Could an Amiga with an 030 do Virtua Fighter?

[Juz400]

Originally Posted by roondar

This gives rise to an interesting notion (well for me anyway )

The various low end Amiga's all have at least one expansion port that can be used to plug an accelerator of sort in. Now, normally these ports were either used for memory or a Motorola 68K processor/memory. But there was really nothing stopping Commodore (or others) from including such a DSP or other high performance specialised chip on the board. We've seen one such thing in the form of the A314 expansion for the A500.

So, if we accept that the Mega Drive can do Virtua Racing with an expensive DSP chip - perhaps we can also accept that the A1200/CD32 can do something similar if it is coupled with such a chip?

It's not really on topic I know and it definitely isn't doable right now (there are no DSP/"foreign CPU" accelerators for the A1200), but I still find it interesting that generally we all consider a game like the MD version of Virtua Racing to be a native MD game - even though it has a massively powerful CPU on board that basically bypasses the MD CPU. But on the other hand, we'd usually not consider a game needing say a 68060 a native A1200 game. To me this is a grey area. Anyway, enough off-topic stuff

Its a real shame that Mr Haynie had two game changing projects canned by CBM namely

Gemini Multi CPU card for ZorroIII
https://bboah.com/index.php?action=a...684&artlang=en

AA3000+ with the AT&T DSP3210
[ Show youtube player ]

The DSP is capable of up to 25 million Flops @50Mhz, with Codecs for Audio and Modem use.
Surely it does not need to be audio data, give this little beastie a list of Float calculations to get on with while 680x0 is busy with game and graphics work

The Gemini project, wow! If that had been finished and sold, Kickstart 3 could have been the first domestic OS with multicpu capablility.
same as the DSP, offload some work to another CPU. say cd32, the 020 does all the screen buffering/C2P and the other does the heavy lifting.

Would the DSP unit have kept the Amiga in the game a little longer if it relished 3D math?
Could we today have Coldfire as a SECONDARY CPU, NOT running Amiga OS but what ever the Programmer wants it to do?
Here Coldy, you get on with this data and send me an poll/interupt when you are done. Oh Well dream on,
We have the Vampire now, I dont think I have seen anyting to make it gasp for air yet.

[trixster]

Just look at what DML has managed to make the Atari Falcon do by leveraging the power of its DSP and FPU

[ Show youtube player ]

[ Show youtube player ]

[Old_Bob]

Originally Posted by Juz400

Its a real shame that Mr Haynie had two game changing projects canned by CBM namely

Gemini Multi CPU card for ZorroIII
https://bboah.com/index.php?action=a...684&artlang=en

AA3000+ with the AT&T DSP3210
[ Show youtube player ]

The DSP is capable of up to 25 million Flops @50Mhz, with Codecs for Audio and Modem use.
Surely it does not need to be audio data, give this little beastie a list of Float calculations to get on with while 680x0 is busy with game and graphics work

The Gemini project, wow! If that had been finished and sold, Kickstart 3 could have been the first domestic OS with multicpu capablility.
same as the DSP, offload some work to another CPU. say cd32, the 020 does all the screen buffering/C2P and the other does the heavy lifting.

Would the DSP unit have kept the Amiga in the game a little longer if it relished 3D math?
Could we today have Coldfire as a SECONDARY CPU, NOT running Amiga OS but what ever the Programmer wants it to do?
Here Coldy, you get on with this data and send me an poll/interupt when you are done. Oh Well dream on,
We have the Vampire now, I dont think I have seen anyting to make it gasp for air yet.

Very interesting, dude. I wasn't aware of that multi 68030 board. At the slight risk of further thread drift, I have occasionally pondered whether the Amiga could have been easily adapted to use multiple CPUs. It seems, as ever, that somebody else already had that idea. *sigh* Another "what if..." on to the the Amiga pile.

Along with, of course, the 3210 DSP. All that extra processing power in our machines certainly wouldn't have hurt, back in 199x.

Originally Posted by trixster

Just look at what DML has managed to make the Atari Falcon do by leveraging the power of its DSP and FPU

[ Show youtube player ]

[ Show youtube player ]

Quite. And this is with, as I understand it, a far less powerful chip.

B

[Juz400]

Originally Posted by Old_Bob

Quite. And this is with, as I understand it, a far less powerful chip.

B

It is quite the eyebrow raiser! Indeed, the Amiga has/had the faster silicon

Begs the question, Could a well programmed Falcon do VF?


[ Show youtube player ]

[Gilbert]

Originally Posted by roondar

Note: I do understand why it wasn't done (the trapdoor/side car slots were not really designed for many repeated insertions/removals), but it was technically definitely possible to make A500/A1200 "cartridges".

Anyway, enough random speculation.

