Click Here
home features news forums classifieds faqs links search
6071 members 
Amiga Q&A /  Free for All /  Emulation /  Gaming / (Latest Posts)
Login

Nickname

Password

Lost Password?

Don't have an account yet?
Register now!

Support Amigaworld.net
Your support is needed and is appreciated as Amigaworld.net is primarily dependent upon the support of its users.
Donate

Menu
Main sections
» Home
» Features
» News
» Forums
» Classifieds
» Links
» Downloads
Extras
» OS4 Zone
» IRC Network
» AmigaWorld Radio
» Newsfeed
» Top Members
» Amiga Dealers
Information
» About Us
» FAQs
» Advertise
» Polls
» Terms of Service
» Search

IRC Channel
Server: irc.amigaworld.net
Ports: 1024,5555, 6665-6669
SSL port: 6697
Channel: #Amigaworld
Channel Policy and Guidelines

Who's Online
26 crawler(s) on-line.
 139 guest(s) on-line.
 0 member(s) on-line.



You are an anonymous user.
Register Now!
 Gunnar:  44 mins ago
 pixie:  1 hr 3 mins ago
 DiscreetFX:  1 hr 42 mins ago
 DWolfman:  1 hr 51 mins ago
 cncparts:  3 hrs 25 mins ago
 saipaman4366:  4 hrs 11 mins ago
 Beajar:  4 hrs 30 mins ago
 Rob:  4 hrs 32 mins ago
 agami:  5 hrs 36 mins ago
 RobertB:  5 hrs 55 mins ago

/  Forum Index
   /  Amiga General Chat
      /  How good or bad was the AGA chipset in 1992/1993.
Register To Post

Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 Next Page )
Poll : How good or bad was the AGA chipset in 1992/1993.
10p Excellent (Best at 2D/3D, colors, and resolution, frame rate etc.)
5p Good / better than most computer.
0p Barely hanging in there.
-5p Below average / slow but usable
-10p useless / horrible
 
PosterThread
bhabbott 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 11-Oct-2022 22:40:59
#281 ]
Regular Member
Joined: 6-Jun-2018
Posts: 330
From: Aotearoa

@Hypex

Quote:

Hypex wrote:

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.

Only Zorro III cards (or controllers on the CPU board) can be 32 bit. Several A1200 accelerators had optional on-board SCSI ports. I had the DKB Cobra with Ferret SCSI board. I don't know if this used DMA (based on the small size and number of 'glue' chips in the Cobra I suspect not). The Blizzard 1230IV that I have now definitely does, but I don't have a SCSI board for it (or any SCSI devices to use with it).

This is how it should have been done for faster IDE too. Putting a 32 bit DMA controller on the motherboard would have been expensive, limited the speed, and probably buggy. The A1200 was designed as a low cost home computer to replace the A500+, but with open-ended expansion to much higher performance via the 32 bit expansion slot. Putting a high performance DMA hard drive interface on the motherboard didn't make sense for its intended market. You could argue that the A4000 should have had a better IDE interface, but it had the slots to take SCSI cards that professionals preferred.

Quote:
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.

No, it wasn't.

Most A600s and A1200s were sold without a hard drive, and many owners never installed one because they were happy enough with floppy disk speeds. Even with 'slow' PIO the A1200 can achieve over 2MB/s with a good hard drive. This is a huge speed boost over floppies. Sure it loads the CPU, but in typical use there isn't much the CPU would be doing anyway during eg. loading a game.

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

Sure, but consider the time frame. Faster PIO modes were not standardized until 1994. The A600 was released in March 1992. The specification sheet for Gayle is dated July 1991, 3 years earlier. How was Commodore expected to support faster modes without a proper standard to work from?

Coincidentally 1991 was the year I started my business selling Amigas and PCs. I remember having a problem with new IDE hard drives and certain motherboards, which according to my supplier was was caused by logic voltage level mismatch. I don't know if this was true, but I do know that we had significant compatibility issues with IDE back in those days.

Quote:
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.

Only one 'serial' port in the CIAs is used, for the keyboard interface. The other one is spare (I used it to remotely control a Panasonic video recorder).

Most other home computers used a standard 'UART' chip for RS232 serial, but this was almost always an option provided on a plug-in card, not a part of the base machine. Some machines used nasty bitbanged serial, which was OK for printers but not good for bidirectional communications.

Quote:
The Amiga could do MIDI but the baud speed wasn't exactly MIDI complaint.

No, it was certainly compliant enough.

The only problem was that in a multitasking environment it was easily possible to overrun the serial receive buffer when reading MIDI commands from a synthesizer keyboard (imagine pressing 10 keys at once, with each one sending 3 bytes for key down). This problem could have been solved by restricting multitasking, but the companies developing MIDI apps didn't want to do that (or didn't know how). Unfortunately many Amiga developers did not understand the hardware they were working with.

Quote:
Also, any kind of 16550 board doesn't plug into an A500 or A1200 so that didn't solve it.

Nonsense. With a simple interface you can plug any standard ISA bus serial card into the A500. Several 16550 boards that plugged into the A1200's clock port were produced. The only reason we didn't see lot of buffered serial solutions for the A500 was lack of demand. Professional users bought a big box Amiga, for which many cards were produced.

I remember when the 16550 became desirable as MODEMs got faster. Most older PCs only had a standard 16450. Some had a socket to install a 16550, but the chips were hard to get (at least here in New Zealand). I had ideas of making a clockport 16550 board for my A1200, but never got around to it because my 28k8 MODEM didn't need it. Then I write a driver for the CNET PCMCIA Ethernet card and it became moot.

Quote:
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.

External MODEMs were expensive, so the industry was moving towards internal MODEMs (which had their own serial chip so the stock serial port buffering was irrelevant). A few were produced for the Amiga, but of course it was 'proprietary' compared to PC cards.

This is where Commodore missed an opportunity. Instead of making Bridgeboards they should have produced a simple card that connected the ISA bus slots to the Zorro II bus. Then customers could use cheap ISA bus cards in their Amiga. Had they done this the early days they might have expanded it to PCI so we could use high performance graphics cards etc.

Quote:
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.

Hand scanners designed for the Amiga could be plugged into it - I sold one in my shop. But like PC hand scanners they were not very good. A flatbed scanner was much better, but expensive. I had a SCSI flatbed scanner on my A1200.

The only thing stopping parallel drives being plugged into the Amiga was that they were designed for PCs. I developed software to use a ZIP parallel drive on the Amiga. I have often thought of getting another ZIP drive since they often go cheap on auction sites, but I can't be bothered because we have better solutions now (those ZIP drives had a nasty habit of dying!).

Quote:
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.

Sure it's possible. Or just throw in a 765 controller like other machines did. But why do we need it? DD disks work fine.

Quote:
They still had to modify the drives.

No, 'they' didn't. The drive manufacturer made the drives. But again, why bother?

Quote:
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.

I had no problem getting DD disks for my shop throughout the 1990s. The only problem most Amiga users had was finding enough disks to copy their latest 'acquisitions' onto.

