@cdimauro

Again, for the avoidance of doubt my suggestion was what would have been done to improve the situation for AGA as a platform which doesn't involve changing the chipset they already had designed. Adding more chip ram etc just wasn't an option in time for launch of what was already supposed to be a temporary solution. So the obvious answer is to include some fast ram as standard. This would not only free some of the limited chip ram that is otherwise used by the application code and non-media data but also anything the OS has taken away by then too for libraries and such.

Gunnar agrees later if you read on but he states this was solved because of ram expansions. However the cost to the user per MB for a trapdoor expansion in 1992 is a lot higher than it would've been for C= to install 1MB on the motherboard of all AGA machines. Simply knowing that all AGA machines have 1MB as fast ram as an absolute minimum and that you have approximately 4x the integer performance of the A500 as standard would definitely have created more opportunities for better games and other software in the early AGA years.

Even Fightin' Spirit. If zero chip ram were wasted for anything not for graphics or sound and even if nothing else visually were changed, you'd have been able to use slightly higher quality audio samples at least and it wouldn't sound like you were listening to domestic abuse down a long distance analogue telephone line.

Last edited by Karlos on 10-Oct-2022 at 11:25 AM.

_________________
Doing stupid things for fun...

@cdimauro

Quote:

Again? It took SEVEN years for the AGA, and it was crap!

No, Commodore did not work on AGA for 7 years.

The Amiga team worked on other ideas, AAA for example.
As the other project were delayed, not ready, cancelled, they cooked AGA "Spaghetti" for us in 10 minutes.

They not worked 7 years on AGA. They had limited time only.

If you have a limited time, then you need to focus.
You can only do "so many" features in this time.
And every feature that you try to do has a risk to fail - and the whole project might then fail.

AGA has significant improvements over OCS.
And they limited the features to the amount that they could do, and could debug.
This was a good decision.

Its better to have a working and released chipset with 5 new improvements,
than an non working non released with 10 new features.

@Gunnar

I don't have an axe to grind about AGA. The only misstep in my view was the lack of fast memory to get the most our of the CPU at the time of launch.

_________________
Doing stupid things for fun...

Karlos Quote:

I don't have an axe to grind about AGA. The only misstep in my view was the lack of fast memory to get the most our of the CPU at the time of launch.

Did Jay Miner fix that problem with Ranger?

https://en.wikipedia.org/wiki/Amiga_Ranger_Chipset

Let us say memory was $100 per MiB sometime in the '80s when Ranger would have been ready and VRAM was 20% more expensive.

2MiB DRAM chip mem = $200 mem cost
2MiB VRAM chip mem = $240 mem cost (Ranger)
2MiB DRAM chip mem + 1 MiB fast mem = $300 mem cost

Certainly 3MiB of total mem would be better if 2MiB wasn't enough for a particular program. There are several valuable advantages to the 2MiB VRAM though.

o dual ported memory could give full mem access to both the chipset and CPU at the same time.
o no data duplication and copying between mem types is necessary
o one continuous memory chunk
o no 64 bit mem fetch incompatibility or restrictions with double the number of 32 bit slots instead
o likely cheaper mem logic/controller cost for one type of mem

It would have been even better if more of the CPU slots could be lent to the chipset. If this was possible, then the higher mem bandwidth should have allowed higher resolutions with more colors. Maybe the Amiga 3000 wouldn't need more memory for the scan doubler/deinterlacer and productivity mode could have displayed full colors without slowdown on a cheaper (S)VGA monitor (no expensive multisync required). Sometimes being cheap costs more money in the long run.

I'm no expert on how a Ranger Amiga was going to use double the mem slots of VRAM. There seems to be limited information on Ranger despite Jay Miner saying he finished the specification and Dale Luck claiming to have a prototype. Dave Haynie thought Ranger was a myth or at best specs on paper only. It appears CBM hid Ranger away well for West Chester engineering to know practically nothing about what Los Gatos had worked on for maybe a year. CBM management was like thanks Jay, let me file the Ranger specs in the "Amiga is not the next C64 drawer". It would be interesting to know if Michele Battilana found them in the trove of CBM documents or whether they were destroyed by CBM amid the animosity with the Los Gatos developers.

