English Amiga Board


Go Back   English Amiga Board > Main > Amiga scene

 
 
Thread Tools
Old 22 July 2019, 13:31   #601
grond
Registered User

 
Join Date: Jun 2015
Location: Germany
Posts: 634
Quote:
Originally Posted by roondar View Post
Like I said, like for like would require a setup I don't have. And sorry, but I'm not going to recompile DOOM just for this thread.
Actually it would suffice to hexedit the binary but I guess that is out of reach, too.


Quote:
That said, you yourself pointed out that DOOM was not so complicated and really rather efficient at what it did when first released. You also said that DOOM didn't care as much about a fast processor and instead really needed memory bandwidth. Expecting a massive improvement in optimisation to the core game IMHO does not fit with these earlier statements.
I understand your points but I think that indeed you can still (or again) optmise code that is written to achieve optimal memory access times on a specific CPU architecture when porting it to another CPU architecture. I also assume that ADoom has more ASM than the original x86 sources had which means that ID left more work to the (presumably better) x86 compilers for economical reasons than the Amiga coders working for idealistic reasons.


Quote:
Everything you post is technically true. And understand, I am not saying c2p should have zero impact at all. By now, I actually accept that the difference between planar and chunky come to about a CPU grade as you said earlier.
I was looking at CPU grades when 030/386 class processors were at about 25 MHz which I consider just enough to play Doom-type games. A 25 MHz 030 needs twice as much CPU time to perform a c2p for a frame as a 50 MHz 030. C2p-related overhead is bigger for slower CPUs when keeping the memory bus constant. It eventually disappears (fast 040s and 060s can do c2p as fast as a mere copy operation). Of course, it comes back the moment you make the graphics bus faster and increase the resolution. An AGA Amiga sweats a lot when doing c2p for a 640x400 resolution.


Quote:
This actually fits with your position that there is about a CPU grade of difference between the two - that is simply not a lot.
I'm pretty sure that when the A3000 was out, the price difference between 16, 25 and perhaps 33 MHz CPUs was a three digit number in any of $, UKP or hindsight-€. We were discussing how a slightly-above-budget or mid-range Amiga could have been made with little extra effort from Commodore that would have at least provided a path to Doom-capable Amigas. Spending 100$ more on the CPU would have eaten all the extra budget that was meant to go into an improved chipset, IDE connector, some fastmem and the CPU card connectors (my scenario). Adding a few gates and wires to the custom chips in order to support a chunky 8 bit mode would have come a lot cheaper.


Quote:
The perceived difference in speed between a CPU and a CPU that is 15% faster is very minimal. The same goes for frame rates: sure, a 40MHz 68030 might end up being 15% slower than a 40MHz 386DX when rendering Doom, but that's not going to be all that notable (this would a difference on the order of 2-3FPS).
Well, if you degrade a 40 MHz 030 to a 33 MHz 030, you'll probably end up with 83% of the FPS. On the lower end of Doom-capable processors this malus may push you over the edge from playable to not-enjoyable.


Quote:
This isn't going anywhere and frankly, I have better things to do with my time than to discuss this subject if this is how you're going to react.
I'm perfectly OK if we decide to stop discussing this. I was enjoying this discussion and have made my points clear. I made that statement in jest and expected you to be able to take it in a good-humoured way.

However, I do hold my point that having to OR together two chunky buffers, a 64 colour one and a 4 shadow steps one, to achieve the same effect as using two bitplanes for shadows is such a negligible disadvantage of chunky graphics that I will consider it irrelevant. I already admitted that my "no advantages" statement was (slightly) exaggerated. I made that statement because I hardly can see a meeting room full of people discussing the next computer generation's graphics architecture in 1989, everybody being in favour of a chunky 8 bit mode until somebody steps up and says: "but what about shadow effects in 2D shooters? Aren't those slightly easier to achieve in bitplanar graphics?" and then everybody is so impressed that they decide to keep it planar for the 8 bit mode. That's why I called it "irrelevant".


Quote:
Which is not really relevant. My point was that they kept doing new planar stuff. You don't do that if you think it's useless. Had they only kept the 8-bit planar modes and not added any new things, that would've been different. But they didn't.
Well, I'm not very clear about the exact planar modes they wanted to have. In the document you linked the term seems to be used rather loosely. I think they just wanted to make sure they didn't overlook anything and thus wanted to have lots of different modes which didn't cost much in R&D anyway.
grond is offline  
Old 22 July 2019, 13:39   #602
grond
Registered User

 
Join Date: Jun 2015
Location: Germany
Posts: 634
Quote:
Originally Posted by roondar View Post
Such as this list of total Amiga sales world wide:

1985 - 100K
1986 - 200K
1987 - 300K
1988 - 400K
1989 - 600K
1990 - 750K
1991 - 1,0M
1992 - 390K
1993 - 155K
1994 - 50K
The A3000 came at the height of the Amiga's popularity. It was ridiculously expensive and lacked colours and colour resolution. If it had had a chunky-AGA chipset, a cut-down version of it in 1991 might have repeated the A500's success with sales of the cut-down A3000 taking off a couple of years after initial publication just like the A500 did. That would have been in time for Doom.
grond is offline  
Old 22 July 2019, 22:01   #603
roondar
Registered User

 
Join Date: Jul 2015
Location: The Netherlands
Posts: 1,242
Quote:
Originally Posted by Foebane View Post
DOS was a pain, as I said, and it wasn't until Windows 95 came out later that year, and I was finally able to get it, that I was able to embrace the PC lifestyle.
I'm not sure if I understood your time line correctly, but just in case: Windows 95 was released august 1995, so you're talking about that year. Right?

Personally I held on a bit longer to my A1200. Partly because I really enjoyed using it (even went the whole 'Doom clone on Amiga route on my poor, overworked 68030). And partly because in 1995 I simply couldn't afford a 'proper' PC. I eventually did get one in 1997 or 1998 (I forget, it was around the time CU Amiga folded). Primarily for professional reasons, though once I had one I started playing Starcraft and the rest was history.

