to break down into any petty or personal.

 

If we can't come and have a great conversation on such things, then there is no point of forums.

 

 

 

Curt

 

 

*que Obama speech applause*

[Chickybaby]

Cue *Moderator Applause* for no thread degredation the last few posts.

 

Thanks Curt for at least someone gettimg my point and not questioning -*Cues Cartman* Respect my Authority!

(It really does get annoying after a bit.)

[kool kitty89]

I don't know kskunk nor agree with him. Business equipment at the time went for special monitors whereas the home computers targetted TVs so 80-column based hardware wasn't needed for A800/A400. Commodore made mistakes in making so many incompatible machines just to make money in different areas. Atari tried to do the same with A5200 although fortunately it was mostly compatible with A800/A400. Atari did make a mistake with A7800 just buying it from someone and sacrificing compatibility again-- following Commodore's mistaken ideas. They should have spent some time integrating the newer 7800 features into existing machines.

 

Hmm, again I'm thinking of a previous discussion with Kskunk mentioning how th ememory model of th e8-bitters (and 650x in general) would be pretty limiting on later designs: http://www.atariage.com/forums/topic/147811-what-if-atari-8bit/page__p__1800968entry1800968

Sorry, I didn't follow that thread and now given some quotes you have given of kskunk, all I have to say is he is just speculating. His statement like:

"Maybe that's why the C64 comes up. The C64 was a great value, because in those days Commodore was totally organized around selling advanced technology cheaply."

 

This is utter rubbish. Any proof or just some fanboyish remark or perhaps he was mislead by some Commodore fanatic. Let me see 1.79Mhz was less advanced than 1.0Mhz, doing bullcrap indeterministic cycles for the color RAM/sprites is more advanced than having exact cycle-to-cycle behavior, having less colors is more advanced than having 256. The guy should take a class in computer technology before he speaks. Atari has more hardware support for most aspects of graphics and sound than Commodore machines. Commodore machines outsold A8 because they were a cheap piece of a crap compared to Atari 800. The only two hardware benefits Commodore had over A8 with color RAM and more sprites had disadvantages of eating up consecutive DMA cycles (causing higher latencies) and making cycle-exact coding inexact (bad for kernels). As I said, if 6502-based processors won the processor wars, you would have 64-bit 6502s instead of 80x86 architecture being supported in modern 64-bit processors.

 

You should listen to people who can prove stuff not just people who speculate their brains out.

 

Well he did say he liked the 6502 persoanlly, but just didn't think such a platform could have grown into something along the linse of PC/MAC/ST/Amiga etc.

 

And the point about commodore didn't sound to me like he was saying the C64 waqs vastly superior to the A8-bit, just that Atari could not compete directly with Commodore in that playfield (low cost-console priced computers) due to Commodore's vertical integration, hence Atari needed an actual video game console to beat such prices. (much narrower hardware profit, with higher software profit plus licence royaltees -the latter necesitating a lockout mechanism to avoid the problem with 2600)

BTW, from what I know Kskunk is an engineer (with experience on both the hardware and software side of things, some of which in specific context), he's a bit more active in the Jaguar forums I think. But in any case, its more than just idle speculation. (honestly that's more on my part as I only have a more general understanding of this stuff, I'm 20 and taking my first programming class this semester -I really love all this history though, and speculative hypotheticals -what ifs- especially)

from Kskunk (2nd page of that thread)

I won't argue that the hardware was primitive, but both the Mac and PC had CPUs with far more potential than a 6502. I don't mean to slam the 6502. It has a special place in my heart, which is why I'm here. And it's still fun to program, but I'm a special kind of nostalgic nut.

 

I also moved on to x86s and 68Ks in the 80s with the rest of the programming world, and so I'm pretty well-versed in all the comparative limitations of those old architectures. When I have to get real work done, it's great to have a CPU with lots of general purpose registers, relocatable code, and real relative addressing (without quirky zero page footwork). The 6502/816 weren't in the same league.

 

 

 

This architecture stuff is off topic though, if it continues, maybe get its own topic as a separate discussion. (it would have worked in that 8-bit thread had it not been locked of course)

 

Doesn't seem like misinterpretation. Re-read his original remarks-- he's claiming that we should completely replace the A800/A400 design (i.e., trash it). He's employing Chewbacca defense now. It's better to make new game system that uses that design and enhances it. A7800 should be A5200 compatible and so on.

