Sholud something like this have been the Amiga's gaming future?

[Khephren]

Looking for a new job, I found 'lucky mountain games' and their title 'racing apex'. Something about it makes me think of the Amiga. Maybe something an accelerated A1200 could do. If we had not been so facinated by textured poly's, could we have had something like this?

better screenshots here:  http://toucharcade.com/forums/showthread.php?t=65000

[Khephren]

yeah thats what I thought, pure sega. i'd love to work on something like that. Although it has gourard shading as well, dunno how intensive that is to generate in software. Certainly plenty od demos do it.

[Khephren]

That's exactly what i'm talking about. An evolution of what went before, rather than an attempt to ape the PC.

Zeewolf, no second prize, frontier or cybercon 3. All done with a lot more colours/polys. Maybe with some vertex shading (gourard shading). Old school, but I bet they would look really nice on an 030 with plenty of RAM.

Genetic species is lovely, but everyone switched to trying to do doom, when the old school way of doing it went out the window.

[Khephren]

That game is massively CPU intensive, so unless Commodore had a secret 3D GPU project at half the cost of Orchid/Matrox/PowerVR OR suddenly the price of 50mhz 060 CPUs were half that of 60Mhz Pentium CPUs then never would have happened IMO.

No I don't see the problem. I have a machine here that's 10mips vs A500's .75. A lot of us have them. I can't see  a good reason why my machine can't throw around a lot more untextured polys, with more colours using a c2p routine? Perhaps we won't reach this quality, certainly not this resolution.

Perhaps I misnamed the thread, perhaps i'm just thinking that we could have stuff not a million miles away from this. I was thinking that 'flying high' for example would have been better if they'd concentrated on what the amiga could do, rather than what it wasn't built for.

Although would we have accepted it? I love looking at 'old school' graphics now...but when we were younger, would we have wanted this over something that attempted textures?

[Khephren]

Yep great colours.  It looks like it could have been a hit A7000 game for 1998.
If only that alternate reality was real.

I don't think it's far off what an fast 030 or perhaps an 040 could manage. At 320x256 of course. Maybe drop the vertex shading (which would be a shame).
  Nothing wrong with talking alternate realities, or even giving some game coders future food for thought.
  There's a lot of retro style stuff out there these days, that I think our old girls would be more than capable of giving it a good shot at.

I was hoping some demo coders would chime in and tell us a bit about polygon throughput and gourard shading on expanded amigas, and I wonder if an FPU would be of any help.

[Khephren]

@Amiga_nut

Fatal racing is a bad example. It's fully textured. Texture maps are far harder computationally than either plain shaded polys of gouraud shading. Phong shading is nothing like gourard shading, it's far harder to do as well.

I'm not so much interested in what what came out when, or some hyperthetical A7000 etc. I'm talking about what the limits of our machines actually are. That's why i'm hoping to get some coder input, i'm sure there's some demo coders who know whats what.

yeti3D on a 16.8mhz arm on the GBA manages texture mapping and gouraud. And i'm not even talking about texturemapping.

[Khephren]

I think the closest thing we have to that game on the Amiga is Leading Lap, which I think is not that bad, using the same engine and making it more arcadee and more for aga and high spec amiga's I seen no reason why something like that could'nt be made.

Yeah, up the colours to get rid of the dither patterns, increase the poly count by perhaps a factor of four. Maybe add a shading option for higher end machines, and perhaps options to up the rez. I don't see what would be so out of this world for a fast 030, with plenty of ram.

Leading lap runs ok on an base A1200. Most 030/50's are 8x faster than that, and many have FPU's.

[Khephren]

FPU may not be used, but they SHOULD be used if available. How do you expect to do perspective correction in fixed point? Why do you think so many PC games made use of DX not SX processors on the PC? Do you honestly think FPU's are only good for Ray Tracing?

FPU are used heavily in games on other platforms, the reason they were not on ours, is because they were not a standard feature of the low end, for whom games were generally produced. So most of us have a second processor, that could be used to off load some of the grunt work, doing bugger all most of the time. Again any coder input woud be cool.

Leading lap may look a bit..erm...shit  but it moves at quite a clip. Which I guess was the point.

Zeewolf was cool, short draw distance though. and very few colours. I'm sure with what we now know about c2p, and the systems most of us have, it could be improved on a lot.

[Khephren]

something like elite/frontier for certain. and a nice pixel art rts would be cool- wasn't the bitmaps 'z' originaly meant to be for amiga before they switched platforms?

@amiga_nut
ARM's are lovely chips, but you are over playing their power by a great margin. I should know, i've been developing DS games for years. The archie was a lovely machine, but some times struggled with scrolling because it was cpu bound with no co-pro's.

the bitplane to chunky problem has been some what over come for some time now. novacoders scumm port is in 8bit, nargoth rpg is in ham8. many demos use 8bit or ham8. you mention 32 colours, i wouldn't even say no to that! most amiga sims are 16 and dithered.

zee wolf is nice, but does not really show how much you could throw around.

"if it was possible we would have seen it" not really. all the games companies began to desert the amiga,and those that remained gave us what we wanted- texture mapping and doom clones in the main  

[Khephren]

@Hattig The ARM 2 was  very powerfull chip. Clocking in at 4 MIPS easy, VS the 68k 1MIP. But these are RISC MIPS, ARM2 had a much simpler instruction set (roughly one third the 68k), where you need to combine some instructions to equal one 68k CISC one. So some your throughput on MIPS are combined instructions. Rough comparison might x2 as powerfull. I would not say any more than that.