Last edited by matthey on 10-Oct-2022 at 11:21 PM.
Last edited by matthey on 10-Oct-2022 at 10:25 PM.

@matthey

We'll probably never know.

_________________
Doing stupid things for fun...

@cdimauro

Quote:

cdimauro wrote: I've just reported what the Amiga Hardware Manual said about the specific question.

But the answer is misleading. What frequency the blitter circuit is 'clocked' at is irrelevant, what matters is how many 'ticks' it takes to perform an operation. Furthermore the results are constrained by memory cycle time, so no matter how fast it might go internally it is limited to 3.5MHz per memory operation.

Quote:

I'm assuming a very simple scenario which is also, more or less, the BEST case.... PAL (256 lines): Total slots 70512 Used slots 27866 Free slots 42646 Free slots (percentage) 60% AGA FMode 3 (64-bit fetch), 5 bitplanes, PAL (256 lines): Total slots 70512 Used slots 8666 Free slots 61846 Free slots (percentage) 88%... Since with AGA and 8 bitplanes you have only 83% (NTSC) and 82% (PAL) of slots available, you can see yourself that those aren't enough for 8 bitplanes = 256 colors, as you were suggesting. Whereas 7 bitplanes = 128 colors is much more realistic, albeit the numbers are still a little bit below of what is really needed.

Whether it's enough depends on what you need to do. With AGA you have 37% more slots in 256 colors than you had with OCS in 32 colors. Since you have more colors to spare you can do more tricks to cut down on blitting. A game that was limited to 32 colors in OCS could well be able to get 256 colors on AGA.

Even if Gunnar's figures are a bit 'optimistic' for a typical scenario, I agree with him that AGA was more than just a 'modest' improvement. 4 times higher bandwidth, sprites up to 4 times larger, 8 times more colors, 24 bit color palette vs 12, superhires HAM8 for photographic quality images. If this was a VGA card you wouldn't call the improvements 'modest'.

@Karlos

Quote:

Karlos wrote: Gunnar agrees later if you read on but he states this was solved because of ram expansions. However the cost to the user per MB for a trapdoor expansion in 1992 is a lot higher than it would've been for C= to install 1MB on the motherboard of all AGA machines.

However the initial purchase price with 1MB on the motherboard would be significantly higher. If the user installed an accelerator card that 1MB would become 'slow' RAM, causing yet more confusion in the memory map.

Quote:

Simply knowing that all AGA machines have 1MB as fast ram as an absolute minimum and that you have approximately 4x the integer performance of the A500 as standard would definitely have created more opportunities for better games and other software in the early AGA years.

But there weren't many 'early' AGA years, and better games were produced without Fast RAM. To do much better it really needed an accelerator card, which would make that on-board 1MB redundant. Better to lower the cost of entry and save for that more important purchase down the road.

Quote:

Even Fightin' Spirit. If zero chip ram were wasted for anything not for graphics or sound and even if nothing else visually were changed, you'd have been able to use slightly higher quality audio samples at least and it wouldn't sound like you were listening to domestic abuse down a long distance analogue telephone line.

A slight improvement not worth the extra expense.

Commodore obviously expected users to purchase RAM boards when they wanted more memory, like they did with earlier models. I believe the thinking was sound. Few would buy a RAM board with only 1MB, and I doubt many Amiga fans would be impressed by only getting a miserable 1MB on the motherboard. They would want a SIMM slot with 2 or 4 MB at least.