They could have made put a device you plug-in which had a slot *on it* which was designed for repeated insertions. If they had upgraded the A500 -it would have beaten back the consoles maybe. Things like how the Amiga couldn't run Streetfighter 2 very well probably damaged it a lot at that time

I think the 1/2 meg memory expansion showed that people would buy expansions in large numbers if the price wasn't too high. The Half-meg expansion must have been one of the few successful hardware upgrades in history!

[Gilbert]

Originally Posted by AmigaHope

The SVP is completely different from the 020 in the CD32. The SVP is a DSP used for vertex calculations and it does it an order of magnitude faster than the 020/14 can.

How much faster though? I read somewhere the CD32 is 5 times faster than than the Amiga 500. If they rendered in dual playfield with less bitplanes in teh foreground they could have maybe made a good fist of it.

Originally Posted by trixster

The cd32 in stock form is incredibly slow.

Go easy man, the CD32 is great! I don't know much about programming but it could probably stream data in from the CD while the game is playing. You could probably use that to make some amazing things, like pre-rendered 3D backgrounds etc

[Samurai_Crow]

@Gilbert

Both the A1200 and CD32 run at half-speed until you add Fast RAM.

[roondar]

Yup, the CPU in the A1200/CD32 are about 2x the speed of the A500. Until you add some 32 bit fast RAM, then they're closer to 4-5x the speed.

[Gilbert]

Originally Posted by Samurai_Crow

@Gilbert

Both the A1200 and CD32 run at half-speed until you add Fast RAM.

Is that because the blitter shares memory cycles with the processor? or another chip?

Why would they do that? Why didn't they make the CD32 with 1.5 MB chip-Ram and 0.5 Meg fast ram?

[Reynolds]

0.5 meg of fast ram is next to nothing, while having 1.5 megz of chip ram is even more a narrow bottleneck than the 2 megs which is never enough on an average use.

[trixster]

Originally Posted by Gilbert

Go easy man, the CD32 is great! I don't know much about programming but it could probably stream data in from the CD while the game is playing. You could probably use that to make some amazing things, like pre-rendered 3D backgrounds etc

I think you have me wrong, I’m not having a go at the cd32, I have one and think it’s great. But it’s also a very slow and underpowered console that should have been given more power. When it was released towards the end of 1993 it was already very outdated.

[roondar]

Originally Posted by Gilbert

Is that because the blitter shares memory cycles with the processor? or another chip?

To a point, but the most important reason is that the DMA controller in Alice doesn't let the CPU access memory every cycle it can, but rather at a slow rate (essentially limiting the CPU to a maximum bandwidth of 7MB/sec). Fast memory has a better connection to the CPU and allows for much faster memory access (2-4x faster access).

Why would they do that? Why didn't they make the CD32 with 1.5 MB chip-Ram and 0.5 Meg fast ram?

Alice was designed in a hurry and on a budget, so it just isn't as good of a memory controller as it could've been. Now, adding some fast memory would've been very nice, but I'd much rather have seen something like 2MB chip + 0.5MB. Even 2MB + 128KB would've made quite the difference. But, such a design would probably have been (much?) more expensive to design and build.

[robinsonb5]

Originally Posted by Reynolds

0.5 meg of fast ram is next to nothing, while having 1.5 megz of chip ram is even more a narrow bottleneck than the 2 megs which is never enough on an average use.

"Next-to-nothing" can be enough to make a significant difference, though. Obviously more is always better, but even as little as 128k of fast RAM - just enough to get the rendering loops out of chip RAM - would have made a huge difference for games that need the horsepower.

[Samurai_Crow]

Originally Posted by roondar

To a point, but the most important reason is that the DMA controller in Alice doesn't let the CPU access memory every cycle it can, but rather at a slow rate (essentially limiting the CPU to a maximum bandwidth of 7MB/sec). Fast memory has a better connection to the CPU and allows for much faster memory access (2-4x faster access).

Alice was designed in a hurry and on a budget, so it just isn't as good of a memory controller as it could've been. Now, adding some fast memory would've been very nice, but I'd much rather have seen something like 2MB chip + 0.5MB. Even 2MB + 128KB would've made quite the difference. But, such a design would probably have been (much?) more expensive to design and build.

Exactly. If Alice was designed to give page mode bandwidth to the blitter and copper, the copper chunky hack would run at 320x240x4096 colors, negating the need for Akiko.

Speaking of Akiko, it cost so much that if it had been exchanged for fast RAM and an 030, it would have cost the same and had a fast enough CPU that C2P ops would have executed just as fast in software.

[AmigaHope]

Originally Posted by Juz400

It is quite the eyebrow raiser! Indeed, the Amiga has/had the faster silicon

Begs the question, Could a well programmed Falcon do VF?