3D0= 25mhz 030, sounds about right. If we work on the basis of X2 as powerfull per clock as a 68k.
Bare in mind the 3D0 had a DSP, FPU and two video co-processors (unsurprising, as it's from ex Amiga designers).

3D0 ARM 60 (FPU)=.83 DMIPS/MHZ
GBA ARM 7DTMI (NO FPU)= .89 DMIPS/MHZ a 40mhz 030, at best then -without a maths unit.

GBA screen= 76.8% of NTSC low resolution, 60% of PAL low.

So bearing in mind that the GBA has no maths unit, and that the 030/50 is probably a bit faster, and that I am talking about plain polygons, and perhaps gouraud, not textured as the 3DO  and GBA throw around, I think your backing up my ideas, thanks very much!

*edit* If what your saying about the speed of the 3D0 is true, maybe the Amiga could run something like Star Fighter 3000, I used to love that on the 3D0.

[Khephren]

The GBA can do some impressive stuff, considering it's all CPU driven:

http://www.youtube.com/watch?v=zr_DkfO_Csg&feature=related
http://www.youtube.com/watch?v=t5_cnZabhig&feature=related
http://www.youtube.com/watch?v=6XNo1auSJhY&feature=related

As iv'e said plain polys are an order of magnitiude easier to draw than Texture mapped ones, this is why you see no texture mapped racers on the A500.
The GBA can draw quite good textured+vertex shaded poly environments. It would be able to draw a hell of a lot more polys if it dropped the texture maps.

Same for the Playstation (although it's a GPU equiped machine, made for textured polys, so you only get a x2 performance boost in poly count by dropping texturing).

No plain poly racer was ever released for either of these machines, so i'm not sure how you can say it would not be possible? (not at the quality of the pic I added, certainly).

NatAmi: They reckon an improved 060 at 150mhz last I heard. I'm still skeptical....but not as skeptical as the last few years!  

anyway, great discussion! I had hoped some coders would chime in and give us some actual figures.

[Khephren]

Yes of course, but you cannot entirely mitigate the fact that you have to combine instructions in a RISC architecture (you have two thirds less than a 68k), this is especially true of ported software. Hatiqs initial data that a 12.5mhz 3D0= 25mhz 030 sounds correct. Your RISC nets you a x2 performance increase per clock cycle, with the rest lost during combiner operations.

It's hard to compare the actual machine architectures though. As I pointed out, the 3D0 has a DSP, FPU and twin co-pro's. A 25mhz Amiga will not match this.

The Amiga has bitplane problems, or a chunky convertion overhead.

The GBA is probably the purest CPU renderer, but hampered by a lack of ram.

Doom on an 030/50 will run at a higher resolution, but about the same frame rate as the GBA version, but the GBA version looks worse than it has to because they down rezed the textures, not the CPU's fault in this case. Still not comparing like with like, because the Amiga and GBA versions are optimised for their architectures. Probably as close as we will get though.

@Hattiq
Yes it's very elegant instruction set. That's probably why it's poised for domination while having 1/3 the instructions it will not need to combine 3x as much. They were very careful in their instruction choice, and use of branch prediction.

Moto used to quote 1 MIPS standard, 2 MIPS peak at 8Mhz. I always thought they inflated their figures though. 1 MIPS sounds about right, given the Amiga never got the best out of a given processor. So .125DMIPS.

*edit* Iv'e posed the question "what's the render penalty for plain shaded polys VS gouraud VS textured on a CPU bound architecture?"  over at 'the chaos engine' profesional game dev forum. Lots of 3D programmers over there, including some old Amiga guys. I reckon a textured poly is about 8x slower, but it's just a guess..

[Khephren]

Doom is the closest iv'e come across. But as I pointed out, they are each optimised for their platform   MIPS are not really all that valid between architectures. We could compare Archie games, but as most were ported from amiga CISC to Acorn RISC, that would be a bit unfair.

[Khephren]

course it matters, most of us now own machines that are over 20 years old, but still love talking about this stuff

Anyway, i don't think the original archie ARM2 was a match for an 030, if thats what your getting at

[Khephren]

Maybe the 8MHz ARM2 would have matched the 16MHz 68030 crippled with 16-bit bus in the Atari Falcon!.

lol yeah, I like the falcon though. Borg that with an A1200, i'd have been happy!

I almost bought an Archie A3010 by mistake you know, thought it was an A1200 in the shop, only when he asked for £500 that I realised it wasn't the new A1200

@Digiman
Yeah, we can use c2p we have to.  Lord Riton over at eab has got a HAM8 engine going, it's pretty nice. Doubt we'd need that many colours, i'm talking about plain polys after all. If we use shading, then maybe.

I got some feed back from the chaos engine game dev site (not as much as i would have liked):

"with a plain poly you can just write the same colour to each part of the scanline the poly covers. There are no calculations or lookups when writing a scanline and often depending on the raw bitplane formats you can write multiple pixels at once.

for gouraud shading you need to calculate the start and end colours for each scanline and interpolate between them, worst case is you lose the ability to write multiple pixels at once but depending on how much memory you have to play with you may be able to still do this.

for texture mapping you have to work out start and end pixels in the texture (these may not be on the same vertical or horizontal line) and then write these out to the correct scan line(s) with possibly variable scaling in the x. This can get nasty if you have polys which have too much throw in the z.


you're probably safe saying each is an order of magnitude slower than the last.
Also performance is dependant on poly size and small polys will look better than big ones."

Also:
 well I know *wink wink* that one of the coders in our demo group is working on a flatshade engine that'll run ~15,000 polys on 060. Without all the trimmings of a game, obviously (Loonies)