Towards the end the number of disks that some games came on was ridiculous. But this was mostly because the games were developed for the PC first. The PC version was uncompressed and installed on the hard drive for playing, but the Amiga version had to play directly from disk because most Amiga owners were too stingy to buy a hard drive. Therefore it couldn't be compressed as efficiently, and so took up many more disks than it could have.

This problem soon want away with CDs. But it was largely too late for the Amiga, as not enough Amiga owners had CDROM drives. The other problem was more fundamental - CDROMs were less attractive to users because at that time they were too hard to pirate. For most Amiga users the ability to get any game for the price of a blank disk or two was its main attraction.

Last edited by bhabbott on 11-Oct-2022 at 10:47 PM.
Last edited by bhabbott on 11-Oct-2022 at 10:41 PM.

 Status: Offline
Profile     Report this post  
bhabbott 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 11-Oct-2022 23:15:44
#282 ]
Regular Member
Joined: 6-Jun-2018
Posts: 330
From: Aotearoa

@BigD

Quote:

BigD wrote:
@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...


Source

That's an interesting prediction especially since it's from 1986. What would be the point of producing an 'earth-shattering improvement' when developers had barely scratched the surface of what was already in the Amiga?

Quote:
Quote:
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.

RAM was expensive. Hard drives were expensive. But 3 years later I had a second hand Microbotics 2MB Expansion on the A1000 that I only paid NZ$300 for (same price I paid for 32 kilobytes of bare RAM chips for the ZX Spectrum a few years earlier), and a second hand 20MB hard drive and PC controller that I hooked up to it via a simple bus adapter.

Of course by then the A2000 was out, which had lots of slots and plenty of room for hard drives etc. with a nice big power supply. Things like RAM and hard drives might not be 'earth-shattering' chipset-wise, but they make a big difference to functionality. The A2000 eventually made things like the Video Toaster and 040 accelerator cards with onboard RTG a reality. So in some ways that 'big box' design was a more significant development than possible improvements to the custom chips.

 Status: Offline
Profile     Report this post  
matthey 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 0:42:46
#283 ]
Super Member
Joined: 14-Mar-2007
Posts: 1968
From: Kansas

cdimauro Quote:

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.


It looks like VRAM is used differently from both ways I thought it may have been implemented.

https://retrocomputingnews.wordpress.com/2015/05/31/the-aui-interview-jay-miner-the-father-of-the-amiga/ Quote:

Video RAM

Commodore now has a high resolution chip set of Amiga chips that I worked on when we were with Amiga in Los Gatos. These chips use video ram and can produce a very high resolution ten twenty four display along with the present Amiga display simultaneously. They increase the display address range to two megabytes. These chips are completed and tested and only require a computer and memory to hold them together. I’d like everyone to know that Amiga in Los Gatos designed these chips! These improved Amiga chips can use a new type of ram called video ram. This new type of ram – video ram – is a giant step in computer improvement because it frees up the bottleneck into memory caused by competition between the computer itself and the memory fetchers required for the high resolution display. Imagine having an additional gigantic parallel output port thousands of bit wide, just for video. You wouldn’t have to access it very often to dump a lot of memory data to a video picture.

The way it works is the video data for the high resolution display is dumped from memory into a large parallel to serial shift register right on the video ram chips. This outputs hundreds of picture bits – pixels – in one memory cycle, leaving 99% of the memory bandwidth for the computer. This is critically important for very high resolution multi-bitplane colour displays.

Video ram can also be used for other things than video. It can provide a very fast path to hardware parallel processors; such as blitter and all kinds of I/O such as audio, hard disk and networking ports. Special purpose multiport chips like the video ram will continue to evolve and we will see multi-shift register ports for dumping many of these datatypes bi-directionally and simultaneously. So keep your eye on video ram and on the next generation of Amiga computers that will probably use it.


It sounds like it frees up chipset and CPU memory bandwidth but isn't like doubling it all the time? There should be much more available bandwidth for the chipset and CPU anyway. Fast memory likely would leave even more chipset bandwidth for the chipset but I expect this would bring a large performance improvement without fast memory.

I also found another really good site on the Ranger with info and pics of the prototype.

https://obligement-free-fr.translate.goog/articles/ranger.php?_x_tr_sch=http&_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en&_x_tr_pto=wapp

It is in French but I set up Google translate.

cdimauro Quote:

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...


It sounds like no more than dual ported memory would be used the way VRAM is implemented? VRAM eventually lost popularity to high bandwidth DDR DRAM. Yes, modern DDR DRAM is setup for wider accesses. Memory for the chipset on a custom(izable) chip is more versatile. There are a lot more options today for memory than there was back then. VRAM was the way to substantially increase memory bandwidth back then and a 20% cost premium wasn't that bad for double the bandwidth. VRAM pretty well separated high end from low end graphics hardware in the late '80s into early '90s.

cdimauro Quote:

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...


The pics in the link above show the Ranger prototype. I can't tell if the prototype chipset is wire wrapped but I expect that is what it would be (custom chips were expensive back then without the project go ahead from CBM). A Ranger prototype was seen at the Comdex show in 1986 but what are the chances it was working? The expansion board size was large and more square which is supposedly one of the things CBM didn't like. It's that chipset that counts though. I expect 2MiB of VRAM in 1988-1989 would have transformed the Amiga with no more "slow" and cramped chip mem. Instead, CBM gave us 512kiB of chip mem with the slowest possible 512kiB "Ranger" mem using part of the Ranger address location that was neither fast or chip mem. Big difference. It sure sounds like the Ranger VRAM bandwidth could have displayed high resolution full color productivity modes on cheap SVGA monitors without extra hardware. Amiga was the next C64 toy so from CBM's perspective a TV was probably good enough.

 Status: Offline
Profile     Report this post  
agami 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 0:53:19
#284 ]
Super Member
Joined: 30-Jun-2008
Posts: 1637
From: Melbourne, Australia

@bhabbott

Quote:
bhabbott wrote:
@BigD

Quote:
BigD wrote:
@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...

Source

That's an interesting prediction especially since it's from 1986. What would be the point of producing an 'earth-shattering improvement' when developers had barely scratched the surface of what was already in the Amiga?

Exactly.
Hardware development in the '80s was a much more expensive prospect when compared to today.

What @cdimauro et al neglect to understand is the timings.
It's not about what engineers are or aren't capable of cooking up. Engineers are always working on something. But CBM was a business, and as much as it irks me, engineering is subservient to the business, which is in turn subservient to the market.

The Amiga, as it turns out, and as was the dream of Hi Toro, was a gaming console masquerading as a personal computer.
It's then best to think of the A1000 as the AGC (Amiga Games Console) dev kit. Then the starting pistol went of in 1987 with the release of the AGC (A500).

Once that little gem and the V2 of the Dev Kit (A2000) started making some money, the business could start looking at how much they are willing to invest in R&D for the next generation of AGC. Engineers are given a target and they work toward it.