At a crap framerate, and if you threw enough RAM at it to hold precomputed frames for the fighters, probably yeah it could. Not as good as the 32X but at least in the ballpark. (Maybe half as fast)

The thing really hampering the DSP in the Falcon is keeping it serviced with stuff to do over the Falcon's crippled data bus. DSP56000 load/stores are 24-bits wide, but Falcon data bus is 16-bits wide even though it has a 32-bit 030 in it, and that bus is clocked at half the frequency of the DSP. Moreover the CPU or blitter still has to do the rasterization, over that same garbage bus they you're trying to keep the DSP fed with.

The Falcon030 had some great components (good CPU, okay blitter, great video chip, great DSP) but was let down by its ugly shitty bus that made even the A1200 look good even though the A1200's individual components were worse in almost all respects. (A1200 could do dual-playfield, sprites, HAM8, and had its old copper still, but had a worse CPU, similarish blitter, no real hi-color mode, and absolutely nothing even remotely like the DSP)

tl;dr the Falcon's DSP gives it the oomph to do vertex calculations that an 030 can't, but you still run into the bus limitations that you do on an AGA Amiga, only even worse because the Falcon DSP lives in the Atari equivalent of chip memory land, on a bus around half as fast as AGA, with a video chip chewing through AGA-level amounts of data.

EDIT: Wanted to point out that while, given vertices, the AGA blitter can do solid polygon rasterization by itself, while the Falcon's can't, it's usually faster to just draw horizontal lines rather than use the Amiga blitter's line+fill. Either way the Falcon is screwed due to its crap bus.

[Gilbert]

Originally Posted by trixster

I think you have me wrong, I’m not having a go at the cd32, I have one and think it’s great. But it’s also a very slow and underpowered console that should have been given more power. When it was released towards the end of 1993 it was already very outdated.

That's ok - yes I agree with that - but it has enough CPU power - it just needs more sprites. Look at What Neo Geo does with just sprites and no backgrounds and only a 68000 CPU. They should have given it 64 (16 pixel wide) sprites and a polygon chip. It doesn't need a blitter and all those graphics modes that games never use.

[Gilbert]

Originally Posted by roondar

To a point, but the most important reason is that the DMA controller in Alice doesn't let the CPU access memory every cycle it can, but rather at a slow rate (essentially limiting the CPU to a maximum bandwidth of 7MB/sec). Fast memory has a better connection to the CPU and allows for much faster memory access (2-4x faster access).

Alice was designed in a hurry and on a budget, so it just isn't as good of a memory controller as it could've been. Now, adding some fast memory would've been very nice, but I'd much rather have seen something like 2MB chip + 0.5MB. Even 2MB + 128KB would've made quite the difference. But, such a design would probably have been (much?) more expensive to design and build.

I guess Commodore just rushed AGA out. I do like the CD32 now but at the time I could see that AGA was not much of an upgrade. I suppose leveraging their computer hardware in a console was a low-cost, low-risk move for them. It made business sense. They could rely on Amiga users to buy it too.

I do think if people pushed CD32 to it's limits it could make some good games because the games on there are mostly shovelware.

[eXeler0]

Originally Posted by Samurai_Crow

8><---------

Speaking of Akiko, it cost so much that if it had been exchanged for fast RAM and an 030, it would have cost the same and had a fast enough CPU that C2P ops would have executed just as fast in software.

Let's not forget AKIKO replaced a a couple other chips so in total it was saving cost.. There's no Budgie or Gayle chip on the CD32. Physically, AKIKO looks a lot like Budgie.. Also the 2 CIA chips on the 1200 were replaced with stuff in AKIKO.. And on top of that it Controlled the CD, the 1k EEPROM..
So 3 less chips and more functionality..
A lot of people associate AKIKO with the c2p function which is only a minor, minor thing it also does among a whole bunch of other things.

[roondar]

After that extremely long A1200 thread it's become pretty clear to me Commodore really ought to have put chunky modes into Lisa to begin with. Apparently it's not nearly as complicated to do that as I always thought.

That said, I'm also of the opinion we only recently started seeing the potential of AGA with some of the new stuff coming out today (Rygar, Reshoot-R/Proxima III, Boss Machine, etc). And all those games are looking pretty good TBH

[eXeler0]

Originally Posted by roondar

After that extremely long A1200 thread it's become pretty clear to me Commodore really ought to have put chunky modes into Lisa to begin with. Apparently it's not nearly as complicated to do that as I always thought.

That said, I'm also of the opinion we only recently started seeing the potential of AGA with some of the new stuff coming out today (Rygar, Reshoot-R/Proxima III, Boss Machine, etc). And all those games are looking pretty good TBH

Atari did several of the things discussed (offtopic) here with the Falcon. VIDEL has a 16-bit color Chunky mode, there's a DSP and a 030 CPU.
I'd like to think that an Amiga would have turned out less crippled by poor design than the Falcon, but who knows.. It is what it is..
Now let's try to salvage this thread by going back to the topic :-)