Also got a PSX around 1997 or 1998. Good times, but in reality the Amiga was long dead by then. A 1992 design can't really compete with PC's that new.
Quote:
Originally Posted by grond View Post
Actually it would suffice to hexedit the binary but I guess that is out of reach, too.
Yup. I've already spent far more time on this thread than I care to admit. Besides, I've never once claimed that c2p didn't slow the Amiga down for 3D games so I really don't see the point.
Quote:
I understand your points but I think that indeed you can still (or again) optmise code that is written to achieve optimal memory access times on a specific CPU architecture when porting it to another CPU architecture. I also assume that ADoom has more ASM than the original x86 sources had which means that ID left more work to the (presumably better) x86 compilers for economical reasons than the Amiga coders working for idealistic reasons.
I don't doubt that the Amiga has more assembly in it. But that by itself proves little in terms of performance gain. The main question is by how much it would be improved. Personally, I'd guesstimate such a reworking may get you a few percent. But gains will likely be fairly limited and are unlikely to make the renderer much faster as that part is clearly bandwidth limited more than anything else.
Quote:
I was looking at CPU grades when 030/386 class processors were at about 25 MHz which I consider just enough to play Doom-type games. A 25 MHz 030 needs twice as much CPU time to perform a c2p for a frame as a 50 MHz 030. C2p-related overhead is bigger for slower CPUs when keeping the memory bus constant. It eventually disappears (fast 040s and 060s can do c2p as fast as a mere copy operation). Of course, it comes back the moment you make the graphics bus faster and increase the resolution. An AGA Amiga sweats a lot when doing c2p for a 640x400 resolution.
I've seen Doom on 25 and 33MHz 386's. It's not pretty. I've also seen Doom on an Amiga with a 28MHz 68030. Also not pretty. Not very different either.

Don't get me wrong: the 68030 is slower here. But because they're both really slow, it doesn't really matter.
Quote:
I'm pretty sure that when the A3000 was out, the price difference between 16, 25 and perhaps 33 MHz CPUs was a three digit number in any of $, UKP or hindsight-€. We were discussing how a slightly-above-budget or mid-range Amiga could have been made with little extra effort from Commodore that would have at least provided a path to Doom-capable Amigas. Spending 100$ more on the CPU would have eaten all the extra budget that was meant to go into an improved chipset, IDE connector, some fastmem and the CPU card connectors (my scenario). Adding a few gates and wires to the custom chips in order to support a chunky 8 bit mode would have come a lot cheaper.
It still doesn't matter. The perceived speed of going up one grade of CPU is so small you just don't really note the difference. What I mean is: objectively Doom will run better on such a PC. Subjectively, they're about the same.
Quote:
Well, if you degrade a 40 MHz 030 to a 33 MHz 030, you'll probably end up with 83% of the FPS. On the lower end of Doom-capable processors this malus may push you over the edge from playable to not-enjoyable.
On the lower end, Doom is not very enjoyable, period. I've shown this already, 386 Doom runs poorly. All the way to the 386DX40. At 25MHz it's basically unplayable. I've also shown that, despite my expectations, Doom on a 68030 Amiga runs surprisingly similarly to a 386 at the same clock speed. There's not much in it.
Quote:
I'm perfectly OK if we decide to stop discussing this. I was enjoying this discussion and have made my points clear. I made that statement in jest and expected you to be able to take it in a good-humoured way.
Perhaps I could've worded it a bit nicer, but my issue here is that you seemingly went out of your way to not have to answer my points. Call it me misunderstanding the joke.
Quote:
However, I do hold my point that having to OR together two chunky buffers, a 64 colour one and a 4 shadow steps one, to achieve the same effect as using two bitplanes for shadows is such a negligible disadvantage of chunky graphics that I will consider it irrelevant. I already admitted that my "no advantages" statement was (slightly) exaggerated. I made that statement because I hardly can see a meeting room full of people discussing the next computer generation's graphics architecture in 1989, everybody being in favour of a chunky 8 bit mode until somebody steps up and says: "but what about shadow effects in 2D shooters? Aren't those slightly easier to achieve in bitplanar graphics?" and then everybody is so impressed that they decide to keep it planar for the 8 bit mode. That's why I called it "irrelevant".
I have three problems with this:

1) If planar to chunky is bad because it slows down graphics, then chunky emulating easy to do planar features at much greater cost (your solution essentially doubles bandwidth requirements for drawing anything) is also bad. No double standards. Either c2p is no big deal because all it does is increase bandwidth requirements (which you just deemed to be irrelevant if they happen for a chunky mode screen), or this is a valid advantage because it saves masses of bandwidth. You can't have it both ways.

2) I named two main advantages and you only really responded to one of them: the smaller one. The big one is multi-layer scrolling. This is much, much cheaper to do using bitplanes (especially once you go over two layers) and was a big thing in 1989-1993.

3) We were not discussing whether or not chunky/planar should've been put in a chip and what engineers would think of this. We were discussing if planar had any advantages over chunky.
Quote:
Well, I'm not very clear about the exact planar modes they wanted to have. In the document you linked the term seems to be used rather loosely. I think they just wanted to make sure they didn't overlook anything and thus wanted to have lots of different modes which didn't cost much in R&D anyway.
Adding a 16 bit planar mode just for the hell of it? Not buying that one. They specifically mention one of the advantages I've pointed before (Dual Layer screen on the cheap) when talking about it.