As is the nature of plans, they change. Engineers are told the business has reduced the amount of money they're willing to spend. Oh and they now have only 12-18 months to put something together. In a time where things can't be fixed post-release with downloadable firmware. So unless they had the foresight in 1988 to design a highly scalable solution, it was going to end up being a hack.

So CBM engineers never really had 7 years. At best they had 4 years, and ultimately they had less than 2.

There's an excellent scene in Apollo 13 that I think illustrates this well
https://youtu.be/ry55--J4_VQ?t=23

Last edited by agami on 12-Oct-2022 at 12:57 AM.
Last edited by agami on 12-Oct-2022 at 12:56 AM.
Last edited by agami on 12-Oct-2022 at 12:55 AM.

_________________
All the way, with 68k

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 4:59:36
#285 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@Karlos

Quote:

Karlos wrote:
@cdimauro

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

https://youtu.be/SwnDqj8pNe4

Very likely the palette is fixed so the conversion from chunky to planar/HAM8 is much faster.

HAM8 needs an optimized palette to give its best quality. It means that quite some time is needed to scan the original 24-bit image and calculate a set of good 64 base colors.

As you can see from the video, the quality is really poor and doesn't show so many colors. Maybe the used textures weren't that much good and/or the scene doesn't give justice to the technique.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 5:03:04
#286 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@BigD

Quote:

BigD wrote:
@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

Indeed. Really prophetic words.

@bhabbott

Quote:

bhabbott wrote:
@BigD

That's an interesting prediction especially since it's from 1986. What would be the point of producing an 'earth-shattering improvement' when developers had barely scratched the surface of what was already in the Amiga?

The quoted part attains only to the HARDWARE part: the chipset. Its evolution.

Developers aren't part of the author's reflections, here.

And BTW some developers already produced nice things on '85 and '86.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 5:11:46
#287 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@matthey

Quote:

matthey wrote:
cdimauro Quote:

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.


It looks like VRAM is used differently from both ways I thought it may have been implemented.

https://retrocomputingnews.wordpress.com/2015/05/31/the-aui-interview-jay-miner-the-father-of-the-amiga/ Quote:

Video RAM

Commodore now has a high resolution chip set of Amiga chips that I worked on when we were with Amiga in Los Gatos. These chips use video ram and can produce a very high resolution ten twenty four display along with the present Amiga display simultaneously. They increase the display address range to two megabytes. These chips are completed and tested and only require a computer and memory to hold them together. I’d like everyone to know that Amiga in Los Gatos designed these chips! These improved Amiga chips can use a new type of ram called video ram. This new type of ram – video ram – is a giant step in computer improvement because it frees up the bottleneck into memory caused by competition between the computer itself and the memory fetchers required for the high resolution display. Imagine having an additional gigantic parallel output port thousands of bit wide, just for video. You wouldn’t have to access it very often to dump a lot of memory data to a video picture.

The way it works is the video data for the high resolution display is dumped from memory into a large parallel to serial shift register right on the video ram chips. This outputs hundreds of picture bits – pixels – in one memory cycle, leaving 99% of the memory bandwidth for the computer. This is critically important for very high resolution multi-bitplane colour displays.

Video ram can also be used for other things than video. It can provide a very fast path to hardware parallel processors; such as blitter and all kinds of I/O such as audio, hard disk and networking ports. Special purpose multiport chips like the video ram will continue to evolve and we will see multi-shift register ports for dumping many of these datatypes bi-directionally and simultaneously. So keep your eye on video ram and on the next generation of Amiga computers that will probably use it.


It sounds like it frees up chipset and CPU memory bandwidth but isn't like doubling it all the time?

Indeed.
Quote:
There should be much more available bandwidth for the chipset and CPU anyway.

Double.
Quote:
Fast memory likely would leave even more chipset bandwidth for the chipset but I expect this would bring a large performance improvement without fast memory.

Correct and I would have liked to have all that bandwidth available. Especially only on chip mem.
Quote:
I also found another really good site on the Ranger with info and pics of the prototype.

https://obligement-free-fr.translate.goog/articles/ranger.php?_x_tr_sch=http&_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en&_x_tr_pto=wapp

It is in French but I set up Google translate.

OK, so it proves that it wasn't a paper project.
Quote:
cdimauro Quote:

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...


It sounds like no more than dual ported memory would be used the way VRAM is implemented?

I think so.
Quote:
VRAM eventually lost popularity to high bandwidth DDR DRAM. Yes, modern DDR DRAM is setup for wider accesses. Memory for the chipset on a custom(izable) chip is more versatile. There are a lot more options today for memory than there was back then.

Unfortunately DDR wasn't available. So VRAM was a good substitute.
Quote:
VRAM was the way to substantially increase memory bandwidth back then and a 20% cost premium wasn't that bad for double the bandwidth. VRAM pretty well separated high end from low end graphics hardware in the late '80s into early '90s.

Yup. And it was good to have.
Quote:
cdimauro Quote:

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...


The pics in the link above show the Ranger prototype. I can't tell if the prototype chipset is wire wrapped but I expect that is what it would be (custom chips were expensive back then without the project go ahead from CBM).

Then was Hayine wrong?
Quote:
A Ranger prototype was seen at the Comdex show in 1986 but what are the chances it was working?

I think close to zero.
Quote:
The expansion board size was large and more square which is supposedly one of the things CBM didn't like.

Indeed. But there was time to reduce its complexity: this was just a prototype.
Quote:
It's that chipset that counts though.

Indeed.
Quote:
I expect 2MiB of VRAM in 1988-1989 would have transformed the Amiga with no more "slow" and cramped chip mem.

Too late and Ranger chipset provided too limited innovations.
Quote:
Instead, CBM gave us 512kiB of chip mem with the slowest possible 512kiB "Ranger" mem using part of the Ranger address location that was neither fast or chip mem. Big difference.

It was nightmare. I want to cry...
Quote:
It sure sounds like the Ranger VRAM bandwidth could have displayed high resolution full color productivity modes

Unfortunately it didn't provide "full color". Only 128 colors, that were too few on '89.
Quote:
on cheap SVGA monitors without extra hardware.

That was nice.
Quote:
Amiga was the next C64 toy so from CBM's perspective a TV was probably good enough.

Yup. As we know...

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 5:52:09
#288 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@agami

Quote:

agami wrote:
@bhabbott

Quote:
bhabbott wrote:
@BigD

That's an interesting prediction especially since it's from 1986. What would be the point of producing an 'earth-shattering improvement' when developers had barely scratched the surface of what was already in the Amiga?

Exactly.
Hardware development in the '80s was a much more expensive prospect when compared to today.

Are you serious? On '80s there were single-man projects whereas today we several hundreds of engineers for a single project.
Quote:
What @cdimauro et al neglect to understand is the timings.
It's not about what engineers are or aren't capable of cooking up.

Well, actually you've to read this: On the Edge: The Spectacular Rise and Fall of Commodore

Specifically on the part where he was talking about the engineers that remained when the ones that created the Amiga left the company.
Quote:
Engineers are always working on something. But CBM was a business, and as much as it irks me, engineering is subservient to the business, which is in turn subservient to the market.

