Overview Features Coding ApolloOS Performance Forum Downloads Products Order Contact

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.
Please visit our Apollo-Discord Server for support.



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Documentation about the Vampire hardware

Fantasy Future (I Passed the Time Driving...)

Smartroad 78

Posts 116
30 Oct 2019 18:27


So I was on a long drive today and got to thinking about the Vampire project and it going forwards. This is pure fantasy although something that would make my inner nerd really happy!

I would love to see a future board that was designed more like the original Amiga's. By that I mean, rather than everything integrated into a single chip (or FPGA at the moment), have everything in discrete chips. Mostly I thought CPU, video, audio and some IO (maybe even a GPIO as that is all the rage right now, thanks RPi!). Also, make the chips removable so that the machine could be upgraded when new and better chips came out.

Maybe even having in a form factor more suited to a keyboard case type Amiga (A500/600/1200) because I really like that format. With modern connections it wouldn't be like machines of old with many bulky connections. Just power and DIGITAL-VIDEO at least, maybe a USB port or two at most.

Like I say, this is a fantasy thought exercise for me, helped me pass the time while I was driving!! Even so I would love a future where a new keyboard case Amiga was created!


Kef Emzy

Posts 50
30 Oct 2019 19:50


Integrating everything on a single chip makes everything much cheaper, smaller motherboard, faster communication between custom chips and cpu, and solves a lot of tracing problems (impedance, signal integrity etc).

Nostalgia can be a good thing but technology moves on...


Markus B

Posts 209
30 Oct 2019 20:23


And that's what Commodore and all other companies always did. They reduced costs by merging more and more functions into a single chip.


Mike Kopack

Posts 268
30 Oct 2019 22:11


Yeah, the "trap" commodore and Atari fell into (but not Apple or IBM) was that by using discrete chips for different function, and having to then have busses between those chips, you were always going to be limited in what you could do to upgrade the system.

For example - Ok, let's say you originally have an audio chip that does 2 channel stereo. It might be a chip that uses say 12 pins. Now down the line you develop a 4 or 8 channel chip - you need a LOT more pins to support that - you're stuck. You can't just do a straight drop in replacement because you need more data lines to route the signals between the audio chip and the cpu or whatever else...

THAT is where having everything in a single chip or single FPGA chip is beautiful - If you update one subcomponent within the core, and let's say that needs more bus to talk to the other components, it's all within the single chip - nothing external changes.

Apple (2) and PC's got around this by making everything communicate via the open bus architectures, so if you needed a different video card or audio card or HD controller or whatever, as long as it was designed for that standard bus, it was largely just remove and pop in the new one. That's a LOT harder to do with everything done directly on the MB the way Atari and Commodore did it.



Sean Sk

Posts 488
30 Oct 2019 23:01


I agree with the idea that it's much more sensible to have much of the Amiga system on the FPGA since much of these custom chips and components are able to operate in parallel with one another in an FPGA. Lightning fast parallel processing is one of the awesome things about FPGA's.
 
Having said that I wanted to divert slightly to something off-topic that is NOT Vampire related. Something that Smartroad 78 said made me think of this:
 
There are a few groups in the Amiga community that are doing a bang-up job of reproducing the old Amiga motherboards, giving people the wonderful opportunity to transplant custom chips from old/damaged motherboards to the new one. One such project is the re-Amiga 1200.
 
Using the Amiga 1200 as an example:
 
In the future those custom chips are going to disappear and someone may like to put a newer version of an A1200 motherboard into their A1200 case for preservation purposes. Or they may like to create their own A1200 with one of the new A1200.net cases and an A1200 mechanical keyboard by Kipper2K. I wonder whether it would be possible to reproduce the A1200 motherboard, but using FPGA's in place of where the custom chips would normally go? This would mean not having to rely on sourcing either new old stock of custom chips or ripping them off old boards.
 
I know that at one stage Jens Shoenfeld was doing the A1200 Reloaded. I don't know if that project is dead now, but he was still relying on sourcing the original custom chips.


Mike Kopack

Posts 268
31 Oct 2019 01:10


sean sk wrote:

using FPGA's in place of where the custom chips would normally go? This would mean not having to rely on sourcing either new old stock of custom chips or ripping them off old boards.

Pretty sure I read something in the last week of somebody having done just that for one of the chips (Gary maybe?) and was starting to work on one of the others.

Hardest part will be matching the pinouts. Might have to use some sort of adapter boards to plug into the sockets.


Kresimir Lukin

Posts 65
31 Oct 2019 01:37


Mike Kopack wrote:

 
sean sk wrote:

  using FPGA's in place of where the custom chips would normally go? This would mean not having to rely on sourcing either new old stock of custom chips or ripping them off old boards.
 

 
  Pretty sure I read something in the last week of somebody having done just that for one of the chips (Gary maybe?) and was starting to work on one of the others.
 