Adding a ton of pixel formats just for fun and because it was 'cheap'? Not really buying that either. Silicon is precious, adding useless features is in no one's interest - least of all engineers fighting to get funding, while working at a company that wants the cheapest possible solution.
Quote:
Originally Posted by grond View Post
The A3000 came at the height of the Amiga's popularity. It was ridiculously expensive and lacked colours and colour resolution. If it had had a chunky-AGA chipset, a cut-down version of it in 1991 might have repeated the A500's success with sales of the cut-down A3000 taking off a couple of years after initial publication just like the A500 did. That would have been in time for Doom.
It might have. But Commodore were never going to do AGA in a low end Amiga in 1991: they sold so many A500's in 1990 and 1991 that replacing it at that time would have been a crazy thing to do. You don't replace something that only just became a cash cow. Even properly run companies don't do that

Now... Prepare a new system for immediate launch once the 'cash cow' starts declining? Yeah, that's what they should've been working on much harder and much earlier. AAA started in 1989 and was still not finished in 1993. Personally, I've always though that the CDTV might have had something to do with that. Apparently it was a very rare thing for Commodore: a machine with a large R&D budget and management support. As I understand it, they poured tons of money into it.
roondar is offline  
Old 22 July 2019, 22:48   #604
chb
Registered User

 
Join Date: Dec 2014
Location: germany
Posts: 154
Quote:
Originally Posted by roondar View Post
Appologies for snipping parts of your reply, the idea is to make this all somewhat less of a wall of text.
No problem, good idea.
Quote:
Originally Posted by roondar View Post
I think we're reading far too much into an off-the-cuff remark made to a receptive audience.
Haynie is clearly very enthusiastic about the Amiga (and the opposite about C= management), but in general not known for outrageous claims. My intention was to analyze what he could have meant with his statement, not to praise AAA.

Quote:
Originally Posted by roondar View Post
It's clear from the AAA documentation you have to bend over backwards to get anywhere near the figure he mentioned. Even when we do so and add a non-existing DSP, we're still nearly 100MB/sec short for the lower end of his range.
No, far from it, see link below.
Quote:
Originally Posted by roondar View Post
I guess that's my main problem with such figures: having a system with an aggregate of about 70MB/sec of bandwidth to VRAM when using the CPU/Blitter combined and then claiming 400-600MB/sec bandwidth feels wrong to me. Even if it were technically correct (see the first bit of my edit), you'll never be able to actually do anything useful with all that bandwidth other than displaying high resolution still images.
I think that's the misunderstanding here: When we talk about bandwidth, we have different definitions. You talk about bandwidth to VRAM, which is indeed poor on AAA, while I am sure Haynie talked about aggregate bandwidth, meaning the total amount of data that can be read or written at the same time from all different memory systems under ideal conditions (after reading the doc linked below I am sure he did not include DSP internal bandwidth, which btw. at one longword per clock cycle does not equal 66 MB/s at 66 MHz). I tried to make this clear several times in my earlier posts: The naked aggregate figure does not tell you much about the performance for your specific task.

But take a look for yourself if you're interested: Acutiator specification, that is probably the system Haynie talked about. Esp. chapter 6.4 is relevant here, but also the other parts are quite interesting (and that "non-existing" DSP is mentioned all over the place ).
Acutiator was planned as the med-to-high end architecture mainly for AAA, but also AGA, it had hardware c2p (Akiko-style), but more importantly, a clever interface between chipmem and CPU to buffer writes, letting the CPU write to the buffer at full speed, after which it could perform other work while the data was written to chipmem, making use of burst modes. Those two alone would have solved quite a number of shortcomings of the A1200.

So while the A1200 clearly was in a lower segment, there were a couple of solutions in that doc that are relevant for this topic (things that Commodore could have done rather easily to improve the A1200), that's why I wrote to this thread. I will not ramble further about the bandwidth topic, I think we just agree to disagree here.

Last edited by chb; 22 July 2019 at 22:55. Reason: stylistics
chb is offline  
Old 23 July 2019, 06:22   #605
sandruzzo
Registered User
 
Join Date: Feb 2011
Location: Italy/Rome
Posts: 1,463
A++ Should have be done

https://en.wikipedia.org/wiki/Commodore_AA%2B_Chipset
sandruzzo is offline  
Old 23 July 2019, 10:18   #606
Steril707
OCS rocks!

 
Join Date: Sep 2016
Location: Amiga Island
Posts: 869
Quote:
Originally Posted by roondar View Post

It might have. But Commodore were never going to do AGA in a low end Amiga in 1991: they sold so many A500's in 1990 and 1991 that replacing it at that time would have been a crazy thing to do. You don't replace something that only just became a cash cow. Even properly run companies don't do that
They should have shown some high end "Halo System" with AGA capabilities back in 1990/91, and tease that this might end up in a keyboard version.

So consumers would have still bought the A500s, but the message would have been that the Amiga is still on top of the graphics game.

And then release the A1200.
Instead we got nothing, and moved over to PCs to play Wing Commander II and Ultima 6 with 256 colours.
Steril707 is offline  
Old 23 July 2019, 12:53   #607
grond
Registered User

 
Join Date: Jun 2015
Location: Germany
Posts: 634
Quote:
Originally Posted by roondar View Post
Besides, I've never once claimed that c2p didn't slow the Amiga down for 3D games so I really don't see the point.
Actually, you have numerous times claimed that it didn't make much of a difference to a point where you found it nigligible. It clearly wasn't.


Quote:
Don't get me wrong: the 68030 is slower here. But because they're both really slow, it doesn't really matter.
It still doesn't matter. The perceived speed of going up one grade of CPU is so small you just don't really note the difference. What I mean is: objectively Doom will run better on such a PC. Subjectively, they're about the same.
On the lower end, Doom is not very enjoyable, period. I've shown this already, 386 Doom runs poorly. All the way to the 386DX40. At 25MHz it's basically unplayable. I've also shown that, despite my expectations, Doom on a 68030 Amiga runs surprisingly similarly to a 386 at the same clock speed. There's not much in it.
Your argumentation with regard to Doom on chunky or planar mode seems to boil down to "it's either too slow anyway or the CPU is fast enough so it doesn't matter". This doesn't make sense. There is a lot of midland between a slow 030 and a CPU that is fast enough regardless of the graphics being planar or chunky. This midland did matter in the early 90s because it meant a lot of $$$ per Amiga (or PC for that matter). The midland did matter because new game techniques tend to be implemented the moment the hardware has just become capable of running them. A hardware that has to bear the additional burden of planar graphics is just not capable at all at that point in time.


