@cdimauro

Quote:

I've no time to reply to comments, because in the last period I was very busy writing article articles on my technical blog plus I've started again working on my architecture .

Excellent. I've been waiting to read this.

What's your architecture? Oh sorry, that implies a comment.

Quote:

However I've written another 16 articles about the packed vs planar graphics to further clarify the topic. Just reporting now what I've written on LinkedIn.

Wow. Another 16. Maybe you could release a book. Amiga: The Packed vs Planar Wars.

@Kronos

Quote:

Kronos wrote: @cdimauro The issue is that you are still 2nd guessing with 30+ years of hindsight.

I think that the primary issue is that you don't read what people are writing. Let me quote me again:

"I had this idea very long time ago.

I've also shared it on Olaf's Amiga coding forum around 13-14 years ago (I don't recall precisely now)."

and

"To me it was a natural, obvious, conclusion once I've mastered both planar and packed / chunky graphics."

In short: you do NOT need AGES to acquire the concept, since it's pretty self-evident (once you had the chance to work with both graphical formats).
Quote:

Same as every historian could tell you why Varus failed and was doomed to fail from the get go. Put them in the same situation with the same background and fog of war they would still end the same.

I don't know this story, so I can't understand the analogy.

Anyway, I think that it's relevant: see above.
Quote:

Reality is, Jay worked with primitive tools, a small budget and against an unrealistic timeframe. So yeah he did start with what he had learned from the AtariXL instead of starting from scratch (just like Intel with the 8086) and someone decided that this new system should do more then 16 "free" colors

Not accurate (development of the Lorraine / Amiga hardware was quite ahead of its software) and irrelevant (he and his team designed from scratch a lot of new things and the Amiga hardware is quite different from the Atari one).
Quote:

meaning that packed pixels would either waste bits or start at any given bit in a byte.

Again, you haven't read the articles and fall on exactly the same mistake.

So, again: you're statement is absolutely wrong!

Actually it's the planar graphic format which wastes more memory (AND bandwidth). NOT the packed one!

If you have no time to read the articles I can understand, but then please do NOT write completely wrong statements.
Quote:

Both a nogo when even the design we got had the tendency to strangle the weak CPU as it was.

See above. Plus, the weak CPU could have GAINED advantages if the system used the packed graphic.
Quote:

Sure with hindsight we could go back and decide that the system would only finished in a way and at a time where shipping it with 256/512k was a non issue, which would have made it sensible to give the chipset it's own separate display buffer (read like a CGA/EGA/VGA card), but none of that was an option when the started.

And nothing of this was required.

The hardware of the "packed" Amiga which I've proposed is exactly like the regular Amiga hardware with the ONLY DIFFERENCE represented by the usage of packed graphic instead of the planar one. Even the Blitter is exactly the same (and being able to do exactly the same thing), but with one change.

The problem, again, is that you have NOT read the articles.
Quote:

It is the same as looking back at (less than) 1MB of binary for Kickstart+WB in the early version and coming to the conclusion that a half talented IT student could do a better job as a sumer project. With today's tools and resources? Maybe. With what was available back then? Doubtful.

Read the articles and you'll see the answers to both questions is certainly yes.


@babsimov

Quote:

babsimov wrote: @Kronos A few years ago I had read or listened to an interview with Jay Miner in which he explained that one of his regrets about the Amiga was not having chosen the chunky display directly. Unfortunately, I searched for this, but I can't find this interview, maybe it's no longer online or maybe one day I'll come across it by chance. That's why I've been waiting a long time for someone to give us some insight into what a chunky Amiga chipset could have been and what it might or might not have done better or not. Here an article about the history of Amiga (french) : http://obligement.free.fr/articles/amiga_histoire_1983.php At a moment they talk about planar or chunky choice : "There was thus a debate on the graphics modes that the new machine should use. Chunky pixel mode would be more suitable for 3D games that Jay Miner and RJ Mical had in mind, while planar mode would be more suitable for producing 2D displays and GUI. Ultimately, planar mode was chosen to make designing a multi-window operating system and 2D features much faster and easier, and because planar bit-planes only required a small amount of memory, so that at the time, memory was still very expensive."

Please pay attention to the part that I've highlighted, because it's purely and entirely wrong: it's the exact opposite!


@babsimov

Quote:

babsimov wrote: @cdimauro By the way, would it be possible to create a simulator of this Amiga Chunky chipset from WinUAE sources so that you can create a game demo for us using the advantages mentioned, such as all the sprites usable in AGA, the different dual play field modes, EHB 128 and 256 color mode. Or the 15/16 bit mode for AGA graphics, the 256/256 couleurs dual playfield AGA. In short, to see in "real" what all this could have given?

Yes and doesn't require a lot of changes (but WinUAE source code could be complex. Never took a look at it).

Some existing Amiga o.s.-friendly (which doesn't assume planar graphics and/or directly accesses the bitmaps) software could run without problems as well.

More software could work by implementing an RTG driver for it.

However 15/16/32-bit Dual-Playfield modes will not be usable (no software supports them) and probably the extended indexed-mode (e.g.: more than 16 colors per playfield) couldn't be used as well.


@kolla

Quote:

kolla wrote: @babsimov “I am not a hardware developer” (Many have asked the same before)

Please, read carefully. Quoting again the guy:

"would it be possible to create a simulator of this Amiga Chunky chipset from WinUAE sources"

WinUAE is a SOFTWARE emulator...