  Hardest part will be matching the pinouts. Might have to use some sort of adapter boards to plug into the sockets.
 

 
  there was suska project to recreate Atari ST and as I know they use that aproach at the beggining of project and later they reduced chips by integreting in smaller number of FPGA chips, I think for the moment vampire approach is best, in future if ASIC is made for CPU, then CPU can be ASIC and all other can be on one FPGA.
 


Adam A

Posts 130
31 Oct 2019 02:18


EXTERNAL LINK 
I like the idea of the ultimate 64 elite motherboard, its fpga but still has the form factor and all connections of the original c64  and most importantly you can still use your own SID chips!

dont get me wrong, I still like the idea of the v4 stand alone, it will spread the vampire idea faster and going cost efficient at this stage is more important than doing a high end product, but since this thread is all about fantasy, this is one of the things that id really like to see in future


Eric Gus

Posts 477
31 Oct 2019 06:23


Adam  A wrote:

  I like the idea of the ultimate 64 elite motherboard, its fpga but still has the form factor and all connections of the original c64  and most importantly you can still use your own SID chips!

The c64 is far far simpler a system to recreate than an Amiga, don't get me wrong Gideon and folks have put a great deal of effort into the Ultimate64 (I own one .. its fantastic) .. but the bar is much higher when you are talking about a significantly more complex machine like the Amiga.


Smartroad 78

Posts 116
31 Oct 2019 07:43


@Mike Kopack

"For example - Ok, let's say you originally have an audio chip that does 2 channel stereo. It might be a chip that uses say 12 pins. Now down the line you develop a 4 or 8 channel chip - you need a LOT more pins to support that - you're stuck. You can't just do a straight drop in replacement because you need more data lines to route the signals between the audio chip and the cpu or whatever else... "

Partly I guess that is why many chips have NC pins for future use?

-

Most seem to still be thinking about connecting chips via a parallel bus, which I agree is difficult and complex especially as the bus speed gets faster and latency can affect when a bit will 'arrive'. Not to mention routing that many address and data lines!

However we have things like the PCIe bus, that could be used to connect the CPU to the chip. Even a single PCIe lane would probably have more data bandwidth (PCIe v1.1 is 250MB?) than the Amiga has right now. Not to mention with a PCIe bus chips could be used in other machines (although I am not sure why you would want to, but the possibility exists and could open up more markets).

Also, regarding your specific example of an audio chip maybe needing more pins for other channels. The chip could be on a carrier that connects to the data connection on the board but has it's own headers/connections for the outputs it needs. There is that A1200+ motherboard that is doing that, the video connections and chip ram are on slimline daughter boards, allowing for multiple configurations depending on need.

I had also been thinking that a product like this would move away from FPGA's and enter into ASIC territory, in this fantasy. Before anyone say I know it is HIGHLY unlikely to go ASIC any time soon or ever because of the huge cost ;) It would allow better power efficiency and faster clocks than FPGA could do, but the cost would be prohibitive :D

At the end of the day this was just me having fun with a thought experiment to pass the time while I was driving for 4 hours LOL It's fun to think of the possibilities, look for the plot holes and see if they can be plugged or not ;)


Nixus Minimax

Posts 416
31 Oct 2019 08:42


It is not so much the buses between the chips. Those are a problem with regards to power consumption (it takes significantly more power to drive those comparatively long lines) and data throughput. The problem is price. It is much cheaper to put a single more complex chip into a single housing and use less PCB space than putting two chips with the same combined functionality. This is the driving power for building smaller and smaller transistors.

Commodore was too cheap to invest into the chip production and to my knowledge never did the step from their stoneage NMOS-technology to less power consuming CMOS-technology. By saving the investment on an up-to-date chip technology, they made their products more expensive than they had to be. (Just have a look at the ridiculously large CIA chips, they never changed in more than a decade). The entire Amiga custom chipset could have easily fit on a single chip and consumed much less power if it had been manufactured using a current manufacturing technology at the time the AGA-chipset came out. That is also why the far more complex AAA was supposed to be a single chip and be manufactured by HP (?) and not Commodore themselves (correct me if I'm wrong, I never cared too much about the AAA hype).



Andy Hearn

Posts 374
31 Oct 2019 09:45


i found this as a starting point
  EXTERNAL LINK   
  so as a TL-DR, as of the time of writing, the author recons you're looking at one meeeeeelion dollars as a baseline for a standard Asic to be masked, silicon to be cut and an initial batch run off.
 
  while we're in fantasy land - one of my favourite places...
  i think we'd get a lot of buy in from the current FPGA nay-sayers if could produce the '080 as an actual hard silicon 040/060 socket "drop-in" replacement. forget Saga and H-D-M-I, forget SD and IDE, just go for pure CPU/FPU performance. You'd probably need to bake the usual 128meg ram into the die as well to actually GET that performance, but hey. fantasy right?
 
  then everything from an a4000 to a mac quadra could get a taste of the good stuff.

[edit] looks like there are other options for super low volume orders (sub 100 production run), still priced at $x00's per mm2 though. in this case, most of the cost is the run up to production to make sure the chip is solid. but i'd say the apollo core is quite mature at this stage?

posts 12