Sure. And it wasn't the same for all other companies & engineers involved on different, concurrent projects?

What's not clear about what I've already reported before:

"ALL other platforms evolved over the years bringing better hardware technologies."

?

Why don't you look at what Commodore's competitors did over the same period of time?
Quote:
The Amiga, as it turns out, and as was the dream of Hi Toro, was a gaming console masquerading as a personal computer.

A very expensive console...
Quote:
It's then best to think of the A1000 as the AGC (Amiga Games Console) dev kit.

It was a concrete product: it was The Amiga...
Quote:
Then the starting pistol went of in 1987 with the release of the AGC (A500).

That was because of the natural strategy in Commodore: produce cost-reduced versions of the hardware.
Quote:
Once that little gem and the V2 of the Dev Kit (A2000) started making some money,

The Amiga 2000 was super-expensive. And offered nothing more to developers compared to the A500.

I know because I had one, unfortunately (better to have bought an A500: my father could have saved a pile of money).
Quote:
the business could start looking at how much they are willing to invest in R&D for the next generation of AGC. Engineers are given a target and they work toward it.

So, basically you're saying that Commodore had no money for designing next generation chipset and it was waiting for the cost-reduced Amiga versions to sell to accumulate the required for starting it. Right?

Where Commodore found all the money wasted on OTHER projects then?
Quote:
As is the nature of plans, they change. Engineers are told the business has reduced the amount of money they're willing to spend. Oh and they now have only 12-18 months to put something together. In a time where things can't be fixed post-release with downloadable firmware. So unless they had the foresight in 1988 to design a highly scalable solution, it was going to end up being a hack.

And this story lasted... for 7 years?
Quote:
So CBM engineers never really had 7 years. At best they had 4 years, and ultimately they had less than 2.

Sure. Whereas for other projects (who said C65?) they had the money.

While competitors didn't move a single finger during that period, right?

Do you really believe on what you write?

 Status: Offline
Profile     Report this post  
bhabbott 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 7:39:31
#289 ]
Regular Member
Joined: 6-Jun-2018
Posts: 330
From: Aotearoa

@cdimauro

Quote:

cdimauro wrote:

Very likely the palette is fixed so the conversion from chunky to planar/HAM8 is much faster.

10k colors it said. This is not a palette, it's true color.

Quote:
HAM8 needs an optimized palette to give its best quality. It means that quite some time is needed to scan the original 24-bit image and calculate a set of good 64 base colors.

In Superhires you can have have any color you like (out of 262144) in 1.5 Hires pixels, without using the base colors. What I suspect is being done is not a fixed palette, but fixed RGB HAM selection., ie. the 1st pixel always sets red, 2nd green, 3rd blue etc. No calculations required, just blast the 18 bit RGB image into the 6 HAM bitplanes.

Quote:
As you can see from the video, the quality is really poor and doesn't show so many colors.

Doesn't look 'really poor' to me. The only poor part is the monitor it is displayed on (and nasty moire patterns and flicker etc. caused by filming it with a digital camera).

Quote:
Maybe the used textures weren't that much good and/or the scene doesn't give justice to the technique.

Most of the scene is mapped with a single texture which is just a test pattern, so yes it 'doesn't give justice' to the technique. But this is a technical demo, not an artistic one.

 Status: Offline
Profile     Report this post  
agami 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 7:55:44
#290 ]
Super Member
Joined: 30-Jun-2008
Posts: 1637
From: Melbourne, Australia

@cdimauro

Quote:
Do you really believe on what you write?

Belief has nothing to do with it. It's reality.

Just because Commodore bought Amiga, doesn't mean that they stopped making everything else. The C64 sold about 1 million units in 1985 and also in 1986. The Amiga 1000 had poor availability in 1985, and sales are estimated to be 35,000 units @ $1,300 USD in 1986, representing gross revenue of $45.5M USD.

After spending $27M USD on buying Amiga in 1984, they would be looking to recover that cost, plus the cost of developing the A1000, probably another $10M. It's optimistic to measure net profit of A1000 sales as 30%, making it $13.5M.

C64 units in 1986 represented $149M with net likely to be 50%, representing $74.5M

CBM R&D budget would be allocated across all initiatives. How much more do you think the bean counters were willing to put into the Amiga before it recovered the initial outlay? The A2000 and A500 were only ever going to be variations on the exact same tech.

Only after the A500 started moving units, was it justifiable to invest more in the Amiga line of computers. CBM mismanaged most parts of their business, and the Amiga line was no exception, but their C-suite did follow basic MBA rules.

Also:
There are single-person projects today.
While any platform that has been around for more than a handful of years evolves, they do not all evolve at the same pace. Apple was stuck on the Apple II for quite a while, then they struggled again in the mid '90s and were mere weeks from insolvency. The ATX platform has seen very little evolution for a few decades now, UI conventions are stagnant, smartphone and tablet evolution has pretty much stopped. The same goes for mirrorless cameras, game consoles, ICE cars, airline airframes, etc.

_________________
All the way, with 68k

 Status: Offline
Profile     Report this post  
Karlos 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 8:19:27
#291 ]
Elite Member
Joined: 24-Aug-2003
Posts: 4394
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@bhabbott

Correct, it's an RGB raycaster. The palette is completely irrelevant. He's using Kalm's HAM8 C2P routines. If you look carefully you'll also note he's doing MIP mapping too. As you approach the walls the textures change. That's what the numbers indicate.

I do wonder which CPU it's running on, but the code is all in github at least.

Last edited by Karlos on 12-Oct-2022 at 08:31 AM.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
bhabbott 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 8:52:10
#292 ]
Regular Member
Joined: 6-Jun-2018
Posts: 330
From: Aotearoa

@cdimauro

Quote:

cdimauro wrote:

Are you serious? On '80s there were single-man projects whereas today we several hundreds of engineers for a single project.

Many projects today are still done by small teams. The difference is that the tools are much better and cheaper than they used to be. The original Amiga prototype (Lorraine) was made from thousands of standard TTL ICs, hand assembled on large wire-wrap boards. It took many months to build. Today you just write some HDL on your PC and simulate it, then download the bitstream into an FPGA for hardware testing. Any amateur can do that today with free software and an FPGA dev board. Might need a digital scope to debug it, and that's about it. You can design the PCB with free software too, and have prototypes made in days for a few bucks.

There's an old saying - too many cooks spoil the broth. Intel had 500 engineers working on Itanium for many years, and it still failed. They might have done better with a small team of more focused engineers.

Quote:
Well, actually you've to read this: On the Edge: The Spectacular Rise and Fall of Commodore

Specifically on the part where he was talking about the engineers that remained when the ones that created the Amiga left the company.

And?

It didn't stop them from designing the A500, A3000, CDTV, A600, A4000, A1200, CD32, several accelerator cards, hard drive controllers etc. All this shows is that lack of talent wasn't the problem, it's just that most of the engineers were working on real products, not fantasies.

Quote:
Quote:
Engineers are always working on something. But CBM was a business, and as much as it irks me, engineering is subservient to the business, which is in turn subservient to the market.