Quote:
I have three problems with this:

1) If planar to chunky is bad because it slows down graphics, then chunky emulating easy to do planar features at much greater cost (your solution essentially doubles bandwidth requirements for drawing anything) is also bad. No double standards. Either c2p is no big deal because all it does is increase bandwidth requirements (which you just deemed to be irrelevant if they happen for a chunky mode screen), or this is a valid advantage because it saves masses of bandwidth. You can't have it both ways.
No. c2p usually affects the entire screen. Shadow effects for precalced graphics (that's all you can really do in planar!) are small snippets of the screen that get ORed into the main screen. Yes, this takes twice as much bandwidth as planar. However, ORing two buffers together is a very simple operation that will have source and destination in fastmem and nonetheless can be carried out at copy speed. I think this disadvantage is very small compared to the fact that planar graphics are very impractical for ALL types of rendered graphics (including flat-shaded polygons) while an 8 bit chunky mode can just as well do almost everything what planar modes can do.


Quote:
2) I named two main advantages and you only really responded to one of them: the smaller one. The big one is multi-layer scrolling. This is much, much cheaper to do using bitplanes (especially once you go over two layers) and was a big thing in 1989-1993.
Sorry, I don't understand. A chunky buffer can scroll in arbitrary directions just like a planar buffer can. Your example seems to be scrolling individual bitplanes in different directions. I'm not sure where this technique would be relevant. You can do dual-playfields in chunky, too. Of course, this would only work well (i.e. not waste DMA bandwidth and chipmem) for 4, 16 or 256 colours per playfield. Again I don't see much relevance in this discussion as I never said that the OCS planar modes where a mistake or should have been abandoned with AGA. I think AGA would have done better even with just an 8bit chunky mode and no added planar modes. Of the two possibilities, chunky 8 bit modes are clearly the more desirable if you have to choose.


Quote:
3) We were not discussing whether or not chunky/planar should've been put in a chip and what engineers would think of this. We were discussing if planar had any advantages over chunky.
I already admitted that there were some cases where planar graphics have some advantages over chunky graphics. I don't think these cases are very relevant when compared to the advantages that chunky graphics offer. You say it is merely a question of personal preferences and 2D bitplane scrollers are more enjoyable to some than rendered graphics and who could judge that. The market agreed with me. And as it seems, Jay Miner agreed because I'd be surprised his statement in the above video was done in the few months between Doom and his death.


Quote:
Adding a 16 bit planar mode just for the hell of it? Not buying that one. They specifically mention one of the advantages I've pointed before (Dual Layer screen on the cheap) when talking about it.
Yes, but I can't remember whether they actually stated that the two layers were supposed to be planar or whether they may have been 8bit chunky. There were other modes where it sounded very much like they really intended to have planes of chunky pixels, e.g. for 24 bit graphics and some of the more complex packed formats. They clearly noticed that HAM, which was a left-over feature from the early days when they wanted to have some YUV-type colour model (or was it HSB?), was considered to be much more useful than they had expected and thus wanted to have some more such modes. I think they were probably overdoing things.


Quote:
Adding a ton of pixel formats just for fun and because it was 'cheap'? Not really buying that either. Silicon is precious, adding useless features is in no one's interest - least of all engineers fighting to get funding, while working at a company that wants the cheapest possible solution.
Supporting more pixel format requires almost no silicon footprint. All it takes are a few more switches and latches to route the data differently.


Quote:
It might have. But Commodore were never going to do AGA in a low end Amiga in 1991: they sold so many A500's in 1990 and 1991 that replacing it at that time would have been a crazy thing to do. You don't replace something that only just became a cash cow. Even properly run companies don't do that
I didn't say they were supposed to stop selling the A500. They could either have made the AGA+chunky an A500+ and enjoyed the popularity of the earlier model carrying over to what would have been marketed as a revision of the A500 or, if it had proven impossible to reach a comparable pricepoint as that of the A500 when it was introduced, could have offered the new machines in addition to the A500. In the latter case sales of the A500 would probably have slowly decreased over a period of about two years while sales of the new machine would have taken off in the same way (with the price of the new machine coming down over time).
grond is offline  
Old 23 July 2019, 14:02   #608
activist
Registered User

 
Join Date: May 2017
Location: here
Posts: 42
Quote:
Originally Posted by roondar View Post
It might have. But Commodore were never going to do AGA in a low end Amiga in 1991: they sold so many A500's in 1990 and 1991 that replacing it at that time would have been a crazy thing to do. You don't replace something that only just became a cash cow. Even properly run companies don't do that
Not that crazy. AA/AGA was never meant to happen remember. With hindsight and data here we know commodore sales fell off a cliff in '92. So 1991, the time was actually overdue for :
a) A300 (not the A600, god dammit why or what were they thinking)
b) low end ocs/ecs successor. Something like this for sure (thanks Sandruzzo, hadn't read) :

Quote:
Originally Posted by sandruzzo View Post

Quote:
Originally Posted by grond View Post
or, if it had proven impossible to reach a comparable pricepoint as that of the A500 when it was introduced, could have offered the new machines in addition to the A500. In the latter case sales of the A500 would probably have slowly decreased over a period of about two years while sales of the new machine would have taken off in the same way (with the price of the new machine coming down over time).
assuming there was a big enough difference at the time between the new machine and A500/A300 (there wasn't with AGA) the A500/A300 is something commodore could have/should have kept going regardless. There'd still be a demand for it today e.g. C64 reboot

Last edited by activist; 23 July 2019 at 14:48.
activist is offline  
Old 23 July 2019, 15:43   #609
grond
Registered User

 
Join Date: Jun 2015
Location: Germany
Posts: 634
AA+ is pretty much what AGA should have been. The developer's comment quoted in the wiki-article translates to "AA+ was going to have all the things we would have done if the AA project had had higher priority and funding". IIRC Haynie said something to the effect that during some time anyone found working on an Amiga-related project was threatened to be fired immediately.
grond is offline  
Old 23 July 2019, 20:43   #610
roondar
Registered User

 
Join Date: Jul 2015
Location: The Netherlands
Posts: 1,242
Quote:
Originally Posted by chb View Post
No problem, good idea.

Haynie is clearly very enthusiastic about the Amiga (and the opposite about C= management), but in general not known for outrageous claims. My intention was to analyze what he could have meant with his statement, not to praise AAA.
As I pointed out in a (late) edit a few posts back:
"I've been thinking. It's possible that he was talking about theoretical limits - the CPU/Blitter may not have the speed to get to a 400-600MB/sec total, but perhaps the VRAM does support it (if the both ports are used to their theoretical limits). And maybe the BUS theoretically does as well. So if you put in a hypothetical CPU/Blitter combo that can push 200MB/sec or more, you'd actually get to those sort of numbers."