@Karlos

Quote:

Karlos wrote: @cdimauro Again, for the avoidance of doubt my suggestion was what would have been done to improve the situation for AGA as a platform which doesn't involve changing the chipset they already had designed. Adding more chip ram etc just wasn't an option in time for launch of what was already supposed to be a temporary solution. So the obvious answer is to include some fast ram as standard. This would not only free some of the limited chip ram that is otherwise used by the application code and non-media data but also anything the OS has taken away by then too for libraries and such. Gunnar agrees later if you read on but he states this was solved because of ram expansions. However the cost to the user per MB for a trapdoor expansion in 1992 is a lot higher than it would've been for C= to install 1MB on the motherboard of all AGA machines. Simply knowing that all AGA machines have 1MB as fast ram as an absolute minimum and that you have approximately 4x the integer performance of the A500 as standard would definitely have created more opportunities for better games and other software in the early AGA years. Even Fightin' Spirit. If zero chip ram were wasted for anything not for graphics or sound and even if nothing else visually were changed, you'd have been able to use slightly higher quality audio samples at least and it wouldn't sound like you were listening to domestic abuse down a long distance analogue telephone line.

And I want to clarify as well that I've nothing against having added extra memory to an Amiga, but with 3 conditions:

- it should be provided with all low-cost machines;

- it should be of the proper type (so, NOT the stupid Slow ram);

- the size should be defined according to the single assets.

To me and according to my experience 128kB of FAST mem on all Amigas (even OCS) should be ok. And good enough for the games for the time.

To be more clear: OCS/ECS with 1MB chip + 128kB fast. AGA 2MB chip + 128kB fast.

The second and third point are very important and I can show you why taking of Fightin' Spirit.

If 1MB of chip ram were standard for the Amigas then the parallax scroll of the bottom 64 lines of the screen was possible. I know it because I've already implemented, using entirely the Blitter to draw those lines.

This to show what difference can make the memory type (vs 512kB of chip + 512kB of Slow/Fast mem).

@Gunnar

Quote:

Gunnar wrote: @cdimauro Quote: Again? It took SEVEN years for the AGA, and it was crap! No, Commodore did not work on AGA for 7 years.

I haven't said it. It was the time which passed to have AGA.
Quote:

The Amiga team worked on other ideas, AAA for example. As the other project were delayed, not ready, cancelled, they cooked AGA "Spaghetti" for us in 10 minutes. They not worked 7 years on AGA. They had limited time only. If you have a limited time, then you need to focus. You can only do "so many" features in this time. And every feature that you try to do has a risk to fail - and the whole project might then fail.

Sorry, but I can't believe that nothing good was ready for a better chipset, and that the engineers had to patch the ECS to create the AGA.

In 7 years I expect that engineers had already something better. Which could be "cut" / scaled down.

Instead of hurring for a horrible patch...
Quote:

AGA has significant improvements over OCS. And they limited the features to the amount that they could do, and could debug. This was a good decision.

AGA sucks and I've shown you already why.

It was an enhancement, yes, but not good.
Quote:

Its better to have a working and released chipset with 5 new improvements, than an non working non released with 10 new features.

Sure. Nothing against it.

@cdimauro

Quote:

In 7 years I expect that engineers had already something better. Which could be "cut" / scaled down.

In other words:
You have no clue about hardware development, but "you assume" and "you expect" and then you talk stupid.

TALK LESS, DO MORE!

Bring out your USA RACING and your TINA.

@Thread

What justifies AGA is the software:

Productivity: DPaint 4 AGA, Deluxe Paint 5, Brilliance etc

Games: Slam Tilt, Banshee, The Acid Collection, Xtreme Racing, Theme Park AGA, Sim City 2000 and Alien Breed 3D, Wing Commander CD32 etc.

'nuff said!

Ruff 'n Tumble and The Settlers aside there were no ECS games that competed on a presentation level!

Yes, SWIV and Battle Squadron were fun but we needed Guardian and Slam Tilt to usher in the mid 90s!

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

@bhabbott

Quote:

Not really. Only the A3000 had SCSI with DMA built in, and it had problems. The A2090 and A2091 cards for the A2000 and the A590 expansion for A500 had DMA, but with only 24 bit addressing. Furthermore the MFM and XT-IDE drives supplied with them were not fast, so they were the bottleneck. The CDTV also had DMA with an optional SCSI interface, but was even more limiting with no Fast RAM. DMA certainly helped it perform better, but that is not really relevant because the machine is so limiting in other ways.