Sure. And it wasn't the same for all other companies & engineers involved on different, concurrent projects?

What's not clear about what I've already reported before:

"ALL other platforms evolved over the years bringing better hardware technologies."

That is a lie.

I have only have to provide one example to refute your claim, and here it is:-

The Amstrad CPC464 was released in 1984, and became very popular in the same time period as the Amiga. How much did its hardware 'evolve'? In 1985 the CPC664 got a disk drive, using the same circuit as the external unit for the 464 with a bog-standard design. In the same year the CPC6128 arrived, with RAM doubled to 128k and a small PAL to bank switch it (I designed a compatible circuit for my 664, but using 256k DRAMs for double the memory of the 6128). The main custom chip remained practically the same. The CPC received no further 'better hardware technologies' until 1990 (6 years later) with the CPC Plus range. So what did it get? Sprites, better hardware scrolling, 4096 color palette, a DMA channel for the AY-3-8912 sound chip. How was this better than the competition? Answer, it wasn't. The Amiga had already done it all (and more) many years before.



Last edited by bhabbott on 12-Oct-2022 at 08:57 AM.
Last edited by bhabbott on 12-Oct-2022 at 08:53 AM.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 21:09:24
#293 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@bhabbott

Quote:

bhabbott wrote:
@cdimauro

Quote:

cdimauro wrote:

Very likely the palette is fixed so the conversion from chunky to planar/HAM8 is much faster.

10k colors it said. This is not a palette, it's true color.

Bruce, I don't understand what kind of knowledge do you have about the Amiga hardware.

HAM8 isn't true color AND it USE a palette of 64 colors. Similar to HAM, which had a base palette of 16 colors.
Quote:
Quote:
HAM8 needs an optimized palette to give its best quality. It means that quite some time is needed to scan the original 24-bit image and calculate a set of good 64 base colors.

In Superhires you can have have any color you like (out of 262144) in 1.5 Hires pixels, without using the base colors. What I suspect is being done is not a fixed palette, but fixed RGB HAM selection., ie. the 1st pixel always sets red, 2nd green, 3rd blue etc. No calculations required, just blast the 18 bit RGB image into the 6 HAM bitplanes.

See above: HAM8 uses a palette, like HAM.

And to get good results you have to carefully select its colors, otherwise the graphic artifacts become much more evident.
Quote:
Quote:
As you can see from the video, the quality is really poor and doesn't show so many colors.

Doesn't look 'really poor' to me. The only poor part is the monitor it is displayed on (and nasty moire patterns and flicker etc. caused by filming it with a digital camera).

It's poor besides the camera which was used for recording the video.
Quote:
Quote:
Maybe the used textures weren't that much good and/or the scene doesn't give justice to the technique.

Most of the scene is mapped with a single texture which is just a test pattern, so yes it 'doesn't give justice' to the technique. But this is a technical demo, not an artistic one.

Then let's see when there's an artistic demo, because currently the results don't look good.

@Karlos

Quote:

Karlos wrote:
@bhabbott

Correct, it's an RGB raycaster. The palette is completely irrelevant.

With HAM and HAM8 the palette IS relevant. Those aren't true color modes: they use a fixed palette as a "base" to reduce the color components transitions and consequently to reduce the graphic artifacts that are generated by those transitions.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 21:15:03
#294 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@agami

Quote:

agami wrote:
@cdimauro

Quote:
Do you really believe on what you write?

Belief has nothing to do with it. It's reality.

Just because Commodore bought Amiga, doesn't mean that they stopped making everything else. The C64 sold about 1 million units in 1985 and also in 1986. The Amiga 1000 had poor availability in 1985, and sales are estimated to be 35,000 units @ $1,300 USD in 1986, representing gross revenue of $45.5M USD.

After spending $27M USD on buying Amiga in 1984, they would be looking to recover that cost, plus the cost of developing the A1000, probably another $10M. It's optimistic to measure net profit of A1000 sales as 30%, making it $13.5M.

C64 units in 1986 represented $149M with net likely to be 50%, representing $74.5M

CBM R&D budget would be allocated across all initiatives. How much more do you think the bean counters were willing to put into the Amiga before it recovered the initial outlay? The A2000 and A500 were only ever going to be variations on the exact same tech.

Only after the A500 started moving units, was it justifiable to invest more in the Amiga line of computers. CBM mismanaged most parts of their business, and the Amiga line was no exception, but their C-suite did follow basic MBA rules.

Then how do you justify the Ranger chipset development, since he started already on 1986?
Quote:
Also:
There are single-person projects today.

Not for projects like a CPU, a GPU, a chipset, etc., which are sold worldwide in big markets.
Quote:
While any platform that has been around for more than a handful of years evolves, they do not all evolve at the same pace. Apple was stuck on the Apple II for quite a while, then they struggled again in the mid '90s and were mere weeks from insolvency. The ATX platform has seen very little evolution for a few decades now, UI conventions are stagnant, smartphone and tablet evolution has pretty much stopped. The same goes for mirrorless cameras, game consoles, ICE cars, airline airframes, etc.

The argument was about a chipset evolution: graphic, audio, I/O, etc..

Take a look at what Commodore / Amiga competitors (Apple, Atari, Acorn, PCs) did in the same period of time.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 21:27:52
#295 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3621
From: Germany

@bhabbott

Quote:

bhabbott wrote:
@cdimauro

Quote:

cdimauro wrote:

Are you serious? On '80s there were single-man projects whereas today we several hundreds of engineers for a single project.

Many projects today are still done by small teams.

Not at the mainstream level.

Remember that we were talking about the Amiga chipset: a system which was sold worldwide as personal computer. The parallel should have been about the same components that today are found on personal computer (and by extension to smartphones and tablets, which acquired a good part of the PCs market).
Quote:
The difference is that the tools are much better and cheaper than they used to be. The original Amiga prototype (Lorraine) was made from thousands of standard TTL ICs, hand assembled on large wire-wrap boards. It took many months to build. Today you just write some HDL on your PC and simulate it, then download the bitstream into an FPGA for hardware testing. Any amateur can do that today with free software and an FPGA dev board. Might need a digital scope to debug it, and that's about it. You can design the PCB with free software too, and have prototypes made in days for a few bucks.

But we weren't taking of amateur projects, rather by mainstream components: see above.
Quote:
There's an old saying - too many cooks spoil the broth. Intel had 500 engineers working on Itanium for many years, and it still failed.

Irrelevant. And a logical fallacy: this means and changes nothing about the discussed topic.
Quote:
They might have done better with a small team of more focused engineers.

Another logical fallacy: you have nothing which proves it. It's just speculation, without any foundation.
Quote:
Quote:
Well, actually you've to read this: On the Edge: The Spectacular Rise and Fall of Commodore

Specifically on the part where he was talking about the engineers that remained when the ones that created the Amiga left the company.

And?

It didn't stop them from designing the A500, A3000, CDTV, A600, A4000, A1200, CD32, several accelerator cards, hard drive controllers etc. All this shows is that lack of talent wasn't the problem, it's just that most of the engineers were working on real products, not fantasies.