Reading the document you've added makes that much more likely as it's filled with all sorts of theoretical options. Such as clusters of CPU's or using one or more DSP's instead of a main CPU. Don't get me wrong - those a cool ideas for 1992 if ever there were some. But they also show that this is a particular kind of document - one that talks about many options, not one that talks about a fixed design.

I'd argue the document I linked (which was much newer) was closer to what they were trying to actually build.
Quote:
I think that's the misunderstanding here: When we talk about bandwidth, we have different definitions. You talk about bandwidth to VRAM, which is indeed poor on AAA, while I am sure Haynie talked about aggregate bandwidth, meaning the total amount of data that can be read or written at the same time from all different memory systems under ideal conditions (after reading the doc linked below I am sure he did not include DSP internal bandwidth, which btw. at one longword per clock cycle does not equal 66 MB/s at 66 MHz). I tried to make this clear several times in my earlier posts: The naked aggregate figure does not tell you much about the performance for your specific task.
Yup, that's pretty much it - I read 600MB/sec and considered that nonsense because it would've been essentially nearly fully useless. But as I pointed out above, it's possible he was talking about theoretical limits. In which case it probably makes sense.

Quote:
But take a look for yourself if you're interested: Acutiator specification, that is probably the system Haynie talked about. Esp. chapter 6.4 is relevant here, but also the other parts are quite interesting (and that "non-existing" DSP is mentioned all over the place ).
Acutiator was planned as the med-to-high end architecture mainly for AAA, but also AGA, it had hardware c2p (Akiko-style), but more importantly, a clever interface between chipmem and CPU to buffer writes, letting the CPU write to the buffer at full speed, after which it could perform other work while the data was written to chipmem, making use of burst modes. Those two alone would have solved quite a number of shortcomings of the A1200.

So while the A1200 clearly was in a lower segment, there were a couple of solutions in that doc that are relevant for this topic (things that Commodore could have done rather easily to improve the A1200), that's why I wrote to this thread. I will not ramble further about the bandwidth topic, I think we just agree to disagree here.
I did look at your document (which is awesome, thanks for the link!) and agree to disagree

Still, it's been nice to dig around in old hardware. We should do that again sometime
Quote:
Originally Posted by grond View Post
Actually, you have numerous times claimed that it didn't make much of a difference to a point where you found it nigligible. It clearly wasn't.
You yourself have claimed the exact same thing numerous times - to the point where you pegged the difference at about 10-15MHz of CPU clock. That is pretty small and simply won't really change the overall Doom experience much. The entire point I've been making since seeing the Amiga version in action is:
  1. Doom on Amiga is slower than Doom on PC due to c2p & AGA bandwidth issues.
  2. In the real world and given several years of optimising c2p routines, the factual difference in FPS is not that big. About 2-3FPS for the low end, 2-3FPS for the medium end and about 5-10FPS for the high end. Objectively, these are all differences that are significant in their percentage. However, in real world the perceived difference is simply not such a big deal.
  3. Low end is unplayable regardless, 8 or 5 FPS is not going to matter. Slower on Amiga, unplayable on both
  4. Medium end is the only interesting one: is Doom unplayable if it doesn't run at 15FPS, but rather at 12? Having now seen both it seems to me the answer is no. There's a difference, but neither is great.
  5. At high end it becomes irrelevant. Sure, 25FPS is a good deal less than 35, but both are perfectly playable.
None of these claims are outlandish and I've explained at length that this is how I think about it since seeing an actual Amiga or two running the darned game. It's how it looks in the real world that causes me to feel this way. 15 vs 12 FPS sounds like a big deal unless you actually look at them one after the other. You just don't really notice as much unless you place them side by side.
Quote:
Your argumentation with regard to Doom on chunky or planar mode seems to boil down to "it's either too slow anyway or the CPU is fast enough so it doesn't matter". This doesn't make sense. There is a lot of midland between a slow 030 and a CPU that is fast enough regardless of the graphics being planar or chunky. This midland did matter in the early 90s because it meant a lot of $$$ per Amiga (or PC for that matter). The midland did matter because new game techniques tend to be implemented the moment the hardware has just become capable of running them. A hardware that has to bear the additional burden of planar graphics is just not capable at all at that point in time.
Yes it does make sense and no, the midland is not relevant. Seriously. Doom doesn't run well on any 'mid-range' PC's in 1993. It only starts running well once you get into 486-with-vlb territory. In 1993, that was high end. Not mid end.

Frankly, if you accept the 15 or so FPS Doom might do on a fast 386 you'll have little problems accepting an Amiga running at 12 FPS. I've seen both in action, there's not much in it.
Quote:
No. c2p usually affects the entire screen. Shadow effects for precalced graphics (that's all you can really do in planar!) are small snippets of the screen that get ORed into the main screen. Yes, this takes twice as much bandwidth as planar. However, ORing two buffers together is a very simple operation that will have source and destination in fastmem and nonetheless can be carried out at copy speed. I think this disadvantage is very small compared to the fact that planar graphics are very impractical for ALL types of rendered graphics (including flat-shaded polygons) while an 8 bit chunky mode can just as well do almost everything what planar modes can do.
Full screen transparency (non-shadow) effects have been done in AGA games already. At full frame rate. You might not like the games that did it, but that doesn't change it.
Quote:
Sorry, I don't understand. A chunky buffer can scroll in arbitrary directions just like a planar buffer can. Your example seems to be scrolling individual bitplanes in different directions. I'm not sure where this technique would be relevant. You can do dual-playfields in chunky, too. Of course, this would only work well (i.e. not waste DMA bandwidth and chipmem) for 4, 16 or 256 colours per playfield. Again I don't see much relevance in this discussion as I never said that the OCS planar modes where a mistake or should have been abandoned with AGA. I think AGA would have done better even with just an 8bit chunky mode and no added planar modes. Of the two possibilities, chunky 8 bit modes are clearly the more desirable if you have to choose.
There is relevance in this discussion because despite your continued protest that 'you can do dual playfields in chunky, too', this is actually 100% false. It's only true if you change the definition of 'chunky' in such a way it becomes much more like planar or a great deal less efficient (unless you use a 16 bit chunky screen, in which case it's merely 'equal' for a two layer screen. But at 16 bits, a planar screen can easily do 4 layers and you're back to square one). Properly done planar graphics for two layers do not need multiple writes to memory for dual layered graphics, nor inefficient nonsense like 2 bit or half byte pixels. Chunky modes with multiple layers definitely do, unless you turn them into planar-mode-lite.