 

I don't see hate for the 400/800 though, just a suggestion that, ideally Atari wouldn't have to convert that into a 2600 successor, and that there were more effecient options. (not that they should cancel their computer line or anything)

 

As for compatibility, that's a dual edged sword, continuing to provide compatible, but enhanced hardware either limits the capabilities, adds significantly to cost, or both. (best when a new design is built specifically around compatibility and highly optimized, but still hindered somewhat by that usually, worst case is having mostly useless old hardware strapped on just for compatibility) In computers and such this makes more sense in terms of standardiszation and different cost concerns, but with games systems, their tight cost margins, and often propritary nature, this is different.

Only a few of the more successful consoles were designed as such, and even fuwer do it seamlessly. (those being Wii, GBC/GBA/DS for GBA, and PS2 -with small compatibility issues for some games; PS3 is kind of iffy, though early models were relatively compatibile with PS2, and all models PS1 compatible) Sega seemed to put an emphesis on this even though their consoles were not especially successful. (direct compatibility of SG-1000 with MkIII/SMS, and hardware SMS compatibility with MegaDrive/Genesis with a passthrough adaptor) Nintendo didn't do this with the SNES, N64, or GameCube, and had no direct problems with that. (the N64 and morseo GC had declining popularity, but this had nothing to do with lack of compatibility, likewise the DVD capabilities of the PS2 were a much bigger factor than PS1 compatibility -though still notable due to the massive PS1 library and popularity)

 

I think a natively 2600 compatible successor would have been practical as such (optimized in a similar manner the 7800 was, granted that wasn't perfect), but going from there might start to get messy depending on how cheap you can encorporate the old hardware and how expandable it can be made. (PS3 is an example of such practical problems with attempting to do so through 2 generations)

 

Sure - having something like the 7800 would've been nice, but that would require the layout and design of the new chip - the MARIA... Again, we are talking about what would've been the best way for Atari to have gotten the 5200 design out the door...

 

To bring up the 7800 like that, you might as well say - hey Sega should've had the Genesis in 1986 instead of the SMS...

 

I understand that, I wasn't saying they could ahve had the 7800 as it was by '81/82, just that going that rout (simplified, low cost, natively compatible system with security/lockout) woud be a more ideal route. And again, I don't know any specifics about SARAH. (maybe you do) But that at least resembles such a system in some respects, and at the same time it's failure is what lead to the rush for an alternative from available hardware (A8-bit) and the resulting 5200. (obvously exacerbated by internal management problems by Atari, otherwise something like the system you suggest may have emerged) In any case, putting TIA+RIOT on the board with the 8-bit hardware wouln't be a good option for compatibility. (at best, an add-on adaptor with built-in 2600 hardware)

 

I was mentioning such a system possibly in the context of a more workable alternative to SARAH in the first place. (so not the same time constraints as the 5200)

 

Anyway, a 1981/82 "7800-ish" design (not in architecture, but terms of simplicity and efficiency) is what opcode suggested here:

Ok, let me do a simple exercise on imagination...

I don't remember from the top of my head the memory bandwidth of a Antic/GTIA, but lets take some competitor stuff as an example here. The TMS9918, created in 1979 by TI for their home computer. The TMS9918 runs at 10MHz and has a memory bandwidth of 171 bytes/scanline, and if TI could do that in 79, Atari could do the same in 81/82 (Coleco was using it in 1982 anyway, and the CV cost $100 less than the 5200). Now let’s think of a very arcade-ish architecture. First, 256 pixels per scanline. Tile mode, 8x8 pixels tiles, 2 bits depth, exactly like most arcade games from the time. So we have 32 tiles per row, 64 bytes per scanline of pattern data (we still have 107 memory accesses). Let’s suppose our video chip supports 256 different tiles (8 bits), which isn't bad at all for the 81/82 standards. That means the chip needs to read 32 bytes to map the tiles to the screen. Since each tile can have 4 colors, we need 2 bits per tile to indicate the palette #, or 8 bytes per scanline. So 40 byes for the name table. So now we have 67 memory accesses still left.