And the topic was the development of the CHIPSET. Understood?
Quote:
Quote:
Sure. And it wasn't the same for all other companies & engineers involved on different, concurrent projects?

What's not clear about what I've already reported before:

"ALL other platforms evolved over the years bringing better hardware technologies."

That is a lie.

Then you've to prove it.
Quote:
I have only have to provide one example to refute your claim, and here it is:-

The Amstrad CPC464 was released in 1984, and became very popular in the same time period as the Amiga. How much did its hardware 'evolve'? In 1985 the CPC664 got a disk drive, using the same circuit as the external unit for the 464 with a bog-standard design. In the same year the CPC6128 arrived, with RAM doubled to 128k and a small PAL to bank switch it (I designed a compatible circuit for my 664, but using 256k DRAMs for double the memory of the 6128). The main custom chip remained practically the same. The CPC received no further 'better hardware technologies' until 1990 (6 years later) with the CPC Plus range. So what did it get? Sprites, better hardware scrolling, 4096 color palette, a DMA channel for the AY-3-8912 sound chip. How was this better than the competition? Answer, it wasn't. The Amiga had already done it all (and more) many years before.

So, according to you the Amstrad CPC464 was an Amiga competitor. In which parallel universe?

Here is it: https://en.wikipedia.org/wiki/Amstrad_CPC#CPC464
The Amstrad CPC (short for Colour Personal Computer) is a series of 8-bit home computers produced by Amstrad between 1984 and 1990. It was designed to compete in the mid-1980s home computer market dominated by the Commodore 64 and the Sinclair ZX Spectrum

The Amiga competitors were Apple Macintosh, Atari ST, Acorn Archimedes and, of course, PCs.

In fact: https://en.wikipedia.org/wiki/Amiga
The original model is one of a number of mid-1980s computers with 16- or 32-bit processors, 256 KB or more of RAM, mouse-based GUIs, and significantly improved graphics and audio compared to previous 8-bit systems. This includes the Atari ST—released earlier the same year—as well as the Macintosh and Acorn Archimedes.
[...]
Poor marketing and the failure of later models to repeat the technological advances of the first systems resulted in Commodore quickly losing market share to the rapidly dropping prices of IBM PC compatibles, which gained 256 color graphics in 1987


As usual, you don't know of what you were talking about!

 Status: Offline
Profile     Report this post  
Karlos 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 21:37:03
#296 ]
Elite Member
Joined: 24-Aug-2003
Posts: 4394
From: As-sassin-aaate! As-sassin-aaate! Ooh! We forgot the ammunition!

@cdimauro

I do understand what HAM is and how it works but for the purposes of the C2P method being used, the base palette is not something you set set up for artistic purposes. The process here is completely unlike the processes normally involved for displaying a 24 bit image on a HAM screen at the same resolution.

The graphics in the demo are rendered in 24 bit RGB and that is in turn converted to the best HAM8 representation. 4 consecutive superhires HAM8 output pixels are used per 1 lores RGB input. There's no fringing effects to worry about here, you are essentially using the superhires pixels as subpixels.

I downloaded the repo and tested it in UAE where it's super smooth. I suspect you'll need an 060 to get useable frame rates otherwise.

Last edited by Karlos on 12-Oct-2022 at 09:44 PM.

_________________
Doing stupid things for fun...

 Status: Offline
Profile     Report this post  
bhabbott 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 22:55:22
#297 ]
Regular Member
Joined: 6-Jun-2018
Posts: 330
From: Aotearoa

@cdimauro

Quote:

cdimauro wrote:

HAM8 isn't true color AND it USE a palette of 64 colors.

You don't have to use the base colors, and then you have 3 possibilities left for each pixel:- set red / hold green and blue, set green / hold red and blue, set blue / hold red and green. It takes 3 pixels to fully change from one 18 bit 'true' (not paletted) color to another. This produces color fringing on vertical edges, but in Superhires the fringe is only 1.5 hires pixels. On a standard TV tube this is about the same size as the dot pitch. In composite its invisible.

Here's something interesting about how the Amiga got HAM mode...
Quote:
Amiga project lead Jay Miner relates:

"Hold and Modify came from a trip to see flight simulators in action and I had a kind of idea about a primitive type of virtual reality. NTSC on the chip meant you could hold the hue and change the luminance by only altering four bits. When we changed to RGB I said that wasn't needed any more as it wasn't useful and I asked the chip layout guy to take it off. He came back and said that this would either leave a big hole in the middle of the chip or take a three-month redesign and we couldn't do that."

3 months just to take a feature off the chip! With today's design tools it would only take a few minutes.

Quote:
HAM8 needs an optimized palette to give its best quality. It means that quite some time is needed to scan the original 24-bit image and calculate a set of good 64 base colors.

Only if you need the absolute 'best possible' quality.

"Oh no!", says the Amiga fan, "The technique used didn't produce the absolute best possible quality! Not acceptable!!!". Hyper-perfectionist Amiga fans, always enemies of the good.

 Status: Offline
Profile     Report this post  
matthey 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 12-Oct-2022 23:41:35
#298 ]
Super Member
Joined: 14-Mar-2007
Posts: 1968
From: Kansas

cdimauro Quote:

OK, so it proves that it wasn't a paper project.


It's strong evidence anyway. It's interesting that the Amiga Corporation developers used the boing ball on the Ranger prototype instead of the checkmark used on the Amiga 1000. In pursuit of ASIC custom chips, I found none for Ranger, not even planned.

https://obligement-free-fr.translate.goog/articles/chipsetamiga.php?_x_tr_sch=http&_x_tr_sl=auto&_x_tr_tl=en&_x_tr_hl=en&_x_tr_pto=wapp

There were at least planned and named custom chips for AA+ and Hombre although Ranger plans may have been hidden away or kept separate from normal CBM activities and documentation. AA+ docs go back to May 1992 which shows how bad of financial shape CBM was in already that they never produced custom chips for AA+ or Hombre (there were custom chips produced for AAA). AA+ development was likely straightforward as much of the logic from AAA development could be backported to AA+. Transistor estimates for AA+ seem suspect to me though.

ECS - 3 chips 60,000 transistors
AGA - 3 chips ? transistors (Lisa 80,000 transistors, Alice 80,000 transistors)
AA+ - 2 chips 200,000 transistors
AAA 32 bit - 4 chips 750,000 transistors
AAA 64 bit - 6 chips 1,000,000 transistors

The link above has a 512kiB Agnus using 21.000 transistors while the 2MiB AGA upgrade Alice uses 80,000 transistors. The AA+ Ariel chip which combines the Alice and upgraded Paula (likely taken from AAA), uses only 100,000 transistors. This suggests that a 16 bit Paula with 8 voices used at most 20,000 transistors. While the memory access logic of Ariel was likely similar to AGA Alice, this suggests the address range upgrade from 2MiB to 4MiB had a very cheap transistor cost. This would also suggest that the Ranger upgrade from 512kiB to 2MiB had a cheap transistor cost. If looks like Paula (10,000-15,000 transistors?) could have been combined with Agnus/Alice much earlier or better, upgraded and combined seeing how 16 bit 8 voice audio likely used fewer than 20,000 transistors which is cheap already in the late '80s and cheaper yet in the '90s. With the Paula transistor estimate, we can estimate AGA used ~175,000 transistors. AA+ using 200,000 transistors likely could have been done in 1992 if CBM had it ready and may have cost less being 2 chips instead of 1. Sony was also planning for the Playstation in 1992 and understood the importance of integration.

https://en.wikipedia.org/wiki/PlayStation_(console)#Inception Quote:

To determine the fate of the PlayStation project, 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.