Doing the latter would prove my point.
Quote:
I already admitted that there were some cases where planar graphics have some advantages over chunky graphics. I don't think these cases are very relevant when compared to the advantages that chunky graphics offer. You say it is merely a question of personal preferences and 2D bitplane scrollers are more enjoyable to some than rendered graphics and who could judge that. The market agreed with me. And as it seems, Jay Miner agreed because I'd be surprised his statement in the above video was done in the few months between Doom and his death.
The market in 1993-1996 didn't really agree though. 2D games consistently outsold 3D games at that time.

Here's the biggest 3D games of 1993-1996 (sales data is life-time sales, not for just the release window)*. Highest data I could find for Doom I & II says it sold about 4 million combined and about 2 million each. Wolvenstein 3D sold under 500.000 units and Quake sold about 1 million units. Duke Nukem 3D sold about 1 million units. Wing Commander 3 sold about 1 million units. Grand Prix 2 sold about a million units.

Let's now look at some PC 2D games from the same era: Command and Conquer sold 3 million units. Myst sold 6 million units. Theme Park sold 3 million units. Warcraft 2 sold 2 million units. Adding in console sales figures from the same time makes it all much more clear*: Street Fighter 2 Turbo: Hyper Fighting sold 4 million copies. Final Fantasy VI sold 3 million copies. Donkey Kong Country 2 sold 5 million units. Aladdin sold over 4 million copies.

Again, IMHO Doom nor 3D games killed the Amiga. A combination of consoles eating into the low end and PC's enveloping on all other markets did it it. 3D undoubtedly also damaged it, but looking at real world sales figures it's pretty clear that 3D games took a long time to dominate. It wasn't until well in the late 1990's/early 2000's that they started to seriously do well. Heck, even today 3D games are regularly outsold by 2D stuff (more so if you count '2.5D' as 3D).

*) I excluded older games or games that have been sold as a pack-in. I've also excluded portable games as that would just have been unfair.
Quote:
Yes, but I can't remember whether they actually stated that the two layers were supposed to be planar or whether they may have been 8bit chunky. There were other modes where it sounded very much like they really intended to have planes of chunky pixels, e.g. for 24 bit graphics and some of the more complex packed formats. They clearly noticed that HAM, which was a left-over feature from the early days when they wanted to have some YUV-type colour model (or was it HSB?), was considered to be much more useful than they had expected and thus wanted to have some more such modes. I think they were probably overdoing things.
The two layer screen modes were very much planar. It's right there in the documentation.
Quote:
Originally Posted by AAA docs
It supports up to 16 bitplanes, which makes dual 8-bit playfields possible
I do agree they were overdoing thing, by the way. But that's not really the point - the SNES did two layers of 256 colours each and no doubt the team at Commodore wanted to do the same. Hence: 16 bitplane, dual 8-bit playfields.
Quote:
Supporting more pixel format requires almost no silicon footprint. All it takes are a few more switches and latches to route the data differently.
Define 'almost no silicon'. Seriously, I'm interested. How many transistors for each of those formats?

If possible, including any and all 'overhead' for selecting the different modes, registers, whatever the video chips needs to do to get it to work basically. Preferably using whatever was known about circa 1992 and not stuff figured out later.

BTW, I have no idea if what I'm asking for is complicated to calculate and don't mean to cause you work - a ballpark estimate is fine for me.
Quote:
I didn't say they were supposed to stop selling the A500. They could either have made the AGA+chunky an A500+ and enjoyed the popularity of the earlier model carrying over to what would have been marketed as a revision of the A500 or, if it had proven impossible to reach a comparable pricepoint as that of the A500 when it was introduced, could have offered the new machines in addition to the A500. In the latter case sales of the A500 would probably have slowly decreased over a period of about two years while sales of the new machine would have taken off in the same way (with the price of the new machine coming down over time).
It still sounds very theoretical to me. Sony didn't replace the PS2 with the PS3 when the former was selling record numbers. Nor did Microsoft replace the XBOX 360 with the XBOX One in it's prime. Nor did Nintendo with the SNES/N64.

What you're describing is essentially never done while the prior product is still selling extremely well. All of these companies I mentioned (which were managed pretty effectively) only released their new products after the old ones had started to decline, which is the opposite of what Commodore saw happen.

Last edited by roondar; 23 July 2019 at 20:51.
roondar is offline  
Old 23 July 2019, 20:53   #611
roondar
Registered User

 
Join Date: Jul 2015
Location: The Netherlands
Posts: 1,242
Sorry for the double post, but the above is so big I fear this may get lost otherwise:

I'm going to cut back the length and time I spend on posts in this thread. It's taking far too much of my time and I'm apparently unable to simple say 'agree' or 'disagree' without going into great detail or looking up all sorts of data.

Do like the discussion and should anyone wonder, this is nothing personal. Purely a matter of time management. I'll also still reply - just less frequently and less in depth.