Ok, we still need sprites. Sprites have pattern and attributes. The TMS9918 reads both data from VRAM, but in order to save some bandwidth we are going to place the sprite attribute table inside the video chip (and after all we are saving in video modes, the TMS has 4, our video chip has just one). Let’s suppose we want 64 sprites on screen, 8x16 pixels per sprite, 256 different patterns. Supposing 16 sprites per scanline, we would need 32 memory accesses for patterns, which leave us with 35 memory accesses. That can be used to VRAM access by the CPU (to load stuff to VRAM for example, that is much bigger window than the TMS has). So now how much video memory we would need? 4KB for tile pattern, 4KB for sprite pattern, 1KB for tile name, 256 bytes for tile color. If we really want to save on video memory we could cut 1.25KB from the sprite pattern table and allow sprites to use patterns from both sprite and tile pattern tables. With that we would need just 8KB of VRAM, half of the memory the TMS9918 uses, but with graphics that are much more appropriated for games. Inside the video chip we would have the entire sprite attribute table (256 bytes) plus palette tables (64 bytes for 16 groups of 4 colors). And then you add all the other features, hardware scroll, etc, etc.

So how about that? Now you match that with a 6502, Pokey and 2/4KB of RAM and voila, arcade quality video game, totally feasible in 1981/1982, probably not more expensive than a 5200. And something the biggest video game company of its time could have done.

 

Though that might seem a tad ambitious. (well at least if one wanted to keep 2600 compatibility as well, RIOT and TIA would still be useful to some degree for sound and I/O, inless you arranged it for POKEY to do a lot of that -analog joysticks plus buttons interpreted as key inputs) Although, with something as capable as that 2600 compatibility might not have been important. (outside of an adaptor to match coleco's terms)

Edited November 4, 2009 by kool kitty89

[atarian63]

*que teabagger sounds*

 

Actually - *que Tea Party speech applause*...

 

 

This is the way it should be Chicky.... having good conversation, putting a lot of idea's, views and possibilities on the table and keeping things on topic and not allowing things to break down into any petty or personal.

 

If we can't come and have a great conversation on such things, then there is no point of forums.

 

 

 

Curt

 

 

*que Obama speech applause*

 

*que Obama bs speech applause*

[atariksi]

I don't know kskunk nor agree with him. Business equipment at the time went for special monitors whereas the home computers targetted TVs so 80-column based hardware wasn't needed for A800/A400. Commodore made mistakes in making so many incompatible machines just to make money in different areas. Atari tried to do the same with A5200 although fortunately it was mostly compatible with A800/A400. Atari did make a mistake with A7800 just buying it from someone and sacrificing compatibility again-- following Commodore's mistaken ideas. They should have spent some time integrating the newer 7800 features into existing machines.

 

Hmm, again I'm thinking of a previous discussion with Kskunk mentioning how th ememory model of th e8-bitters (and 650x in general) would be pretty limiting on later designs: http://www.atariage.com/forums/topic/147811-what-if-atari-8bit/page__p__1800968entry1800968

Sorry, I didn't follow that thread and now given some quotes you have given of kskunk, all I have to say is he is just speculating. His statement like:

"Maybe that's why the C64 comes up. The C64 was a great value, because in those days Commodore was totally organized around selling advanced technology cheaply."

 

This is utter rubbish. Any proof or just some fanboyish remark or perhaps he was mislead by some Commodore fanatic. Let me see 1.79Mhz was less advanced than 1.0Mhz, doing bullcrap indeterministic cycles for the color RAM/sprites is more advanced than having exact cycle-to-cycle behavior, having less colors is more advanced than having 256. The guy should take a class in computer technology before he speaks. Atari has more hardware support for most aspects of graphics and sound than Commodore machines. Commodore machines outsold A8 because they were a cheap piece of a crap compared to Atari 800. The only two hardware benefits Commodore had over A8 with color RAM and more sprites had disadvantages of eating up consecutive DMA cycles (causing higher latencies) and making cycle-exact coding inexact (bad for kernels). As I said, if 6502-based processors won the processor wars, you would have 64-bit 6502s instead of 80x86 architecture being supported in modern 64-bit processors.