@Hypex

Quote:

Hypex wrote: @cdimauro Quote: I've no time to reply to comments, because in the last period I was very busy writing article articles on my technical blog plus I've started again working on my architecture . Excellent. I've been waiting to read this. What's your architecture? Oh sorry, that implies a comment.

LOL

Already talked about it here, some time ago. But I had some other ideas to improve it and I'm working on the peephole analyzer to implement some patterns to improve the code density and/or number of executed instructions. Current results look promising.
Quote:

Quote: However I've written another 16 articles about the packed vs planar graphics to further clarify the topic. Just reporting now what I've written on LinkedIn. Wow. Another 16. Maybe you could release a book. Amiga: The Packed vs Planar Wars.

That's what I've stated: you can think at the articles like the chapters of a book, because I've written A LOT of stuff.

@babsimov

Quote:

A few years ago I had read or listened to an interview with Jay Miner in which he explained that one of his regrets about the Amiga was not having chosen the chunky display directly.

I know the one. It's a popular one that would be in Amiga Forever. Someone posted a YouTube link but I can't find it. I looked through every video here in the whole thread and didn't see it.

What I do recall is Jay being asked if he would do anything differently. He brought up the bitplanes and made a comment about using pixel graphics instead. I can remember him caliing it that. Which in modern terms or 90's terms even would be called packed graphics. Known as chunky to Doomsters.

Quote:

That's why I've been waiting a long time for someone to give us some insight into what a chunky Amiga chipset could have been and what it might or might not have done better or not.

I did some experiments with this using Doom engine. With the limited scope of my hardware. What I found is that if a packed mode is assumed without any pixel block conversion then it is held back by chip ram speeds. Writing to chip ram a pixel byte at a time is slower than copying a frame to it with pixel 4 bytes or a long word at a time. The Amiga needed VRAM.

https://amigaworld.net/modules/newbb/viewtopic.php?topic_id=44651&forum=15

@cdimauro

Quote:

Quote: babsimov wrote: @Kronos Here an article about the history of Amiga (french) : http://obligement.free.fr/articles/amiga_histoire_1983.php At a moment they talk about planar or chunky choice : "There was thus a debate on the graphics modes that the new machine should use. Chunky pixel mode would be more suitable for 3D games that Jay Miner and RJ Mical had in mind, while planar mode would be more suitable for producing 2D displays and GUI. Ultimately, planar mode was chosen to make designing a multi-window operating system and 2D features much faster and easier, and because planar bit-planes only required a small amount of memory, so that at the time, memory was still very expensive." Please pay attention to the part that I've highlighted, because it's purely and entirely wrong: it's the exact opposite!

The quote from the article I chose was mostly to show that there was internal debate over what type of display to use. It was the planar that won, perhaps for a reason erroneous at the time. I feel like it was primarily a memory cost issue though. The analysis at the time was perhaps wrong. Jay Miner's later comments seem to say he should have gone straight for chunky.
In hindsight, chunky would have been a more sustainable choice, but at the time they must not have known that.


Quote:

Quote: babsimov wrote: @cdimauro By the way, would it be possible to create a simulator of this Amiga Chunky chipset from WinUAE sources so that you can create a game demo for us using the advantages mentioned, such as all the sprites usable in AGA, the different dual play field modes, EHB 128 and 256 color mode. Or the 15/16 bit mode for AGA graphics, the 256/256 couleurs dual playfield AGA. In short, to see in "real" what all this could have given? Yes and doesn't require a lot of changes (but WinUAE source code could be complex. Never took a look at it). Some existing Amiga o.s.-friendly (which doesn't assume planar graphics and/or directly accesses the bitmaps) software could run without problems as well. More software could work by implementing an RTG driver for it. However 15/16/32-bit Dual-Playfield modes will not be usable (no software supports them) and probably the extended indexed-mode (e.g.: more than 16 colors per playfield) couldn't be used as well.

What I envisaged was rather a WinUAE which would precisely make it possible to manage the 15/16/32 bit mode of the dual play field. And therefore to write a specific game demo to use these functions of this "WinUAE OCS/AGA chunky". But NutsAboutAmiga explains that it is not necessary to rewrite WinUAE to see what it could have given. It would just be enough to make an animation or video of a game as it could have been with these possibilities of your hypothetical Amiga Chunky chipset. I would be very happy to see that one day to better visualize, if one day you find the time to do that.


Anyway, I'm already very happy to have been able to read your articles on this subject and to already have some answers on what it could have brought in relation to the planar. It was unexpected to see someone take the time to do that. Thanks again.

@Hypex

Quote:

Quote: Hypex wrote: @babsimov [quote]A few years ago I had read or listened to an interview with Jay Miner in which he explained that one of his regrets about the Amiga was not having chosen the chunky display directly. I know the one. It's a popular one that would be in Amiga Forever. Someone posted a YouTube link but I can't find it. I looked through every video here in the whole thread and didn't see it. What I do recall is Jay being asked if he would do anything differently. He brought up the bitplanes and made a comment about using pixel graphics instead. I can remember him caliing it that. Which in modern terms or 90's terms even would be called packed graphics. Known as chunky to Doomsters.

I think you're right about the video. It must be that one. I don't remember everything, but what I remembered was that Jay Miner seemed to "regret" not having chosen the chunky directly. The interview was to date from the release of AGA and there we could clearly see that the lack of chunky posed a problem for 3D games which were becoming fashionable.