Last edited by roondar; 23 July 2019 at 20:57. Reason: Clarified what I meant.
roondar is offline  
Old 23 July 2019, 21:48   #612
swinkamor12
Registered User

 
Join Date: May 2017
Location: Vienna/Austria
Posts: 58
Amiga 500 has better and faster graphics than pc, better sound than pc, better games than pc.
Amiga 1200 has equal but many times slower graphics than pc, equal sound, worse games than pc.
Of course some fanatics will deny that, but
the ugly truth about Amiga 1200 is Commodore ask too much for Amiga 1200.
Amiga 1200 should have three things
- chunky pixels
- slots for fast ram
- simple mmu to protect first page of ram
Adding this will make price Amiga 1200 maybe few DM higher.
But Commodore screw it and thats why they bankrupt.
Lack of chunky pixels make AGA many times slower than afordable SVGA cards from 1992.
Higher resolutions with more colors was too slow to be usable on Amiga 1200 without fast ram.
Amiga 1200 need expensive additional hardware to add fast ram.
Amiga 1200 was underpowered overpriced shit.
I never regret that I sell Amiga 1200.
In 1992 Amiga still has better OS than PC but Amiga hardware was worthless.
swinkamor12 is offline  
Old 23 July 2019, 22:26   #613
roondar
Registered User

 
Join Date: Jul 2015
Location: The Netherlands
Posts: 1,242
Quote:
Originally Posted by swinkamor12 View Post
Amiga 500 has better and faster graphics than pc, better sound than pc, better games than pc.

Amiga 1200 has equal but many times slower graphics than pc, equal sound, worse games than pc.
All of the above is either completely false, mostly false or highly debatable.
Quote:
Of course some fanatics will deny that, but
https://en.wikipedia.org/wiki/Ad_hominem

Not going to bother to even respond to the rest of this, it's more of the same lies, half truths and exaggeration.
roondar is offline  
Old 23 July 2019, 22:33   #614
TEG
Registered User

 
Join Date: Apr 2017
Location: France
Posts: 34
When the 1200 came out, I immediately bought one to replace the 500+ and we had a 386 too so I remember well those times.

The Amiga had still a big advantage over the PC: the Workbench when the PC was running under DOS. But their was problems.

- First point was the hard drive. With the PC I was now accustomed with hard drive. And the hard drive for the A1200 was a 2"1/2 one. It was very expensive at the time compared to 3"1/2 ones. So the PC had a big advantage on this point for the price.

- The second point was the resolution. We bought the Commodore multisync monitor to have access to the AGA higher resolutions modes but it was deceptive because it was falling back to lower resolutions for most of the applications. And 4 colours is very few. At the same time, the PC video cards were rapidly improving and price falling. I remember that cards was soon able to display 16 millions colours and it was affordable. Imagine 16 millions colours ! Whoua !

And yes, Doom came and it was the hype. Windows came too and the affordable Amiga was still stuck with his fixed hardware.

I got a big memory expansion for the A1200 but it was very expensive compared with memory for the PC. Buying an accelerator card was unreachable, it was better to bought another PC, a 486DX2-66 which we overclocked if I remember well. The rendering time for ray-traced images was so cut down then because of the integrated math copro of the 486.

And A3000/4000 were too expensive with too few advantages, same resolution as an A1200.

However the Amiga kept an advantage for a long time over the PC for the sound ! Before the soundblaster card, sound cards were often incompatible with games, a nightmare. And the soundblaster was very expensive at first.

Last edited by TEG; 23 July 2019 at 23:07.
TEG is offline  
Old 23 July 2019, 23:03   #615
chb
Registered User

 
Join Date: Dec 2014
Location: germany
Posts: 154
Quote:
Originally Posted by roondar View Post
I did look at your document (which is awesome, thanks for the link!) and agree to disagree
Yep, let's leave it at that.
Quote:
Still, it's been nice to dig around in old hardware. We should do that again sometime
Would be great, but I fear it would neither help your time management nor mine.
Quote:
I'd argue the document I linked (which was much newer) was closer to what they were trying to actually build.
If you refer to the AAA documentation: They describe pretty non-overlapping parts of a complete system. The AAA doc is about the chipset (basically everything that works in chip ram) and has only brief mentions of fast ram, expansion busses etc., while the Acutiator doc is about the system architecture and does not go into detail about AAA. AAA was not tied to Acutiator, and Acutiator not to AAA. Both docs are form the same time frame, btw. (October 1992 for Acutiator and January 1993 for AAA).

Haynie is quite univocal about AAA and Acutiator:
Quote:
Originally Posted by https://www.bigbookofamigahardware.com/bboah/product.aspx?id=35
Had Commodore survived, the first AAA systems would have been based around the Acutiator architecture.
There's a lot of could-have-been stuff, and of course C= management could have messed with it in any possible way, but it's sad they could probably have survived if they had this ready 1-2 years earlier...
chb is offline  
Old 24 July 2019, 02:41   #616
Bruce Abbott
Registered User

Bruce Abbott's Avatar
 
Join Date: Mar 2018
Location: Hastings, New Zealand
Posts: 235
Quote:
Originally Posted by TEG View Post
the hard drive for the A1200 was a 2"1/2 one. It was very expensive at the time compared to 3"1/2 ones...
A bit more expensive yes, but the Amiga's OS is ROM based and most games can run from floppy disk, so a suitable drive didn't cost more.

Unless you had a serious case of PC envy and just had to have the biggest drive possible, in which case it was possible to stuff a 3.5" drive into the A1200.

Quote:
We bought the Commodore multisync monitor to have access to the AGA higher resolutions modes but it was deceptive because it was falling back to lower resolutions for most of the applications. And 4 colours is very few.
Yes, the higher resolutions were a bit deceptive, just like they were on most PCs at the time.

But if an application only needs 4 colors then what's the problem? On the PC it was a problem because applications either had to run in the desktop's resolution, or switch to DOS. On the Amiga it wasn't a problem because they could open a custom screen tailored to their needs. So why the angst? PC envy again...

Quote:
At the same time, the PC video cards were rapidly improving and price falling.
That's true. With the rate that PC hardware was improving, any design was bound to be obsolete within a few years. Still, it would be several years before the AGA chipset's limits were reached - if they actually have been even now!

Quote:
I remember that cards was soon able to display 16 millions colours and it was affordable. Imagine 16 millions colours ! Whoua !
But the A1200 already had 16 million colours, so...

Quote:
I got a big memory expansion for the A1200 but it was very expensive compared with memory for the PC. Buying an accelerator card was unreachable, it was better to bought another PC, a 486DX2-66
A1200 memory appeared to cost more because you first had to buy a RAM board, but the base A1200 was much cheaper than a PC and the SIMM modules cost the same so it wasn't actually much more expensive - it just seemed that way when doing the upgrade. And if you bought an accelerator card then the RAM slots were 'free'.