 

You should listen to people who can prove stuff not just people who speculate their brains out.

 

Well he did say he liked the 6502 persoanlly, but just didn't think such a platform could have grown into something along the linse of PC/MAC/ST/Amiga etc.

...

It's speculative. I don't think the 6502 developers had that much funding to go the direction intel/motorola went and as fast. But MAC/ST/Amiga were later. At time of 6502 machines you basically had 6800 and 8088 which aren't that much better with some trade-offs.

 

...of which in specific context), he's a bit more active in the Jaguar forums I think. But in any case, its more than just idle speculation. (honestly that's more on my part as I only have a more general understanding of this stuff, I'm 20 and taking my first programming class this semester -I really love all this history though, and speculative hypotheticals -what ifs- especially)

I don't see any proof where A8 technology is less advanced than C64; it's the opposite for most of its hardware.

 

I don't see hate for the 400/800 though, just a suggestion that, ideally Atari wouldn't have to convert that into a 2600 successor, and that there were more effecient options. (not that they should cancel their computer line or anything)

...

I didn't say about cancelling the computer line. The technology of the computers was advanced at the time the A5200 was developed so it was correct to use it rather than trash it. It's more than one place where the hatred was expressed.

 

As for compatibility, that's a dual edged sword, continuing to provide compatible, but enhanced hardware either limits the capabilities, adds significantly to cost, or both. (best when a new design is built specifically around compatibility and highly optimized, but still hindered somewhat by that usually, worst case is having mostly useless old hardware strapped on just for compatibility) In computers and such this makes more sense in terms of standardiszation and different cost concerns, but with games systems, their tight cost margins, and often propritary nature, this is different....

Enhanced hardware does not limit the capabilities as the chip's transistor density increased gradually. How is the VGA limited that it supports older CGA/EGA capabilities? It's not. It may add some time initially to keep in tact but the cost of production of the new chips is negligible and you get the advantage of using existing software base and offering it as an upgrade. Of course, if you want to make more money and think you can get new customers, you want people to rewrite the software from scratch-- bad for consumer but good for company's money making. I think with A2600 the market was different from the computer buyers and it was new market so the compatibility didn't play that big of a role at that time but after the customer base was established then compatibility becomes more significant.

[opcode]

 

Anyway, a 1981/82 "7800-ish" design (not in architecture, but terms of simplicity and efficiency) is what opcode suggested here:

Ok, let me do a simple exercise on imagination...

I don't remember from the top of my head the memory bandwidth of a Antic/GTIA, but lets take some competitor stuff as an example here. The TMS9918, created in 1979 by TI for their home computer. The TMS9918 runs at 10MHz and has a memory bandwidth of 171 bytes/scanline, and if TI could do that in 79, Atari could do the same in 81/82 (Coleco was using it in 1982 anyway, and the CV cost $100 less than the 5200). Now let’s think of a very arcade-ish architecture. First, 256 pixels per scanline. Tile mode, 8x8 pixels tiles, 2 bits depth, exactly like most arcade games from the time. So we have 32 tiles per row, 64 bytes per scanline of pattern data (we still have 107 memory accesses). Let’s suppose our video chip supports 256 different tiles (8 bits), which isn't bad at all for the 81/82 standards. That means the chip needs to read 32 bytes to map the tiles to the screen. Since each tile can have 4 colors, we need 2 bits per tile to indicate the palette #, or 8 bytes per scanline. So 40 byes for the name table. So now we have 67 memory accesses still left.

Ok, we still need sprites. Sprites have pattern and attributes. The TMS9918 reads both data from VRAM, but in order to save some bandwidth we are going to place the sprite attribute table inside the video chip (and after all we are saving in video modes, the TMS has 4, our video chip has just one). Let’s suppose we want 64 sprites on screen, 8x16 pixels per sprite, 256 different patterns. Supposing 16 sprites per scanline, we would need 32 memory accesses for patterns, which leave us with 35 memory accesses. That can be used to VRAM access by the CPU (to load stuff to VRAM for example, that is much bigger window than the TMS has). So now how much video memory we would need? 4KB for tile pattern, 4KB for sprite pattern, 1KB for tile name, 256 bytes for tile color. If we really want to save on video memory we could cut 1.25KB from the sprite pattern table and allow sprites to use patterns from both sprite and tile pattern tables. With that we would need just 8KB of VRAM, half of the memory the TMS9918 uses, but with graphics that are much more appropriated for games. Inside the video chip we would have the entire sprite attribute table (256 bytes) plus palette tables (64 bytes for 16 groups of 4 colors). And then you add all the other features, hardware scroll, etc, etc.