24 bit DMA was common and usually sat in chip ram. However, I found these limitations tended to affect the big boxes like A4000. Despite being more suited for pro use the A4000 SCSI cards I found tended to be limited and needed particular mask and max transfer settings. But a SCSI card on my A1200 could be left on full 32 bit settings. Suppose it comes down to the hardware but I expected A4000 cards to use the full 32 bit width.

Quote:

IDE was always about being cheap, since the drives were generally cheaper than SCSI drives. However even before that a lot of non-DMA ST506 and SCSI drive interfaces were produced by 3rd parties. The reason is that a DMA controller is expensive and hard to get working right, while not improving basic loading times significantly. Most people were happy just getting the extra speed and capacity of a hard drive at all, and DMA was only sought after by those who wanted the best possible performance and/or had specific requirements for video recording etc.

Despite buffering issues, I think Commodore could have improved the interface without needing DMA on the controller. They could have added a small cache so full blocks could be transferred without the CPU hand holding the controller. A hybrid DMA solution I suppose. The A600 and to some extent the A1200 were low power models so reducing CPU load was needed.

Quote:

I know of at least one DMA IDE controller card, the Masoboshi Mastercard. But most other 3rd party IDE cards use PIO, even those produced before the A600 was released. So PIO was an 'industry standard' on the Amiga long before Commodore adopted it.

That's an interesting one. Also came with optional SCSI on board. And RAM slots.

PIO is better if it's not stuck at PIO 0.

Quote:

Nitpick - the serial port is part of Paula, the CIA is only used for RS232 control and status lines.

The Amiga chips were always a funny mix. Paula is best known for audio while also doing floppy. Where as the CIA chips have dedicated serial ports that are used internally.

Quote:

The 68000 was designed to use 6800/6502 peripherals, including Commdore's VIAs - so it made perfect sense to use them in the Amiga. When the Amiga was designed the idea of high speed serial and parallel hadn't yet caught on. PCs didn't have buffered serial chips, and printers couldn't work at full bus speeds (the CIA's strobe pulse is deliberately slowed down to meet Centronics printer timing).

The 6800 at least. Since Motorola didn't approve of the 6502.

Quote:

Next sticking point is that the Amiga didn't have a general purpose onboard DMA controller. It was expected that addon cards would provide their own if needed. The A3000 had one for SCSI, but even just doing that one job was difficult. If you put RAM on an accelerator card then the onboard DMA has to work over its bus at the same speed, which is problematic when the card is faster than the motherboard. If that is not possible then data still has to be copied into accelerator card RAM by the CPU. Another problem for serial ports is that DMA is generally block based, while serial is stream based and often needs to be monitored character by character.

Serial ports were one of the first to slow it down when demands increased. Like parallel the rely on CPU to keep speed up. But it drags the system down.

Quote:

The serial port in Paula was just a bonus for low speed uses that were common at the time (it did at least do MIDI, which standard PC serial ports couldn't because they didn't have the required baud rate). But the Amiga had an open hardware interface that was expected to be used for more demanding applications. There was nothing stopping from you installing a board with 16550 buffer serial chip if you needed it, or even a sophisticated comms board with local RAM and Z80 CPU.

The Amiga could do MIDI but the baud speed wasn't exactly MIDI complaint. Also, any kind of 16550 board doesn't plug into an A500 or A1200 so that didn't solve it.

Quote:

Actually only the PS/2 line got a bidirectional printer port. Most clones stuck with the standard output-only port, and had to use 'nibble mode' via the status lines for input. Another thing they didn't have was power on the connector, so any device plugged into the port had to be separately powered.

The first model apparently had a bidirectional parallel port but wasn't that popular so was dropped for a few years.

Quote:

It wasn't necessary. Any device that needed high speed parallel could use a plugin card. The primary use for EPP/ECP on PCs was GDI printers, which had secret proprietary protocols and needed large amounts of RAM and CPU power to operate.