There are several transistors per logic gate so "his LSI chip" was several million transistors. This required moving to a more expensive chip processes like AAA required. AA+ using ~200,000 transistors would have been low cost and CBM could have surely moved to a better process and made it one chip, even with a 68k CPU included if they wanted a single chip Amiga SoC.

68020 200,000 + 200,000 = 400,000 transistors
68030 273,000 + 200,000 = 473,000
68040 1,200,000 + 200,000 = 1,400,000
68060 2,500,000 + 200,000 = 2,700,000

An Amiga SoC with at least 68030 and AA+ specs should have been feasible in 1992 (AIM Agreement was in 1991 so 68k licensing should have been reasonable). AAA would have cost more to produced just the chipset than AA+ with a 68030. CBM did integrate the CPU with the chipset in the case of Hombre but was planning on 2 chips (no transistor counts given). Hombre was to receive all the exotic hardware enhancements refused for the C64 replacement Amiga though. The HP PA-7150@125MHz with only 2kiB of I cache planned is going to fall on its face as that is like having a 512 byte cache on the 68k, a similar I cache miss rate as a 68020. It's those caches which use transistors. The 68060 8kiB I+D caches use 786,432 transistors which is nearly 1/3 of the CPU transistors. With RISC, they may end up needing a 3rd chip for Hombre. That's what HP did with the PA-8000 moving all the caches onto an external cache chip which was very expensive.

cdimauro Quote:

Then was Haynie wrong?


You would think Dave Haynie would know about Ranger considering he was designing AAA. At the very least, part of the Ranger design could have been reused like for VRAM access. Jay Miner said he finished the Ranger design and gave it to CBM before he left. Maybe CBM never authorized Ranger development and Jay started developing anyway. However, it sounds like Los Gatos and CBM Germany were in a competition for the high end Amiga so I don't know. The Ranger specs appear to have been deeply buried at CBM and at least some of the prototypes appear to have stayed with Los Gatos developers (Dale Luck). There was a lot of animosity with the Los Gatos developers. Remember the "We made Amiga, They ####ed it up" which really pissed off CBM managment?

We made Amiga, they ####ed it up (ENGLISH)
https://www.youtube.com/watch?v=omMOuyTLmyg

CBM pulled tens of thousands of Amigas off the shelves and replaced the ROMs while losing 3 months of Amiga sales in a major market. This incident may have resulted in Los Gatos being closed.

cdimauro Quote:

Indeed. But there was time to reduce its complexity: this was just a prototype.


Yea, the board layout just needs to be changed. Rerouting can be done automatically today although some people still prefer to do it by hand.

cdimauro Quote:

Too late and Ranger chipset provided too limited innovations.

...

Unfortunately it didn't provide "full color". Only 128 colors, that were too few on '89.


Ranger would have been better than VGA in 1988-1989 and would have stood up against SVGA until the next generation of Amiga chipset was available. The chipset performance was like AGA but not as colorful. Still, 128 colors looks almost as good as 256. It would have been better if 8 bit plane EHB with 128 color registers and HAM8 could have been included (shouldn't have needed major changes so maybe could have been added). Higher resolutions would have allowed better SVGA compatibility and monitor use while 640x400/480 would likely have been usable for some games. I think fewer customers would have been lost to the competition compared to ECS in the late '80s and early '90s. There was better and more expensive hardware but this would have moved the base Amiga up quite a bit like AGA did too late. It may have given CBM time to see that customers wanted upgraded Amigas which they saw too late with AGA while they were still producing 68000+ECS Amigas (a major factor in their downfall). Ranger would have likely made the 68020 standard as well. Ranger was designed to work with the 68010 also but by 1988-1989 the more powerful 68020 had dropped considerably in price and the 68010 was clearly on the way out.

 Status: Offline
Profile     Report this post  
agami 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 13-Oct-2022 1:01:14
#299 ]
Super Member
Joined: 30-Jun-2008
Posts: 1637
From: Melbourne, Australia

@cdimauro

Quote:
Then how do you justify the Ranger chipset development, since he started already on 1986?

As I said, engineers are always working on something. And just because the new Amiga team in Commodore received some funding in those early years, does not mean that they received sufficient funding to deliver a new chipset at speed.
The fact that Ranger never went past an early prototype is indicative of the "tinkering" project that it was.

You have no idea how much I wish the purchase of Amiga by CBM was a strategic one, instead of the tactical move to deny Atari the tech.
Having a talented team and bunch of patents, without a clear direction and purpose meant that CBM were going to play some Spaghetti-Wall. They were not a chipset company before that, and they were certainly not magically transformed into a chipset company simply by purchasing Amiga.

Neither were Apple, Atari, or Acorn, chipset companies. They were computer companies like Commodore, that for the most part developed computers using "graphics, audio, I/O, etc." from other companies who's business was to focus on those components. In the nascent days of the personal computer revolution, it was not uncommon to discover that none of the component OEMs had a part that fits a requirement, so some customisation is commissioned. That doesn't make these companies chipset companies.

The PC certainly doesn't count as it was not a single company, and it was the original Open Computing Platform.

You can focus on chipsets if you like, but ultimately it comes down to business management. Companies that managed R&D better, released improvements at a faster pace. Companies like Commodore that were not very good at managing R&D, released improvements at a slower pace. It's really not that complicated.

I've mentioned it before: Commodore should've kept up with Moore's Law and maintained a 2-year release cadence. The engineers I'm sure could've manage it, but that would require some serious business acumen; something Commodore never had.

Things that make it challenging for a single person to develop and bring to market CPU/GPU innovation in 21C:
- Moore's Law has increased the complexity since 1985 by 2E+19. Sure, we have much more advanced tools, and the costs of production are much more affordable, but
- We are not in the "Wild West" of the '80s personal computer market. Things are a lot more commoditised, and beyond the academic curiosity, a single person would be going up against multi-national giants in the industry.
- The broken patent system also keeps most single persons from even trying to start a CPU/GPU innovation project. They might have the engineering skills, but they know they are clueless about IP law and have read enough ugly stories in the media to dissuade them from going out on their own, and then choose to join one of the big firms instead.
- The definition of base infrastructure has changed. Where in the '80s and most of the '90s it was about coding directly to HW, today it's all about the APIs. Remember in the '90s when even the average consumer cared about what GPU was in their game console? 16-bit, 32-bit, 64-bit. Today, I'm not even sure that most game devs care about what GPU is in the Xbox Series X or PS5. Can it Run UE5? is the main question.