So how about that? Now you match that with a 6502, Pokey and 2/4KB of RAM and voila, arcade quality video game, totally feasible in 1981/1982, probably not more expensive than a 5200. And something the biggest video game company of its time could have done.

 

Though that might seem a tad ambitious.

 

Actually not as ambitious as it may sound at the first. Again, the TMS9918 was created in 1979. The solution I proposed above has the same memory bandwidth, all the same decoding (the TMS9918 also has all the same tables as my solution, I just re-arranged data on each table a bit to make the solution more game oriented), and it requires half of the VRAM the TMS9918 uses. The only radical change I did was to embed the sprite attribute table inside the chip. On the other hand I saved on VRAM and removed the extra video modes the TMS9918 has (text mode, lo-res, and another graphic mode that is a sub-set of the main mode). Oh, and it is important to remember that the TMS9918 was also created with computers in mind (the TI99/4A).

So I was just trying to illustrate my point that something, lets say, more adequate for games could have been done back in 81/82, with technology that was totally realistic for the time. I may I add that such architecture would have been a breeze to program for, very arcade-ish, very appropriated for an arcade company.

[kool kitty89]

Actually not as ambitious as it may sound at the first. Again, the TMS9918 was created in 1979. The solution I proposed above has the same memory bandwidth, all the same decoding (the TMS9918 also has all the same tables as my solution, I just re-arranged data on each table a bit to make the solution more game oriented), and it requires half of the VRAM the TMS9918 uses. The only radical change I did was to embed the sprite attribute table inside the chip. On the other hand I saved on VRAM and removed the extra video modes the TMS9918 has (text mode, lo-res, and another graphic mode that is a sub-set of the main mode). Oh, and it is important to remember that the TMS9918 was also created with computers in mind (the TI99/4A).

So I was just trying to illustrate my point that something, lets say, more adequate for games could have been done back in 81/82, with technology that was totally realistic for the time. I may I add that such architecture would have been a breeze to program for, very arcade-ish, very appropriated for an arcade company.

Well, with that, they could even have turned around and used if as the basis for arcade machines. (Nintendo and Sega both did it with their consoles; in fact Sega continued to do it after Dreamcast, with the GameCube and Xbox hardware)

 

In yor example, were you planning on including native VCS compatibility as well? And if you wanted to further save cost, you coude use a single bus design, like the 2600, 8-bit, and 7800 use, thoguh that could pose bandwidth limitations (CPU and VDP fighting for bus time), that's a trade-off though (and something that could be circumvented in arcade applications). Hmm, if you did things like the A8-bit or TMS9918, you'd use cheap DRAM as well (opposed to SRAM), so maybe a single 16 kB block of shared RAM with the VDP having refresh hardware built-in (like the 9918), and with a single bus, you'd also allow both the CPU and VDP access to caartridge ROM, right? (just as with the 7800, or NES for that matter -though in that case there are multiple busses) In which case, you could have some of the video data in ROM, so maybe you could get away with 8 kB of RAM. (that is, inless you wanted to support a 256x192 4-color bitmap mode, in which case that 16 kB would come in handy)

[atariksi]

Actually not as ambitious as it may sound at the first. Again, the TMS9918 was created in 1979. The solution I proposed above has the same memory bandwidth, all the same decoding (the TMS9918 also has all the same tables as my solution, I just re-arranged data on each table a bit to make the solution more game oriented), and it requires half of the VRAM the TMS9918 uses. The only radical change I did was to embed the sprite attribute table inside the chip. On the other hand I saved on VRAM and removed the extra video modes the TMS9918 has (text mode, lo-res, and another graphic mode that is a sub-set of the main mode). Oh, and it is important to remember that the TMS9918 was also created with computers in mind (the TI99/4A).