Once again cards don't plug into a desktop Amiga. But why should the buyer have to purchase more hardware? A new computer should be equipped to handle the needs of the day. The A4000 was expensive. And it didn't meet the current standards of the computer market.

Quote:

It wasn't necessary. Any device that needed high speed parallel could use a plugin card. The primary use for EPP/ECP on PCs was GDI printers, which had secret proprietary protocols and needed large amounts of RAM and CPU power to operate.

Not on common Amigas. The Amiga suffered when scanners and parallel drives could not be plugged into it. This should have been possible with the A1200 but they left 7 year old hardware on board.

Quote:

Indeed. Parnet on the Amiga is faster and more reliable than Laplink parallel on a PC. I used it regularly for transferring files between Amigas.

I should hope so!

Quote:

Not quite. CIAs didn't need to be speeded up for HD floppies. Paula MFM decoding could easily have been speeded, but more DMA slots would have to be made available - with bigger buffers for it. Still doable, but not quite as simple and possibly breaking compatibility.

The MFM encoding and decoding was done in software. Or by the blitter. Technically, there is roughly enough bandwidth to transfer the data bits, but it's transferred as MFM so takes up double he space and needs double the speed on average. If the MFM was offloaded to hardware it could be possible. In any case, compatibility should be no problem when hardware is set to default state, and software doesn't poke around.

Quote:

Perhaps not. One way was a clever hack using a couple of TTL logic chips in the floppy drive. The other involved critical redesign of the custom chips.

They still had to modify the drives. Over just dropping a standard part in which was easily replaceable. But by AGA all the chips should have already had a critical redesign and been ready so it shouldn't have been a issue.

Quote:

I think they should have done nothing, then Amiga fans wouldn't be disappointed by the results. 880k was plenty enough for normal uses, and CDROM was about to make floppies redundant for software distribution. Today PC don't have a floppy drive and you can't even buy HD disks new.

Lots of software already came on multiple disks including games. Not only would speed benefit. 1.44MB PC disks became common leaving 720K disks behind making it hard to transfer data with other computers.

Quote:

It's the same interface, just with different signals on the pins. Older floppy drives had jumpers to set the pin signals, to suit other devices which used a similar configuration to the Amiga. As competition heated up in the PC market, drive manufacturers dropped the jumpers and hardwired them to save money. Then they stopped making drives altogether. If Commodore had had survived they would have continued to have correctly configured drives made for them.

It was already a problem before Commodore were on the way out. Replacing an Amiga floppy drive meant sourcing an Amiga drive or hacking a PC one. The lines are different enough that it won't work without a hardware interface. Including the logic needed for drive ID on boot and providing a disk ready signal. One thing common PC drives lacked. We saw what happened with the AT debacle installing drives without disk ready in hardware. It broke compatibility more so than the A1200 itself.

Quote:

What good have 16 bit have done? A stock A1200 doesn't have the CPU power to decode MP3's or the memory to store large samples. If you want CD quality music then use a CDROM drive. Amiga MODs were always 8 bit - and sounded great. 8 track MODs generally don't add much. If you really wanted more then cards were available, but few people bought them because - unlike PCs - every Amiga has a 'sound card' built in.

16 bit would have given it better quality. An A1200 doesn't need to decode MP3, it just needs to play samples. There was demand for it and the Falcon had it. At the time we saw utilities like Play16 simulating the missing 16 bit Mary sound chip. Playing MP3 on an Amiga is just a novelty to listen to low quality music at even lower quality. Average modules were made of low quality samples. 8 track mods pushed the hardware too much as they tasked the CPU and 8 bit wasn't enough headroom. It needed more tracks minimum. If the CPU was forced to mixdown the quality was lower because it was all stuck in 8 bit. And then there were those 14 bit hacks showing demand and how desperate we became. The sound cards for the A1200 were paltry, with some being one stereo track of 12 bit and maybe one stereo track of 16 bit if you were lucky, without MIDI. Not much use for anyone there.