Last edited by agami on 13-Oct-2022 at 01:07 AM.
Last edited by agami on 13-Oct-2022 at 01:03 AM.

_________________
All the way, with 68k

 Status: Offline
Profile     Report this post  
bhabbott 
Re: How good or bad was the AGA chipset in 1992/1993.
Posted on 13-Oct-2022 1:22:54
#300 ]
Regular Member
Joined: 6-Jun-2018
Posts: 330
From: Aotearoa

@cdimauro

Quote:

cdimauro wrote:

So, according to you the Amstrad CPC464 was an Amiga competitor. In which parallel universe?

In the 'parallel universe' that I lived in. In 1985 I bought a CPC664. I could have had an A1000 if I tried hard enough (they weren't sold in New Zealand at the time), but the Amstrad was a better fit to my needs at the time. 2 years later I did buy an A1000, but I kept using the Amstrad for several years after that, partly because there was an active local user group and nothing for the Amiga.

Quote:
The Amiga competitors were Apple Macintosh, Atari ST, Acorn Archimedes and, of course, PCs.

First off, I don't class the PC as a 'home computer' in the traditional sense. In New Zealand it was used almost solely by businesses. The Acorn Archimedes was hardly competitive. It was very expensive here and had little support. No retail stores sold it. Ditto for the Apple Macintosh. These machines filled a very niche market here (schools and universities).

But let's take a look at the Archimedes:-
Quote:
The first models were released in June 1987, as the 300 and 400 series. The 400 series included four expansion slots and an ST-506 controller for an internal hard drive...

Speculation gathered pace about new machines in the Archimedes range in early 1989, with commentators envisaging a low-cost, cut-down model... The new model sported only a single internal expansion slot, which was physically different from that of the earlier models.

In late 1991, the A5000 was launched to replace the A440/1 machine in the existing product range. With the existing A400/1 series regarded as "a little tired", being largely unchanged from the A400 models introduced four years previously,... The A5000 featured the new 25 MHz ARM3 processor, 2 or 4 MB of RAM, either a 40 MB or an 80 MB hard drive and a more conventional pizza box-style two-part case... offering resolutions of 1024 x 768 in 16 or 256 colours and with 24-bit palettes... The A5000 (along with the earlier A540) supported the SVGA resolution of 800 x 600 in 16 colours, although the observation that "Archimedes machines have simply not kept pace" arguably remained...

In 1992, several new models were introduced to complement the A3000 and to replace the low-end A400 series models - the A3010, A3020 and A4000... The ARM250, running at a higher 12 MHz clock frequency and used in conjunction with faster 80ns memory chips, compared to the 8 MHz of the ARM2 and the 125ns memory of the A3000, gave a potential 50% performance increase over such older systems

A bit late to the party, but by 1992 the Archimedes equivalent of the A1200 and A4000 sported a mere 50% performance increase and no significant change to the graphics and sound subsystems in 5 years. 'Evolution'? Not so much.

"But but", you say, "their high-end models evolved much more than the Amiga!"

Quote:
The introduction of the 68040... put ARM3-based Archimedes models at an increasing performance disadvantage. An Amiga 4000 with 68040 CPU (or suitably upgraded Amiga 2000) could achieve a reported 18.7 - 21.6 VAX MIPS... Against such performance ratings only Acorn's Risc PC 600 (18.4 VAX MIPS[360] to 21.8 VAX MIPS[363]) fitted with an ARM610 CPU would be able to respond.


The ST was competition for sure. So how much did it 'evolve'?
Quote:

In late 1989, Atari released the 520STE and 1040STE (also written STE), enhanced version of the ST with improvements to the multimedia hardware and operating system. It features an increased color palette of 4,096 colors from the ST's 512 (though the maximum displayable palette without programming tricks is still limited to 16 in the lowest 320 × 200 resolution, and even fewer in higher resolutions), Genlock support, and a blitter co-processor (stylized as "BLiTTER") which can quickly move large blocks of data (particularly, graphics data) around in RAM. The STE is the first Atari with PCM audio; using a new chip, it added the ability to play back 8-bit (signed) samples at 6258 Hz, 12517 Hz, 25033 Hz, and even 50066 Hz, via direct memory access (DMA). The channels are arranged as either a mono track or a track of LRLRLRLR... bytes. RAM is now much more simply upgradable via SIMMs.

4 years of 'evolution', copying features the Amiga already had and not even getting to OCS level.

But...
Quote:

The final model of ST computer is the Falcon030. Like the TT, it is 68030-based, at 16 MHz, but with improved video modes and an on-board Motorola 56001 audio digital signal processor. Like the Atari STE, it supports sampling frequencies above 44.1 kHz; the sampling master clock is 98340 Hz (which can be divided by a number between 2 and 16 to get the actual sampling frequencies). It can play the STE sample frequencies (up to 50066 Hz) in 8 or 16 bit, mono or stereo, all by using the same DMA interface as the STE, with a few additions. It can both play back and record samples, with 8 mono channels and 4 stereo channels, allowing musicians to use it for recording to hard drive. Although the 68030 microprocessor can use 32-bit memory, the Falcon uses a 16-bit bus, which reduces performance and cost. In another cost-reduction measure, Atari shipped the Falcon in an inexpensive case much like that of the STF and STE. Aftermarket upgrade kits allow it to be put in a desktop or rack-mount case, with the keyboard separate.

Released in 1992, the Falcon was discontinued by Atari the following year.

Finally some real evolution! After 7 years. But still not much competition, as it didn't even last a year. By this time AGA Amigas were out.

But even before then Commodore had 'evolved' the Amiga by a fair bit, with ChipRAM increased to 2MB, productivity mode, flicker fixer, 020 and 030 accelerators, 7 port serial card and Ethernet card for the A2000, hard drive + FastRAM for the A500, PC in an addon box for the A1000 and 'Bridgeboard' cards for the A2000. If we consider 3rd party devices (as we would for the PC) then by 1992 the Amiga had a 40MHz 040 with buffered serial port and PC compatible parallel port, 24 bit graphics at up to 1600x1200 with 440 MB/s graphics bus bandwidth and 50 MB/s CPU bus bandwidth, 16 bit audio with 8 tracks at 44kHz and direct to disk recording, Video CD/CDI MPEG 1 video decoding and playback.

 Status: Offline
Profile     Report this post  
Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 | 16 | 17 | 18 | 19 | 20 | 21 | 22 | 23 | 24 | 25 | 26 | 27 | 28 | 29 | 30 | 31 | 32 | 33 | 34 | 35 | 36 | 37 | 38 | 39 | 40 | 41 | 42 | 43 | 44 | 45 Next Page )

[ home ][ about us ][ privacy ] [ forums ][ classifieds ] [ links ][ news archive ] [ link to us ][ user account ]
Copyright (C) 2000 - 2019 Amigaworld.net.
Amigaworld.net was originally founded by David Doyle