So I was just trying to illustrate my point that something, lets say, more adequate for games could have been done back in 81/82, with technology that was totally realistic for the time. I may I add that such architecture would have been a breeze to program for, very arcade-ish, very appropriated for an arcade company.

Well, with that, they could even have turned around and used if as the basis for arcade machines. (Nintendo and Sega both did it with their consoles; in fact Sega continued to do it after Dreamcast, with the GameCube and Xbox hardware)

 

In yor example, were you planning on including native VCS compatibility as well? And if you wanted to further save cost, you coude use a single bus design, like the 2600, 8-bit, and 7800 use, thoguh that could pose bandwidth limitations (CPU and VDP fighting for bus time), that's a trade-off though (and something that could be circumvented in arcade applications). Hmm, if you did things like the A8-bit or TMS9918, you'd use cheap DRAM as well (opposed to SRAM), so maybe a single 16 kB block of shared RAM with the VDP having refresh hardware built-in (like the 9918), and with a single bus, you'd also allow both the CPU and VDP access to caartridge ROM, right? (just as with the 7800, or NES for that matter -though in that case there are multiple busses) In which case, you could have some of the video data in ROM, so maybe you could get away with 8 kB of RAM. (that is, inless you wanted to support a 256x192 4-color bitmap mode, in which case that 16 kB would come in handy)

 

I think the ANTIC method of accessing video memory is much more flexible than other methods being used at the time. And the later 130XE memory banking helped add the video memory flexibility.

[kool kitty89]

It's speculative. I don't think the 6502 developers had that much funding to go the direction intel/motorola went and as fast. But MAC/ST/Amiga were later. At time of 6502 machines you basically had 6800 and 8088 which aren't that much better with some trade-offs.

 

That wasn't the context of that discussion, that was int he context of further development of the 8-bitters in '84/85, by which time the PC was prominent (and clones becoming more numerous), the MAC was out there, and the Amiga and ST were almost there. The point was that none of the 8-bit architectures evolved into competitive later designs. (Motorola had the 6809, but they went ahead with the 68000 which went big, intel didn't use 8080 compatibility in the 8086, Zilog didn't make their Z8000 Z80 compatible either, there was the Z800 several years later, but that was both too late and not particularly successful)

 

Do you think it was a bad idea for Apple to switch from the 68k architecture to PowerPC?

 

I don't see any proof where A8 technology is less advanced than C64; it's the opposite for most of its hardware.

Commodore was in a situation to sell them for much lower prices though, that's the point.

 

I didn't say about cancelling the computer line. The technology of the computers was advanced at the time the A5200 was developed so it was correct to use it rather than trash it. It's more than one place where the hatred was expressed.

 

His point is not that the 8-bit is a weak design, but rather that a cheaper, more capable design could have been engineered specifically for a video game console by that point in time. (the 8-bit hardware, at best, would still be relatively expensive, and too expensive if one wanted to encorporate VCS compatibility) So are you saying compatibility of the 8-bit computers is important, but having such a coneole natively compatible with the VCS is unimportant?

 

Enhanced hardware does not limit the capabilities as the chip's transistor density increased gradually. How is the VGA limited that it supports older CGA/EGA capabilities? It's not. It may add some time initially to keep in tact but the cost of production of the new chips is negligible and you get the advantage of using existing software base and offering it as an upgrade. Of course, if you want to make more money and think you can get new customers, you want people to rewrite the software from scratch-- bad for consumer but good for company's money making. I think with A2600 the market was different from the computer buyers and it was new market so the compatibility didn't play that big of a role at that time but after the customer base was established then compatibility becomes more significant.