And the Amiga didn't need as much RAM as a PC for typical use anyway (try running Windows 95 in 4MB - it runs out of memory and starts swapping to the hard drive before it even gets to the desktop!). But again PC envy gets in the way. 8MB sounds much better than 4MB, even if half of it gets gobbled up by the OS.

Quote:
And A3000/4000 were too expensive with too few advantages, same resolution as an A1200.
Expensive yes, but so were high-end PCs at the time. And they did have one big advantage, the Zorro-III bus which could take RTG graphics cards etc. My A3000 with RTG card and 060 CPU was able to match the performance of a Pentium 90 PC.

Quote:
However the Amiga kept an advantage for a long time over the PC for the sound ! Before the soundblaster card, sound cards were often incompatible with games, a nightmare. And the soundblaster was very expensive at first.
That's true. I remember lots of frustration trying to get Sound Blaster 'compatible' cards to work with games, and later trying to get them going in Windows 95. Even today I can't get my Sound Blaster PCI CT4810 to work with a K6-II 500 in Windows XP.

Last edited by Bruce Abbott; 24 July 2019 at 03:15.
Bruce Abbott is offline  
Old 24 July 2019, 03:12   #617
Bruce Abbott
Registered User

Bruce Abbott's Avatar
 
Join Date: Mar 2018
Location: Hastings, New Zealand
Posts: 235
Quote:
Originally Posted by TEG
Buying an accelerator card was unreachable, it was better to bought another PC, a 486DX2-66
Depending on when this was that may be true, but when the 486DX2-66 CPU came out it was very expensive. The advert below is from the January 1993 issue of PC Magazine.

"QTech. the Best PC Values ANYWHERE!" - US$1399 for a 486SX-25, US$2499 for a 486DX2-66.

That's $1100 more for the DX2-66 CPU!

In 1998 a Blizzard 1230-IV 50MHz with 8MB RAM cost £114.95 (~$192).

Somewhere in between those dates it might have been possible to buy a 486DX2-66 machine for less than the cost of upgrading your A1200, but I bet you figured it was worth paying more for the PC anyway - just like all those PC owners who bought 386's around the time that the A1200 was released.
Attached Thumbnails
Click image for larger version

Name:	pc mag jan 1993.jpg
Views:	26
Size:	199.1 KB
ID:	63880  
Bruce Abbott is offline  
Old 24 July 2019, 14:01   #618
grond
Registered User

 
Join Date: Jun 2015
Location: Germany
Posts: 634
Quote:
Originally Posted by Bruce Abbott View Post
A bit more expensive yes
I googled a bit an found an ad with these prices:

A600: 549 DM
A600+40MB HDD: 948 DM

A1200: 889 DM
A1200+64MB HDD: 1450 DM
A1200+130MB HDD: 1850 DM
A4000/030+80MB: 2700 DM
A4000/030+120MB: 2850 DM
A4000/040+120MB: 4099 DM
A4000/040+210MB: 4599 DM

The ad must date from around spring 1993 because in summer 1993 the A600 cost only 299DM (that was when I bought it).

Looking at the above numbers shows that the 40MB 2.5" harddisk cost a whopping 400 DM. The 64MB 2.5" in the A1200 cost 560 DM, the 130MB 2.5" almost 1000 DM, more than the A1200 and 400 DM more than the smaller capacity model.

A similar step up in the 3.5" harddisks in the A4000 from 80 to 120 MB cost only 150 DM. The 3.5" 80 MB harddisk probably cost no more than 300 DM, more likely between 250 and 280 DM.

80 MB 3.5" for 300 DM max compared to 64MB 2.5" for 560DM? I'd say that is more than "a bit more expensive" as you claimed...

The stupid thing about the 2.5" harddisks is that they were meant to go into the budget Amiga while the much cheaper 3.5" harddisks went into the expensive big box Amigas.
grond is offline  
Old 24 July 2019, 14:16   #619
demolition
Unregistered User
demolition's Avatar
 
Join Date: Sep 2012
Location: Copenhagen / DK
Age: 39
Posts: 3,920
Quote:
Originally Posted by grond View Post
80 MB 3.5" for 300 DM max compared to 64MB 2.5" for 560DM? I'd say that is more than "a bit more expensive" as you claimed...
Which is also why those of my friends who were using A1200s used 3.5" harddrives as the 2.5" versions were so ridiculously expensive. They basically had the harddrive sitting on the desk outside the Amiga with the IDE cable running between the two halves of the chassis and it was also common to use a modified PC AT PSU to power both the HDD and the Amiga. It wasn't pretty and definitely not very practical when bringing the setup to parties, but it was the only way to get a lot of space for cheap.
demolition is offline  
Old 24 July 2019, 14:20   #620
vulture
Registered User
 
Join Date: Oct 2007
Location: Athens , Greece
Posts: 966
I know that's not indicative or some kind of universal truth, but the shop I got my A1200 from -per their suggestion- installed a 130MB 3.5'' hdd for me inside, at just above the cost of the device itself. Faster, cheaper and more reliable (still works to this day).
vulture is offline  
 


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools

Similar Threads
Thread Thread Starter Forum Replies Last Post
A1200 RF module removal pics + A1200 chips overview eXeler0 Hardware pics 2 08 March 2017 00:09
Sale - 2 auctions: A1200 mobo + flickerfixer & A1200 tower case w/ kit blakespot MarketPlace 0 27 August 2015 18:50
For Sale - A1200/A1000/IndiAGA MkII/A1200 Trapdoor Ram & Other Goodies! fitzsteve MarketPlace 1 11 December 2012 10:32
Trading A1200 030 acc and A1200 indivision for Amiga stuff 8bitbubsy MarketPlace 17 14 December 2009 21:50
Trade Mac g3 300/400 or A1200 for an A1200 accellerator BiL0 MarketPlace 0 07 June 2006 17:41

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +2. The time now is 15:50.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2019, vBulletin Solutions Inc.
Page generated in 0.12193 seconds with 16 queries