Quote:

Quote: That's why I've been waiting a long time for someone to give us some insight into what a chunky Amiga chipset could have been and what it might or might not have done better or not. I did some experiments with this using Doom engine. With the limited scope of my hardware. What I found is that if a packed mode is assumed without any pixel block conversion then it is held back by chip ram speeds. Writing to chip ram a pixel byte at a time is slower than copying a frame to it with pixel 4 bytes or a long word at a time. The Amiga needed VRAM. https://amigaworld.net/modules/newbb/viewtopic.php?topic_id=44651&forum=15

Thank you for the link. For 3D games, the chunky was an advantage. For the VRAM I don't know, I'm not technical enough to judge. Jay Miner was planning VRAM I believe. But already with standard RAM the Chunky chipset presented to us seems to do excellent things. Much better than what we experienced in reality both for OCS and even more for AGA

@babsimov

Amiga main advantage was that it enable the graphics to be synchronized to the vertical sync, the ability to accurately place something at the screen at the correct time, Allowed Amiga graphics, in some cases to get away with single buffering, or partially buffering the screen, PC on the other hand did not have this in the start. Cooper as it was is essential to managing the rendering of the display.

However WinUAE is displays everything on a 16 or 32 bit screen, and has to converge everything to PC format, There's a lot of instructions required to emulate the old graphics, it does take a lot of time to render it on the screen, it's certainly eats away at the available CPU cycles on the lower end system, It has to deal with the clock cycles rising edges falling edges timing delays, the 16 bit data bus of the Amiga, these are restraints, not necessarily for rendering the same type of graphics, accurate emulation it's only required when you're actually running amiga software.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

@Hypex

Quote:

Hypex wrote: @babsimov Quote: A few years ago I had read or listened to an interview with Jay Miner in which he explained that one of his regrets about the Amiga was not having chosen the chunky display directly. I know the one. It's a popular one that would be in Amiga Forever. Someone posted a YouTube link but I can't find it. I looked through every video here in the whole thread and didn't see it. What I do recall is Jay being asked if he would do anything differently. He brought up the bitplanes and made a comment about using pixel graphics instead. I can remember him caliing it that. Which in modern terms or 90's terms even would be called packed graphics. Known as chunky to Doomsters.

FROM: Port AmigaOS 4 to x86
Quote:

Quote: That's why I've been waiting a long time for someone to give us some insight into what a chunky Amiga chipset could have been and what it might or might not have done better or not. I did some experiments with this using Doom engine. With the limited scope of my hardware. What I found is that if a packed mode is assumed without any pixel block conversion then it is held back by chip ram speeds. Writing to chip ram a pixel byte at a time is slower than copying a frame to it with pixel 4 bytes or a long word at a time. The Amiga needed VRAM. https://amigaworld.net/modules/newbb/viewtopic.php?topic_id=44651&forum=15

But that's different from the packed vs planar debate.


@babsimov

Quote:

babsimov wrote: Quote: cdimauro wrote: Please pay attention to the part that I've highlighted, because it's purely and entirely wrong: it's the exact opposite! The quote from the article I chose was mostly to show that there was internal debate over what type of display to use. It was the planar that won, perhaps for a reason erroneous at the time. I feel like it was primarily a memory cost issue though. The analysis at the time was perhaps wrong. Jay Miner's later comments seem to say he should have gone straight for chunky. In hindsight, chunky would have been a more sustainable choice, but at the time they must not have known that.

IMO they haven't paid enough attention or lacked out-of-order thinking to better evaluate ALL pros and cons of both graphical formats.
Quote:

babsimov wrote: Quote: cdimauro wrote: Yes and doesn't require a lot of changes (but WinUAE source code could be complex. Never took a look at it). Some existing Amiga o.s.-friendly (which doesn't assume planar graphics and/or directly accesses the bitmaps) software could run without problems as well. More software could work by implementing an RTG driver for it. However 15/16/32-bit Dual-Playfield modes will not be usable (no software supports them) and probably the extended indexed-mode (e.g.: more than 16 colors per playfield) couldn't be used as well. What I envisaged was rather a WinUAE which would precisely make it possible to manage the 15/16/32 bit mode of the dual play field. And therefore to write a specific game demo to use these functions of this "WinUAE OCS/AGA chunky". But NutsAboutAmiga explains that it is not necessary to rewrite WinUAE to see what it could have given. It would just be enough to make an animation or video of a game as it could have been with these possibilities of your hypothetical Amiga Chunky chipset. I would be very happy to see that one day to better visualize, if one day you find the time to do that.

As I've said, I don't know how much complex WinUAE is. It's definitely possible and using packed graphics a lot of code will also be simplified, but there's no idea of how long it could take.

Besides that, I don't see it useful, because it's a platform with little software but, and that's worse, nobody would develop software for it because it isn't a "real" system.
Quote:

Anyway, I'm already very happy to have been able to read your articles on this subject and to already have some answers on what it could have brought in relation to the planar. It was unexpected to see someone take the time to do that. Thanks again.

Quote:

@Hypex Quote: I did some experiments with this using Doom engine. With the limited scope of my hardware. What I found is that if a packed mode is assumed without any pixel block conversion then it is held back by chip ram speeds. Writing to chip ram a pixel byte at a time is slower than copying a frame to it with pixel 4 bytes or a long word at a time. The Amiga needed VRAM. https://amigaworld.net/modules/newbb/viewtopic.php?topic_id=44651&forum=15 Thank you for the link. For 3D games, the chunky was an advantage.

Let's say this: packed graphics shines with 3D games, but it's also better than planar even with 2D games.
Quote:

For the VRAM I don't know, I'm not technical enough to judge. Jay Miner was planning VRAM I believe.

It would have been another mistake. As I've proved on the theoretical article, the higher is the bus width (hence, the Chip/VRAM access granularity) the worse performs planar graphics...
Quote:

But already with standard RAM the Chunky chipset presented to us seems to do excellent things. Much better than what we experienced in reality both for OCS and even more for AGA

Indeed. Amiga gave a great time to all us.

@NutsAboutAmiga

Quote:

NutsAboutAmiga wrote: @babsimov Amiga main advantage was that it enable the graphics to be synchronized to the vertical sync, the ability to accurately place something at the screen at the correct time, Allowed Amiga graphics, in some cases to get away with single buffering, or partially buffering the screen, PC on the other hand did not have this in the start. Cooper as it was is essential to managing the rendering of the display. However WinUAE is displays everything on a 16 or 32 bit screen, and has to converge everything to PC format, There's a lot of instructions required to emulate the old graphics, it does take a lot of time to render it on the screen, it's certainly eats away at the available CPU cycles on the lower end system, It has to deal with the clock cycles rising edges falling edges timing delays, the 16 bit data bus of the Amiga, these are restraints, not necessarily for rendering the same type of graphics, accurate emulation it's only required when you're actually running amiga software.

Thank you for these explanations. I had indeed read over time that the Copper was one of the most interesting coprocessors of the Amiga, even one of its major assets.


@cdimauro

Quote:

@babsimov[quote] cdimauro wrote: IMO they haven't paid enough attention or lacked out-of-order thinking to better evaluate ALL pros and cons of both graphical formats.

Possible, but to their credit, they designed a completely new architecture for the time, the first multimedia machine before the term even existed. There were already enough things to create (blitter, copper, memory controller (notion of chip and fast), dual play field mode, HAM mode, sound component). We can forgive them for not having been at the end of the line for the choice of display mode. Especially since all these tests had to cost them money and Hi-Toro didn't have unlimited time and resources. The Amiga chipset as we know it was already ahead of the others of the time. It was later that Commodore failed to make it evolve properly and quickly enough.

Quote:

As I've said, I don't know how much complex WinUAE is. It's definitely possible and using packed graphics a lot of code will also be simplified, but there's no idea of how long it could take. Besides that, I don't see it useful, because it's a platform with little software but, and that's worse, nobody would develop software for it because it isn't a "real" system.

Modifying WinUAE was the extreme assumption. The solution proposed by NutsAboutAmiga (in bold) would make it easier to visually see what it could have looked like. It's just something that would be very visual for someone with technical knowledge like me for example. Even if the summary at the end of your articles already gives a theoretical idea of ​​the contribution of a native chunky Amiga chipset.

Quote:

It would have been another mistake. As I've proved on the theoretical article, the higher is the bus width (hence, the Chip/VRAM access granularity) the worse performs planar graphics....

We can't really blame Jay Miner for wanting to use a new type of more powerful memory to evolve the Amiga chipset. He was an innovator and it is logical that he wanted to use this new technology. But the cost of these RAM deterred Commodore from following through.

Quote:

Indeed. Amiga gave a great time to all us.

I agreed

@babsimov

It would look exactly like what we have with planar display, the question is whether it would be any faster.

Last edited by kolla on 23-May-2023 at 05:25 PM.

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@babsimov

Quote:

babsimov wrote: @cdimauro Quote: @babsimov[quote] cdimauro wrote: IMO they haven't paid enough attention or lacked out-of-order thinking to better evaluate ALL pros and cons of both graphical formats. Possible, but to their credit, they designed a completely new architecture for the time, the first multimedia machine before the term even existed.

"Multimedia" was already present, but on workstations. Amiga had the merit to have delivered it to the masses.
Quote:

There were already enough things to create (blitter, copper, memory controller (notion of chip and fast), dual play field mode, HAM mode, sound component). We can forgive them for not having been at the end of the line for the choice of display mode. Especially since all these tests had to cost them money and Hi-Toro didn't have unlimited time and resources.

Sure, but this doesn't mean that they were bounded to the planar format: packed was a legit choice as well.
Quote:

The Amiga chipset as we know it was already ahead of the others of the time.

Nothing to say about that.
Quote:

It was later that Commodore failed to make it evolve properly and quickly enough.

Yes, to a greater extent.

However the bad design decisions with the first chipset crippled the evolution. You can clearly see it with the AGA, where the planar graphics has exacerbated the issues of this format.
Quote:

Quote: As I've said, I don't know how much complex WinUAE is. It's definitely possible and using packed graphics a lot of code will also be simplified, but there's no idea of how long it could take. Besides that, I don't see it useful, because it's a platform with little software but, and that's worse, nobody would develop software for it because it isn't a "real" system. Modifying WinUAE was the extreme assumption. The solution proposed by NutsAboutAmiga (in bold) would make it easier to visually see what it could have looked like. It's just something that would be very visual for someone with technical knowledge like me for example.

If it's only this then any graphic artist could do it.
Quote:

Even if the summary at the end of your articles already gives a theoretical idea of ​​the contribution of a native chunky Amiga chipset.

Quote:

Quote: It would have been another mistake. As I've proved on the theoretical article, the higher is the bus width (hence, the Chip/VRAM access granularity) the worse performs planar graphics.... We can't really blame Jay Miner for wanting to use a new type of more powerful memory to evolve the Amiga chipset. He was an innovator and it is logical that he wanted to use this new technology. But the cost of these RAM deterred Commodore from following through.

That's ok because having more bandwidth is a natural requirement when talking about graphics. So, adopting a more advanced memory like VRAM was a wise decision (IF the cost wasn't too high).

However with planar graphics it would have ended-up with the same issues that later were shown with the AGA chipset (as reported on my articles).

As you can see, the "original sin" is always the same: having chosen the planar graphics instead of the packed one.


@kolla

Quote:

kolla wrote: @babsimov It would look exactly like what we have with planar display,

The point is that is would have been impossible to do the same with the planar graphics, as it's reported on the last of my articles.
Quote:

the question is whether it would be any faster.

With the same color depth? Yes, on average.

With different color depths (more for packed and less for the planar) it depends on the specific workload.

@babsimov

Quote:

I think you're right about the video. It must be that one. I don't remember everything, but what I remembered was that Jay Miner seemed to "regret" not having chosen the chunky directly. The interview was to date from the release of AGA and there we could clearly see that the lack of chunky posed a problem for 3D games which were becoming fashionable.

I find it hard to imagine as planar is so entrenched in the Amiga chipset design. In a manner of speaking planar is Amiga. I suppose it's one of those core design ideas where the method is first chosen and then hardware designed to implement that method so it would be chosen before or at the drafting stage.

But the decision can be strange as one thing Jay was intent upon was replicating a flight simulator. F18 Interceptor does okay with 3d filled polygons. But it also looks slow and demanding pushing the chipset to render in hardware 3d. HAM was the door to realistic imagery but HAM with planar meant it was only practical for stills. Still cool in the 80's but not as much as live and fluent HAM motion would have been.

Quote:

Thank you for the link. For 3D games, the chunky was an advantage. For the VRAM I don't know, I'm not technical enough to judge. Jay Miner was planning VRAM I believe. But already with standard RAM the Chunky chipset presented to us seems to do excellent things. Much better than what we experienced in reality both for OCS and even more for AGA

Yes Jay did want VRAM. With Ranger IIRC. As an example what could have been possible is hardware like Graffiti that implements an external chunky mode based on bitplanes.

@cdimauro

Quote:

FROM: Port AmigaOS 4 to x86

That's it! Thanks. I knew it was in some big thread.

Quote:

But that's different from the packed vs planar debate.

Yes but it is inspired by it. It's hard to know how the Amiga would have differed had it had pixel and not planar graphics, with regards to chip ram, and other hardware speeds. Being it could have read pixels direct in serial fashion, without a parallel to serial CLUT conversion, it should have been faster but I don't know exactly since it wasn't configured that way in hardware.

Also, I did recall one thing, writing a pixel byte to chip ram was slower than writing four bytes at once. So chip ram lacking cache was perhaps more of a strain than needing faster VRAM.

On top of this, I think one thing that incurred an extra planar penalty, was that the Amiga lacked planar in the PC way. The PC planar way allowed pixel values to be written that the hardware converted to planar. And had features so it could write multiple columns at once. After a comparison I think it needed these extra speed ups and programming tricks to be more practical.

@kolla

Quote:

kolla wrote: @babsimov It would look exactly like what we have with planar display, the question is whether it would be any faster.

If I understood everything correctly, it seems that it is.

@cdimauro

Quote:

"Multimedia" was already present, but on workstations. Amiga had the merit to have delivered it to the masses..

Your right

Quote:

Quote: There were already enough things to create (blitter, copper, memory controller (notion of chip and fast), dual play field mode, HAM mode, sound component). We can forgive them for not having been at the end of the line for the choice of display mode. Especially since all these tests had to cost them money and Hi-Toro didn't have unlimited time and resources. Sure, but this doesn't mean that they were bounded to the planar format: packed was a legit choice as well.

I'm not saying they were right, just that they had so many choices to make, that at some point they made a decision and stuck with it hoping they had made the wisest choice. With hindsight, Jay Miner will say that he may have had to choose the chunky directly, which shows that he probably saw that he had made a questionable choice.


Quote:

Quote: The Amiga chipset as we know it was already ahead of the others of the time. Nothing to say about that. Quote: It was later that Commodore failed to make it evolve properly and quickly enough. Yes, to a greater extent.

To their defence, I find that the Ranger chipset proposed by Jay Miner was perhaps a little better with the VRAM, but in fact it did not provide enough. Not even a 256 color mode in 1988, it's disappointing from my point of view. Besides it seems that in fact the ECS and the AGA are in fact the Ranger without the management of the VRAM
https://retrocomputing.stackexchange.com/questions/2706/what-exactly-was-the-amiga-ranger/2707#2707
http://eab.abime.net/showthread.php?t=77091
https://eab.abime.net/showthread.php?t=86674&page=55

Moreover, in Brian Bagnall's recent books, the Ranger is mentioned and it does not have the 128 color mode that is often mentioned. The 7th plan is in fact a monochrome plan of 1024x1024 which functions in parallel with the displays of the OCS. It was to have something in front of the monochrome mode of the Atari ST. But obligation of VRAM to display that.


Quote:

However the bad design decisions with the first chipset crippled the evolution. You can clearly see it with the AGA, where the planar graphics has exacerbated the issues of this format.

Yes the lack of chunky was a big flaw in the AGA.

Commodore planned to add it in the AA+ and moreover it is what should have been released in 1990 instead of the ECS.

The thesis according to which the AGA is the Ranger without the VRAM seems to me coherent in passing, since roughly the AGA is at the level of the VGA of 87. But in 1992 it is obsolete compared to what exists on PC , even low end. The Ranger planned release date was 1988.


Quote:

Quote: Modifying WinUAE was the extreme assumption. The solution proposed by NutsAboutAmiga (in bold) would make it easier to visually see what it could have looked like. It's just something that would be very visual for someone with technical knowledge like me for example. If it's only this then any graphic artist could do it.

I'm not an artist :(
I hope one day an artist show us that.

Quote:

Quote: It would have been another mistake. As I've proved on the theoretical article, the higher is the bus width (hence, the Chip/VRAM access granularity) the worse performs planar graphics.... We can't really blame Jay Miner for wanting to use a new type of more powerful memory to evolve the Amiga chipset. He was an innovator and it is logical that he wanted to use this new technology. But the cost of these RAM deterred Commodore from following through.

Quote:

That's ok because having more bandwidth is a natural requirement when talking about graphics. So, adopting a more advanced memory like VRAM was a wise decision (IF the cost wasn't too high). However with planar graphics it would have ended-up with the same issues that later were shown with the AGA chipset (as reported on my articles). As you can see, the "original sin" is always the same: having chosen the planar graphics instead of the packed one.

All we need now is a Delorean with you inside to go back to 1983 and tell it to Jay Miner and his team


@Hypex

Quote:

Quote: I think you're right about the video. It must be that one. I don't remember everything, but what I remembered was that Jay Miner seemed to "regret" not having chosen the chunky directly. The interview was to date from the release of AGA and there we could clearly see that the lack of chunky posed a problem for 3D games which were becoming fashionable. Quote: I find it hard to imagine as planar is so entrenched in the Amiga chipset design. In a manner of speaking planar is Amiga. I suppose it's one of those core design ideas where the method is first chosen and then hardware designed to implement that method so it would be chosen before or at the drafting stage. But the decision can be strange as one thing Jay was intent upon was replicating a flight simulator. F18 Interceptor does okay with 3d filled polygons. But it also looks slow and demanding pushing the chipset to render in hardware 3d. HAM was the door to realistic imagery but HAM with planar meant it was only practical for stills. Still cool in the 80's but not as much as live and fluent HAM motion would have been.

If I remember correctly from what I was able to read and hear, 3D and simulators were what Jay Miner and RJ Mical wanted, but the others and especially Dave Morse had asked for 2D. 3D was not fashionable in 1983 and particularly lousy compared to 2D games. 3D took off when it could be textured and fast and therefore look a little prettier. They made a compromise. F18 interceptor for the time was really very good. It was one of the first games I bought for my Amiga that I just got in 1988.


Quote:

Quote: Thank you for the link. For 3D games, the chunky was an advantage. For the VRAM I don't know, I'm not technical enough to judge. Jay Miner was planning VRAM I believe. But already with standard RAM the Chunky chipset presented to us seems to do excellent things. Much better than what we experienced in reality both for OCS and even more for AGA Yes Jay did want VRAM. With Ranger IIRC. As an example what could have been possible is hardware like Graffiti that implements an external chunky mode based on bitplanes.

Some time ago I came across some youtube video of graphiti. It's a plus, but it doesn't replace a native chunky mode from what I've seen and read here or there.

Last edited by babsimov on 24-May-2023 at 10:03 PM.

@pixie

Quote:

pixie wrote: @Hammer Quote: For Quake, it's unlikely Pistorm32-Emu68-RPI CM4 on A1200's AGA would reach 60 fps 320x200/240/245 resolution. Indeed, it got 69.2 FPS on RTG, I guess it wouldn't.

https://www.youtube.com/watch?v=7_X1sfYC1GQ

On A1200's AGA 256 color mode with PiStorm32 Lite with RPi 4B, Quake demo3 reached 42 fps.

Since the Amiga chipsets are the bottleneck, A1200's AGA is twice as fast when compared to A500's OCS/ECS.

With enough CPU power, AGA has beaten many SVGA chipsets from https://thandor.net/benchmark/33

ET4000AX ISA with Pentium 100Mhz reached 22 fps. Faster SVGA PCI cards can exceed ET4000AX's 22 fps with the same Pentium 100 Mhz CPU.

For Quake and with enough 68K CPU power, Amiga AGA can scale like Pentium 200 with a PCI SVGA graphics card.

Last edited by Hammer on 31-May-2023 at 12:09 PM.
Last edited by Hammer on 31-May-2023 at 12:06 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@babsimov

The high-resolution high-color display is a distraction for gaming. The focus should be on the RISC-based math coprocessor with reasonable 2D graphics capability.

High-resolution high-color display capability is useful for workstation desktop OS.

The Play Station prototype was operational in 1993 and R&D started a few years before 1993 i.e. The original Sony and Nintendo contract was signed in 1988.

Sony's Ohga chaired a meeting in June 1992, consisting of Kutaragi and several senior Sony board members. Kutaragi unveiled a proprietary CD-ROM-based system he had been secretly working on which played games with immersive 3D graphics. Kutaragi was confident that his LSI chip could accommodate one million logic gates, which exceeded the capabilities of Sony's semiconductor division at the time.

Intel focused on improved Pentium FPU and it was ready in 1993 which then set the stage for the floating point geometry driven 1996 Quake PC.

For the Amiga chipset evolution, the next step after line draw and color fill acceleration is 3D and Jay wanted 3D.


1987 IBM VGA is slow despite throwing K7 Athlon XP 2200+ (1800 Mhz) at it. https://www.youtube.com/watch?v=octArwHpaiY
Results of timedemo were:
360x480 - 3,2FPS
320x240 - 7FPS
320x200 - 8,6FPS
Tested on Athlon XP 2200+, Soltek SL75-KAV (Via KT133A), 512MB SDR CL3 and IBM VGA.


For playing Quake and enough CPU power, Amiga OCS HAM mode beats IBM VGA!

PiStorm-equipped (ARM Cortex A53 @ 1.4 Ghz JIT'ing emulating a fast 68040 CPU) Amiga 500's OCS can drive Quake in HAM mode faster than IBM VGA. https://www.youtube.com/watch?v=btoU_CQSg7A

Last edited by Hammer on 31-May-2023 at 05:03 PM.
Last edited by Hammer on 31-May-2023 at 04:57 PM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@Hammer

Commodore bankrupt because AGA has not chunky pixel.
c2p was too slow on 68020. nobody care how fast is c2p on faster cpu than 020.

@Hypex

Quote:

Hypex wrote: @babsimov Quote: I think you're right about the video. It must be that one. I don't remember everything, but what I remembered was that Jay Miner seemed to "regret" not having chosen the chunky directly. The interview was to date from the release of AGA and there we could clearly see that the lack of chunky posed a problem for 3D games which were becoming fashionable. I find it hard to imagine as planar is so entrenched in the Amiga chipset design. In a manner of speaking planar is Amiga.

It would have been packed, if this graphic format was used, instead.
Quote:

I suppose it's one of those core design ideas where the method is first chosen and then hardware designed to implement that method so it would be chosen before or at the drafting stage.

Jay Miner has stated that planar was used because it was more efficient compared to packed. So, this decision has driven the design implementation.
Quote:

But the decision can be strange as one thing Jay was intent upon was replicating a flight simulator.

Hum. I don't recall this statement, but if it's true then it's even worse: planar is definitely NOT suitable for 3D graphics.
Quote:

F18 Interceptor does okay with 3d filled polygons. But it also looks slow and demanding pushing the chipset to render in hardware 3d.

Indeed. Packed, at the same color depth, completely destroys planar graphics in terms of performance for rendering 3D stuff.

16 colors was common for 3D graphics (due to the limited hardware) and 4-bits packed graphic would have been a perfect suit.
Quote:

HAM was the door to realistic imagery but HAM with planar meant it was only practical for stills. Still cool in the 80's but not as much as live and fluent HAM motion would have been.

How much it was used the HAM mode? Very very little. It was nice for static graphics, for sure, but for common user experience wasn't much useful.

BTW and from what I recall, Jay Miner wanted to remove it from the design, but it was late and it would have left a big hole on the chip, so they decided to keep it. That to show that it wasn't even considered useful from the team.
Quote:

Hypex wrote: @cdimauro Quote: But that's different from the packed vs planar debate. Yes but it is inspired by it. It's hard to know how the Amiga would have differed had it had pixel and not planar graphics, with regards to chip ram, and other hardware speeds.

"Just" read all my articles and you'll get a very close idea of how it would have been. However you need to allocate some time for it, because it's like reading a book (I've written a lot of stuff).
Quote:

Being it could have read pixels direct in serial fashion, without a parallel to serial CLUT conversion,

Exactly. The graphic was already... ready for accessing the CLUT.
Quote:

it should have been faster but I don't know exactly since it wasn't configured that way in hardware.

WinUAE could be modified to test it.
Quote:

Also, I did recall one thing, writing a pixel byte to chip ram was slower than writing four bytes at once. So chip ram lacking cache was perhaps more of a strain than needing faster VRAM.

Then imagine how much worse was it with planar graphics...
Quote:

On top of this, I think one thing that incurred an extra planar penalty, was that the Amiga lacked planar in the PC way. The PC planar way allowed pixel values to be written that the hardware converted to planar. And had features so it could write multiple columns at once. After a comparison I think it needed these extra speed ups and programming tricks to be more practical.

Yes, it's an advantage in this case, but it's worse if you need to access single pixel's data, since you have to change the bitplanes mask using the infamous I/O ports.

That's why you've to "split" graphic accesses per single planes, to reduce (to 4 time) the masks programming.

@ppcamiga1

Quote:

ppcamiga1 wrote: @Hammer Commodore bankrupt because AGA has not chunky pixel. c2p was too slow on 68020. nobody care how fast is c2p on faster cpu than 020.

Nope.

Chunky Pixels on A1200's 68EC020 without Fast RAM are still slow.

Throwing AAA chipset with 68EC020 @ 14 Mhz without Fast RAM is nearly pointless.

ET4000AX 16-bit ISA bus with 68EC020 @ 14 Mhz with 16-bit system RAM is nearly pointless. Hint: pointless Atari Falcon.

For late 1993 Doom, both AGA and ET4000AX need the full 32-bit 68030 / 80386 DX equipped with CPU-only system ram at 40 to 50 Mhz.

Commodore UK wanted A1200 SKU with bundled CPU accelerator and Commodore International rejected it i.e. protecting Commodore Germany's A4000 existence.


Both 68020 and 68030 have similar IPC (instructions per clock).

The 1-bit plane can be considered to be a chunky pixel.

Last edited by Hammer on 01-Jun-2023 at 08:19 AM.
Last edited by Hammer on 01-Jun-2023 at 08:14 AM.
Last edited by Hammer on 01-Jun-2023 at 08:11 AM.
Last edited by Hammer on 01-Jun-2023 at 08:10 AM.
Last edited by Hammer on 01-Jun-2023 at 08:06 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@babsimov

Quote:

babsimov wrote: @cdimauro Quote: Sure, but this doesn't mean that they were bounded to the planar format: packed was a legit choice as well. I'm not saying they were right, just that they had so many choices to make, that at some point they made a decision and stuck with it hoping they had made the wisest choice. With hindsight, Jay Miner will say that he may have had to choose the chunky directly, which shows that he probably saw that he had made a questionable choice.

I don't agree. Jay clearly stated (in the above interview) that planar was chosen because it was faster.

So, even after so many years and with "hindsight", it continued to state the same thing. Which is completely wrong, as it was proven.
Quote:

Quote: Yes, to a greater extent. To their defence, I find that the Ranger chipset proposed by Jay Miner was perhaps a little better with the VRAM, but in fact it did not provide enough. Not even a 256 color mode in 1988, it's disappointing from my point of view.

Exactly. And still planar. Again: no sign of "hindsight".
Quote:

Besides it seems that in fact the ECS and the AGA are in fact the Ranger without the management of the VRAM https://retrocomputing.stackexchange.com/questions/2706/what-exactly-was-the-amiga-ranger/2707#2707 http://eab.abime.net/showthread.php?t=77091 https://eab.abime.net/showthread.php?t=86674&page=55

I don't agree to this history rewriting: ECS is a very simple OCS improvement and has nothing to do with Ranger.
Quote:

Moreover, in Brian Bagnall's recent books, the Ranger is mentioned and it does not have the 128 color mode that is often mentioned. The 7th plan is in fact a monochrome plan of 1024x1024 which functions in parallel with the displays of the OCS. It was to have something in front of the monochrome mode of the Atari ST. But obligation of VRAM to display that.

A monochromatic screen with a different resolution compared to the one displayed on the original one is a complete non-sense.

Unless the output is redirected to another display.
Quote:

Quote: However the bad design decisions with the first chipset crippled the evolution. You can clearly see it with the AGA, where the planar graphics has exacerbated the issues of this format. Yes the lack of chunky was a big flaw in the AGA. Commodore planned to add it in the AA+ and moreover it is what should have been released in 1990 instead of the ECS. The thesis according to which the AGA is the Ranger without the VRAM seems to me coherent in passing, since roughly the AGA is at the level of the VGA of 87.

AGA is much better than VGA, except for the lack of packed graphics.
Quote:

But in 1992 it is obsolete compared to what exists on PC , even low end. The Ranger planned release date was 1988.

Absolutely.
Quote:

Quote: However with planar graphics it would have ended-up with the same issues that later were shown with the AGA chipset (as reported on my articles). As you can see, the "original sin" is always the same: having chosen the planar graphics instead of the packed one. All we need now is a Delorean with you inside to go back to 1983 and tell it to Jay Miner and his team

Quote:

@Hypex Quote: But the decision can be strange as one thing Jay was intent upon was replicating a flight simulator. F18 Interceptor does okay with 3d filled polygons. But it also looks slow and demanding pushing the chipset to render in hardware 3d. HAM was the door to realistic imagery but HAM with planar meant it was only practical for stills. Still cool in the 80's but not as much as live and fluent HAM motion would have been. If I remember correctly from what I was able to read and hear, 3D and simulators were what Jay Miner and RJ Mical wanted, but the others and especially Dave Morse had asked for 2D. 3D was not fashionable in 1983 and particularly lousy compared to 2D games.

Nevertheless, there were several 3D games at the time, even for underpowered home computers.

But at least the ones with packed graphic had less issues for rendering it.
Quote:

3D took off when it could be textured and fast and therefore look a little prettier. They made a compromise. F18 interceptor for the time was really very good. It was one of the first games I bought for my Amiga that I just got in 1988.

Indeed. At the time 3D was already used on some cool games, and F18 was amazing.

@babsimov

FYI, Commodore created two 256-color chipsets i.e.
1. C65's chipset was completed in December 1990.

256 colors (8 bitplanes) with a 4096 color palette on A500 level chip ram bandwidth.

2. AGA chipset was completed around Q1 1991. The Lisa chip has access to 32-bit bus DRAM with twice the clock speed. The Alice is effectively Agnus ECS 16-bit.

The company's R&D resources were split between AGA / AAA and C65 groups.

Blame Mehdi Ali / Bill Sydnes (IBM PC Jr fame) for AGA's delay. Bill Sydnes directed R&D resources towards "Amiga Jr" i.e. ECS Amigas with 68020/68030 CPUs.

Blame Mehdi Ali / Bill Sydnes for canceling A500 for A600 i.e. large-scale revenue killer.

AAA project started around 1989.

Last edited by Hammer on 01-Jun-2023 at 08:56 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)

@babsimov

Quote:

babsimov wrote: @cdimauro By the way, would it be possible to create a simulator of this Amiga Chunky chipset from WinUAE sources so that you can create a game demo for us using the advantages mentioned, such as all the sprites usable in AGA, the different dual play field modes, EHB 128 and 256 color mode. Or the 15/16 bit mode for AGA graphics, the 256/256 couleurs dual playfield AGA. In short, to see in "real" what all this could have given?

Use Graffiti.

https://www.youtube.com/watch?v=yfBOmXOKnKU
Graffiti with 68030 @ 50 Mhz and 32-bit Fast RAM running Amiga's Doom.

FYI, Indivision AGA Mk3 includes the Graffiti feature. The Lisa chip didn't include a chunky pixel raster function.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68)