 

You still have to put resourses into engineering the disign for comatibility, designing the ASIC to encorporate the older hardware, possibly taking up board space as an additional cip, and at very least using added silicon and development resourses. It depends though, with a good, carefully engineered design, it may be possible to effectivly make something like that. The Master System used the same took the SG-1000 design and heavily built upon the TMS VDP, while retaining the necessary older modes for compatibility, used the same CPU and just added a bit more RAM. With the Genesis, they again, built on to the VDP and kept compatibility (and increaded VRAm to 64 kB), kept the PSG sound (not on board the VDP ASIC), relagated the Z80 to coprocessor status, added a YM2612 for enhanced sound, keeping its own 8 kB and being mainly relegated to audio work, and added a 68000 as the main CPU with 64 kB of work RAM. Now, one could argue (and I have) that instead of the Saturn, a compatible, evolutionary development of the Genesis could have worked. (with soem advantages, and tradeoffs of course)

 

So maybe something liek that could have worked for the 8-bit as well, relagating the 6502 as a coprocessor (maybe even an audio controller), use a new, integrated display processor, compatible with GTIA+ANTIC, and a new CPU. (probably a 68000; the TRS-80 model 16 did something like this too, added a 68k and kept the Z80 for compatibility and I/O)

 

back to this:

See what stupid decision Commodore made by making all those different incompatible machines although later learned from it and made the Amigas backward compatible. And you can see that Atari ST was a new technology and many people stayed with the 8-bit or went to other systems because of incompatibility. Someone's got to have a screw loose to recommend such a methodology after looking at history.

 

Do you really think a big reason the ST didn't sell that well (in the US at least) was because it wasn't compatible with the 8-bit line? (in nay case, that's explainable as Atari Inc was gone by then, Tramiel already had his design in the works, though I suppose they could have somehow managed a compatibile arrangement, it probably would have driven cost into the Amiga range)

 

And do you think Commodore should have had the Amiga compatible with the C64? (part of that would be that, the Amiga was already an established design, so to work in C64 compatibility at that point could have been rather inefficient)

 

I am actually glad they didn't use the PC processors. At that time, you could never tell who would win out on the processor wars. And in hindsight, Tandy's specialized graphics/audio didn't stick around-- it's dead. And Atari was a big gun so why should it try to imitate/copy others. They did some PC clones later on when they were no longer the big guns. If 6502 had won the processor wars -- perhaps we would have a 2Ghz 64-bit 6502 backward compatible with 8-bit 6502. And we all know even to this, there's something unique about A8 systems along with some other retro machines like Amigas.

 

I never responded to part of this before, but I'd meant that comment in the context of Tramiel Atari in particular, with a PC clone rather than the ST. (and maybe going with enhanced EGA compatible graphics rather than simply building on CGA like Tandy did, or just go with encorporating a GUI, maybe the PC version of GEM)

 

I'm not yet knoledgeable about this to really claim anything, but do you really think it would have been practical to further develop the 650x architecture to the extent 68k and x86 was? (and if so, why didn't other companies build on their 8-bit processors like that?)

Edited November 5, 2009 by kool kitty89

[atariksi]

It's speculative. I don't think the 6502 developers had that much funding to go the direction intel/motorola went and as fast. But MAC/ST/Amiga were later. At time of 6502 machines you basically had 6800 and 8088 which aren't that much better with some trade-offs.

 

That wasn't the context of that discussion, that was int he context of further development of the 8-bitters in '84/85, by which time the PC was prominent (and clones becoming more numerous), the MAC was out there, and the Amiga and ST were almost there...

 

Do you think it was a bad idea for Apple to switch from the 68k architecture to PowerPC?

Yep, it's bad to drop customer support and come up with a new design. Of course, the fact that 6502 could not be further devleoped into 16-bit or 32-bit or 64-bit due to lack of research funds contributes to the misfortune. But making incompatible machines is a bad idea always after a user base is established.

 

I don't see any proof where A8 technology is less advanced than C64; it's the opposite for most of its hardware.

Commodore was in a situation to sell them for much lower prices though, that's the point.

...

I don't see the proof that it was ADVANCED at the time. There were architectures better and more advanced. And there were machines at the time that were cheap.

 

His point is not that the 8-bit is a weak design, but rather that a cheaper, more capable design could have been engineered specifically for a video game console by that point in time. (the 8-bit hardware, at best, would still be relatively expensive, and too expensive if one wanted to encorporate VCS compatibility) So are you saying compatibility of the 8-bit computers is important, but having such a coneole natively compatible with the VCS is unimportant?