@Hypex

Quote:

PIO is better if it's not stuck at PIO 0.

A4000 was supposed to support faster IDE modes
A4000 Register $DD1020 is for enable it.
The Vampire has this register working and does over 20MB/sec

@Gunnar

"The Vampire" meaning what?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

@kolla

Quote:

"The Vampire" meaning what?

Meaning all systems with my Fast-IDE.
This means these cards: V500/V666/V1200/V4SA/FIREBIRD/ICEDRAKE/MANTICORE/...

@Gunnar

Quote:

Gunnar wrote: @cdimauro Quote: In 7 years I expect that engineers had already something better. Which could be "cut" / scaled down. In other words: You have no clue about hardware development, but "you assume" and "you expect" and then you talk stupid.

It's enough to take a look at the history to rebut your false statement.

ALL, again: ALL other platforms evolved over the years bringing better hardware technologies.

Only the Amiga platform was stuck SEVEN years to bring some concrete enhancement with the AGA chipset.

ECS, released two years before AGA, brought some technology which was already available on the archaic Hitachi 46505...
Quote:

TALK LESS, DO MORE!

Says the one which is still NOT implementing a PMMU on his "100% Amiga compatible" toy and which continuously reports lies when some people ask for it:
http://apollo-core.com/knowledge.php?b=4¬e=39145&z=v1bg8G

"Please understand that the Apollo-Team are 110% Amiga and Atari fans.
We have zero interest in Linux or BSD."
Quote:

Bring out your USA RACING

You're a Broken Disk. Already replied here: https://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=44581&forum=15&start=500&viewmode=flat&order=0#855857
Quote:

and your TINA.

"If you repeat a lie often enough, people will believe it, and you will even come to believe it yourself." - Joseph Goebbels

Same as before. Replied here: https://amigaworld.net/modules/newbb/viewtopic.php?topic_id=44169&start=200&post_id=855345&order=0&viewmode=flat&pid=0&forum=17#855345

@matthey

Quote:

matthey wrote: Karlos Quote: I don't have an axe to grind about AGA. The only misstep in my view was the lack of fast memory to get the most our of the CPU at the time of launch. Did Jay Miner fix that problem with Ranger? https://en.wikipedia.org/wiki/Amiga_Ranger_Chipset Let us say memory was $100 per MiB sometime in the '80s when Ranger would have been ready and VRAM was 20% more expensive. 2MiB DRAM chip mem = $200 mem cost 2MiB VRAM chip mem = $240 mem cost (Ranger) 2MiB DRAM chip mem + 1 MiB fast mem = $300 mem cost Certainly 3MiB of total mem would be better if 2MiB wasn't enough for a particular program. There are several valuable advantages to the 2MiB VRAM though. o dual ported memory could give full mem access to both the chipset and CPU at the same time.

Which isn't optimal.

The Amiga had different buses to let both chipset and CPU proceed in parallel while using the their own memory as much as they could do without interfering each other.
Quote:

o no data duplication and copying between mem types is necessary

That's a good, but if you have different memory types and you load on them only their specific assets, then you don't need do to any copy and the system works at its maximum capability.
Quote:

o one continuous memory chunk

Not so important.
Quote:

o no 64 bit mem fetch incompatibility or restrictions with double the number of 32 bit slots instead

This works only for dual ported system. But advancing the memory technology you have to resort to bigger busses anyway which demand for bigger alignments.

You cannot create quadruple, octuple, etc., ported memories...
Quote:

I'm no expert on how a Ranger Amiga was going to use double the mem slots of VRAM. There seems to be limited information on Ranger despite Jay Miner saying he finished the specification and Dale Luck claiming to have a prototype. Dave Haynie thought Ranger was a myth or at best specs on paper only. It appears CBM hid Ranger away well for West Chester engineering to know practically nothing about what Los Gatos had worked on for maybe a year. CBM management was like thanks Jay, let me file the Ranger specs in the "Amiga is not the next C64 drawer". It would be interesting to know if Michele Battilana found them in the trove of CBM documents or whether they were destroyed by CBM amid the animosity with the Los Gatos developers.