No, it's always important to have compatibility but I'm saying at the time when computer market was new, it's not as bad as it is later on with a user base established. No, I don't agree that a MORE CAPABLE DESIGN that's CHEAPER was possible at that time and I'm talking time of A5200.

 

Enhanced hardware does not limit the capabilities as the chip's transistor density increased gradually. How is the VGA limited that it supports older CGA/EGA capabilities? It's not. It may add some time initially to keep in tact but the cost of production of the new chips is negligible and you get the advantage of using existing software base and offering it as an upgrade. Of course, if you want to make more money and think you can get new customers, you want people to rewrite the software from scratch-- bad for consumer but good for company's money making. I think with A2600 the market was different from the computer buyers and it was new market so the compatibility didn't play that big of a role at that time but after the customer base was established then compatibility becomes more significant.

 

You still have to put resourses into engineering the disign for comatibility, designing the ASIC to encorporate the older hardware, possibly taking up board space as an additional cip, and at very least using added silicon and development resourses. It depends though, with a good, carefully engineered design, it may be possible to effectivly make something like that....

No, it's easier to enhance existing software if you wrote it than to rewrite from scratch. The problem Atari had was their existing engineers weren't around so they had to start from scratch with Atari ST. If they were around, it would be easier to enhance where needed.

 

So maybe something liek that could have worked for the 8-bit as well, relagating the 6502 as a coprocessor (maybe even an audio controller), use a new, integrated display processor, compatible with GTIA+ANTIC, and a new CPU. (probably a 68000; the TRS-80 model 16 did something like this too, added a 68k and kept the Z80 for compatibility and I/O)

Thanks, that would be 2nd best option. Best would be if 6502 developers had the funds to continue development of the processor.

[Retro Rogue]

The problem Atari had was their existing engineers weren't around so they had to start from scratch with Atari ST. If they were around, it would be easier to enhance where needed.

 

What are you talking about? The RBP was started *before* Jack bought Atari Consumer, and based around the existing ex-Commodore engineers. It was never an issue of "well we don't have existing engineers from Atari Inc., so we have to make due with what we have." In fact they could have hired on some more of the Inc. engineers if they wanted to during the transition, and in fact did hire a few for other areas.

[atariksi]

The problem Atari had was their existing engineers weren't around so they had to start from scratch with Atari ST. If they were around, it would be easier to enhance where needed.

 

What are you talking about? The RBP was started *before* Jack bought Atari Consumer, and based around the existing ex-Commodore engineers. It was never an issue of "well we don't have existing engineers from Atari Inc., so we have to make due with what we have." In fact they could have hired on some more of the Inc. engineers if they wanted to during the transition, and in fact did hire a few for other areas.

 

I am talking about making things backward compatible-- that it would be easier if you have the same engineers around. It's harder for new engineers to come in and continue. I am talking about A2600/A800/A5200 time period. What do you mean by "RBP"?

[Retro Rogue]

I am talking about making things backward compatible-- that it would be easier if you have the same engineers around. It's harder for new engineers to come in and continue. I am talking about A2600/A800/A5200 time period. What do you mean by "RBP"?

 

You made a statement regarding the Atari ST. RBP is the original name of the ST.

[+pboland]

The idea is to keep this system as a GAME CONSOLE - and as an entry level ONLY computer

 

16K maximum (no easy upgrade beyond soldering)

Plug for a keyboard attachment

Move the SIO to the rear and either change it slightly with a custom connector for a custom 5210 (5200 version of the 410 recorder) for Tape use ONLY.

 

It is 100% compatible with all 16K 800 software

 

We are NOT concerned with the XL's because its 1981 - they don't exist.

 

The system is purposely limited - remember this is Atari's marketing making the machine - this is not a wishlist machine from the consumer buyer side. Atari wants the system limited because then it is encourages those who get the taste for owning an entry level computer to then step up to an Atari 800 or our fictional Atari 1200 with 80 columns and 128K of memory...

 

Remember - we are playing retrospective devils advocates here and trying to realistically create what the Atari 5200 should've been from a more sensible approach...

 

 

Curt

I do believe that this kind of 5200 would have been a hit. I would have begged my parents for a system like this back in the day as long as they didn't put pastel colored buttons on it. I always thought the idea behind the XEGS was a few years too late.