Indeed. Because if it was confirmed what Hayine said, then it looks like that Commodore engineers produced too many papers and too few concrete products...

@bhabbott

Quote:

bhabbott wrote: @cdimauro Quote: cdimauro wrote: I've just reported what the Amiga Hardware Manual said about the specific question. But the answer is misleading. What frequency the blitter circuit is 'clocked' at is irrelevant, what matters is how many 'ticks' it takes to perform an operation.

Well, those ticks are coming... from the Blitter clock.
Quote:

Furthermore the results are constrained by memory cycle time, so no matter how fast it might go internally it is limited to 3.5MHz per memory operation.

This generally valid for all the chipset, which was working based on the color clock for all memory accesses.
Quote:

Quote: I'm assuming a very simple scenario which is also, more or less, the BEST case.... PAL (256 lines): Total slots 70512 Used slots 27866 Free slots 42646 Free slots (percentage) 60% AGA FMode 3 (64-bit fetch), 5 bitplanes, PAL (256 lines): Total slots 70512 Used slots 8666 Free slots 61846 Free slots (percentage) 88%... Since with AGA and 8 bitplanes you have only 83% (NTSC) and 82% (PAL) of slots available, you can see yourself that those aren't enough for 8 bitplanes = 256 colors, as you were suggesting. Whereas 7 bitplanes = 128 colors is much more realistic, albeit the numbers are still a little bit below of what is really needed. Whether it's enough depends on what you need to do. With AGA you have 37% more slots in 256 colors than you had with OCS in 32 colors.

No, as I've already calculated and posted here: https://amigaworld.net/modules/newbb/viewtopic.php?mode=viewtopic&topic_id=44362&forum=2&start=240&viewmode=flat&order=0#855920

It's 83% vs 63% (NTSC) = 20% more slots for AGA. 82% vs 60% = 22% more slots for AGA.
Quote:

Since you have more colors to spare you can do more tricks to cut down on blitting.

Which tricks? If you have to blit 3 more bitplanes and you only have an increased load on the Blitter.
Quote:

A game that was limited to 32 colors in OCS could well be able to get 256 colors on AGA.

I've already mathematically proved that this is impossible: you don't have enough slots! Here are again the calculations:

NTSC:
Slots needed per 8 bitplanes 101%

PAL:
Slots needed per 8 bitplanes 97%

Since with AGA and 8 bitplanes you have only 83% (NTSC) and 82% (PAL) of slots available, you can see yourself that those aren't enough for 8 bitplanes = 256 colors, as you were suggesting.
Quote:

Even if Gunnar's figures are a bit 'optimistic' for a typical scenario, I agree with him that AGA was more than just a 'modest' improvement. 4 times higher bandwidth, sprites up to 4 times larger,

And if you enable the scroll with 64-bit fetches 7 of them disappear...
Quote:

8 times more colors, 24 bit color palette vs 12,

That was nice. But the CLUT implementation sucked.
Quote:

superhires HAM8 for photographic quality images.

It was very very slow.
Quote:

If this was a VGA card you wouldn't call the improvements 'modest'.

Well, check what IBM brought with the VGA from the EGA, in only 3 years, and the improvements were huge.

Not even counting IBM's 8514 which arrived with the VGA, which did much more. But for the professional market.

@cdimauro

Regarding HAM8 speed, I dunno, this is quite impressive to me:

https://youtu.be/SwnDqj8pNe4

_________________
Doing stupid things for fun...

@Thread

Some quite prophetic words from Aug 1986:

Quote:

My gut feeling is that the prohibitive cost of updating all the custom chips in the Amiga will prevent the Amiga 2 from being an earth-shattering improvement on the current model, anyway. What I guess will happen is that third-party companies will work to extend the Amiga's open architecture with upgrades. But judging by current prices being charged for samplers, extra RAM, and hard disks, it's also a fair bet that enhancements will come at a price.

Source

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios