Why did Newtek abandon the hobbyist computer graphic artist?

[alexh]

Originally Posted by fxgogo

Don't confuse low price with low quality. Their products were used in professional studios as well as low end facilities.

Low end professional studios. High end professional studios used software and equipment from companies such as Quantel.

[ImmortalA1000]

Originally Posted by alexh

Low end professional studios. High end professional studios used software and equipment from companies such as Quantel.

The thing with the Amiga 1000 + Digi-view & Digi-paint was it was that or a 5 figure sum bespoke bit of kit back in 1986. I used to watch educational documentaries and there was one on the BBC about the Quantel Paintbox which I found fascinating. Beyond all other reasons HAM mode and Digi-view was the biggest reason I got an Amiga 1000.

There was nothing in between, a gulf of about £15,000 minimum if you wanted better than HAM on an Amiga 1000.

I think it's the 1986 Xmas episode of Micro Live where they use some really expensive PC based system with a 24bit graphics card to show some Photoshop style editing of a pie to remove the cut out slice etc and they mentioned the price.

Loads of artists and advertising studios used an Amiga, they did something called animatics for proposed advertising projects IIRC.

It was a revolution that people forget, they think we went from nowhere directly to Photoshop, which of course is completely wrong.

[alexh]

Originally Posted by ImmortalA1000

It was a revolution that people forget, they think we went from nowhere directly to Photoshop, which of course is completely wrong.

There is an era you're forgetting/don't know about that overlaps the Amiga era and is Apple Macs. The DTP and video editing software on Macs soon became significantly superior. Time meant everything. You could easily save the price differential of a Mac in a month by improved productivity.

[pandy71]

Originally Posted by ImmortalA1000

There was nothing in between, a gulf of about £15,000 minimum if you wanted better than HAM on an Amiga 1000.

Well... it is not entirely correct - there was some solutions such as https://www.computer.org/publication...visions-targa/ - i own 2 such first generations board... but there is nothing there beyond framebuffer and flash ADC component board so card is able to capture component, display picture stored in video memory and eventually overlay it (or rather sync to external component source) - it was not so expensive but also limited as ISA bus was too slow to feed this board data to perform 30 fps (even if you have some external ridiculous size sufficiently fast storage). This was time where silicone video disk was size of 42U 19 inch rack full of huge boards filled tightly with some 4164 DRAM chips to create capability to play like 100 seconds of video... this cost millions $ in those times and was area restricted for companies such as Quantel, to play 12 seconds of video you could buy something smaller [ Show youtube player ] - this shows technology limitations from times when Amiga was born - huge gap between 64Kib and 1Mib DRAMS but still impressive.

[fxgogo]

Originally Posted by alexh

There is an era you're forgetting/don't know about that overlaps the Amiga era and is Apple Macs. The DTP and video editing software on Macs soon became significantly superior. Time meant everything. You could easily save the price differential of a Mac in a month by improved productivity.

Sure Mac owned the DTP space, but when are you talking about video editing space, things are not so clear cut. It was 1989 when the first systems for PC and Mac appeared and they were extremely expensive and could just squeeze out quality for offline edits. Broadcast quality non-linear editing only really became a reality for most facilities in the mid to late nineties, by which time Amiga was in its death throws.

Regarding Quantel, that certainly was very high end, but was not the exclusive tool for high end studios. There were many low cost tools, Mac included, that got used for high end productions.

[Pyromania]

NewTek didn’t abandon the Amiga. Commodore died and abandoned NewTek. Thankfully NewTek survived.

[Bruce Abbott]

Originally Posted by ImmortalA1000

It was a revolution that people forget, they think we went from nowhere directly to Photoshop, which of course is completely wrong.

But 'we' did (ie. people who had Macs or PCs).

Tim Jenison chose the Amiga in 1985 because it was designed to produce broadcast standard NTSC and was genlockable, features no other home computer had at the time. Of course a PC or Mac could do it with the right hardware addons, but they didn't have it built in so there was less incentive to exploit it (similar to how the ST had MIDI built in, making it more attractive to musicians).

But most PC and Mac users were barely aware that the Amiga existed, and certainly wouldn't buy one just to do a bit of video editing. So from their perspective there was no 'revolution' until it hit their preferred platform.

There was nothing in between, a gulf of about £15,000 minimum if you wanted better than HAM on an Amiga 1000.

Yes, anything was 'available' on other platforms if you had enough money, but you needed an Amiga to do it on the cheap. Of course this simply added to the perception that the Amiga was 'unprofessional', so PC and Mac fans pooh poohed it as they scrambled to match it.

[grelbfarlk]

It's funny, Newtek tried hard to promote their platform on the Amiga to Mac fans. They actually put a label over the A2000 case and sold it as a dumb switcher to the Mac crowd, you operated the Toaster over Toasterlink.
The Toaster 4000 also included stickers to cover up the A4000 badge.

[pandy71]

Originally Posted by Bruce Abbott

Of course a PC or Mac could do it with the right hardware addons, but they didn't have it built in so there was less incentive to exploit it (similar to how the ST had MIDI built in, making it more attractive to musicians).

Mac definitely was not capable to accommodate video extensions comparable with Amiga until Mac II where possibility to add video hardware was introduced in 1987.

PC was not equipped with bus capable to feed sufficiently fast data to hypothetical video HW so or you need some HW acceleration in video HW (never standardized and introduced quite late when compared to others) or forget about multimedia (practical ISA limit is approx 2..3MBps and usually CPU have no direct access to video memory).

Amiga advantage and curse at the same time was UMA architecture - something not present in PC world till Intel 810 chipset and Mac albeit being UMA (Mac I) was very primitive video HW incapable to compete with Amiga. At many points Amiga was close to Personal Workstation than Mac or PC philosophy.

[Bruce Abbott]

Originally Posted by pandy71

Mac definitely was not capable to accommodate video extensions comparable with Amiga until Mac II where possibility to add video hardware was introduced in 1987.

PC was not equipped with bus capable to feed sufficiently fast data to hypothetical video HW so or you need some HW acceleration in video HW (never standardized and introduced quite late when compared to others) or forget about multimedia (practical ISA limit is approx 2..3MBps and usually CPU have no direct access to video memory).

Er, that's not quite right. Video RAM was memory mapped on CGA and EGA, not that it mattered much because you wouldn't want to use those cards for video work. And this is my point. The PC/AT had a 16 bit bus that could transfer data to/from the CPU at about the same rate as the Amiga, but the stock machine didn't have good enough graphics.

Theoretically you could plug in a card to do anything you wanted on the PC, which was often done when a specific need was identified. IBM introduced the Professional Graphics Adapter in 1984. It did 640×480 with 256 colors from a palette of 4,096 colors, and had an onboard CPU to accelerate drawing operations. But it was optimized for CAD, not animation. The 'card' consisted of 3 full length boards sandwiched together, and cost a whopping US$3000 (over twice the price of an A1000).

In 1985 the Mac was even worse, with a closed architecture and underperforming hardware - and it was twice the price of an A1000. But it did have a neat graphical interface and a serial port, so you could connect it to an external box that did video stuff - if anyone bothered to make such a thing.

Amiga advantage and curse at the same time was UMA architecture...

It was a big advantage for doing fast animation etc. at low cost, not so much for adding more powerful graphics hardware. But the A1000 did expose the entire CPU bus which was designed to be expandable. The disadvantage over the PC was that it didn't have bus slots built in. This oversight was corrected with the A2000.

The 'curse' as you say is that with excellent graphics hardware built in there was little incentive to make the OS hardware-agnostic. In contrast the PC was designed to take any graphics card you put in it, with the OS only requiring it to do basic text. There was no limit to what the PC could do with the right plug in cards. The Amiga wasn't limited in that way either, but it was perceived to be by most users because you couldn't just yank out the chipset and replace it with more advanced hardware.

The other huge advantage the PC had was those 3 letters I, B, and M. The buisiness world went gaga for it from day one, and the by the time the Amiga arrived 'personal computer' meant IBM compatible. With dozens of manufactuers competing to produce the most advanced hardware at the lowest price there was no way any other platform could compete for long.

PCs eventually went UMA too, but with much more powerful CPUs etc. on dirt-cheap motherboards. I remember when the first 'SIS' brand motherboards with shared memory came out. Performance wasn't great compared to a good graphics card, but they were cheap and so quite popular. Over time they got so much better that only hardcore gamers need dedicated graphics cards today.

[Pyromania]

The PC had disadvantages too like the letters MSFT.

[Bruce Abbott]

Originally Posted by Pyromania

The PC had disadvantages too like the letters MSFT.

Microsoft had their hooks into many home computers too. That's how they were able to court IBM. Don Estridge, head of IBM's PC development team, chose Microsoft over IBM's own BASIC because "Microsoft BASIC had hundreds of thousands of users around the world. How are you going to argue with that?".

I was never impressed with Microsoft BASIC on 8 bit home computers. It used floating point for everything and the code was crammed into the smallest possible ROM, which made it slow and lacking in features. If you wanted full support for your machine you often had to buy an 'extended' BASIC just to get a few extra commands (= more $ for Microsoft!).

Amiga BASIC looked good until you used it. Writing structured code without line numbers was great, except that the editor was painfully slow. Not having a proper file requester didn't help, though in fairness a lot of early Amiga programs didn't either (and those that did were often clunky). But the worst sin was ignoring Commodore's programming guidelines, which made it incompatible with accelerator cards. It also had an incorrectly encoded instruction that made it crash on the 68020. I'm pretty sure it was ported from the Macintosh version (or at least used the same codebase).

[Thomas Richter]

Originally Posted by Bruce Abbott

Er, that's not quite right. Video RAM was memory mapped on CGA and EGA, not that it mattered much because you wouldn't want to use those cards for video work.

Well, sort of. video data does not sit in the same RAM the CPU executes from, and there is the graphics controller that sits between the video RAM and the ISA-bus that allows you to select the read page and the write page(s). You cannot realistically store instructions in video RAM as the would vanish or alter on reprogramming the GC. The actual video RAM sits on the card, not near the CPU. So I would no really call that UMA. It is memory mapped, but your average super VGA chip also uses memory mapped video RAM, though it does no longer attempt to squeeze it into the IBM PC VGA video RAM window.

Originally Posted by Bruce Abbott

In 1985 the Mac was even worse, with a closed architecture and underperforming hardware - and it was twice the price of an A1000. But it did have a neat graphical interface and a serial port, so you could connect it to an external box that did video stuff - if anyone bothered to make such a thing.

The Mac hardware had an UMA video architecture, with the graphics framebuffer sharing RAM with the CPU (at actually two different frame buffer locations). No hardware acceleration, only black and white initially (one-bit framebuffer) a fairly primitive thing. However, it had a well-designed (unlike our "graphics") graphics package aka "Quickdraw" in its ROM.

Originally Posted by Bruce Abbott

PCs eventually went UMA too, but with much more powerful CPUs etc. on dirt-cheap motherboards. I remember when the first 'SIS' brand motherboards with shared memory came out. Performance wasn't great compared to a good graphics card, but they were cheap and so quite popular. Over time they got so much better that only hardcore gamers need dedicated graphics cards today.

Most "integrated graphics" CPUs have a UMA architecture, too, as of today. They are generally slower, but good enough for every-day use. However, you need to set aside the framebuffer RAM from the beginning, and it is then set aside and isolated from the rest of the RAM, not a relocatable framebuffer as on the Amiga. The big advantage is that any on-chip GPU can directly access data in CPU RAM, and textures etc. do not have to be moved through the PCIe bottleneck. This is for some applications even an advantage. There are some dedicated NVidia boards for GPU computing that make use of this architecture.

[Thomas Richter]

Originally Posted by Bruce Abbott

I was never impressed with Microsoft BASIC on 8 bit home computers. It used floating point for everything and the code was crammed into the smallest possible ROM, which made it slow and lacking in features. If you wanted full support for your machine you often had to buy an 'extended' BASIC just to get a few extra commands (= more $ for Microsoft!).

That was not all so uncommon back then. Most Basics had "float for everything", to address the letter "B" in Basic (as in "Beginner"). This problem also existed for other non-M$-Basic dialects, e.g. Atari Basic or the TI-Basic. M$ choose at least a sane floating point format (unlike Atari Basic, which uses floating point with the basis of 100, not 2. Yes, really). However, Atari Basic or TI-Basic offered some statements for sound and graphics support, even on the 8-bit machines back then. M$ Basic was pretty bare-bone.

Originally Posted by Bruce Abbott

Amiga BASIC looked good until you used it. Writing structured code without line numbers was great, except that the editor was painfully slow. Not having a proper file requester didn't help, though in fairness a lot of early Amiga programs didn't either (and those that did were often clunky).

Back then, the Os did not offer a standard requester, so programs had to implement its own, often with "mixed" success. The first de-facto standard requester was probably that of arp, but this happened a lot later. The AmigaBasic editor was really a PITA. I never really the Basic and bought GFA Basic later on, which was at least a lot faster (but also a lot buggier).

Originally Posted by Bruce Abbott

But the worst sin was ignoring Commodore's programming guidelines, which made it incompatible with accelerator cards. It also had an incorrectly encoded instruction that made it crash on the 68020. I'm pretty sure it was ported from the Macintosh version (or at least used the same codebase).

I agree. It looks very similar to MS-Basic on the Mac, and its memory management problems are likely a consequence of implementation details of the MacOs handle system which stores handle attributes in the top 8 bits of an address. There was, I believe, a revised version of Microsoft Basic, the CBM bug database has traces of it in multiple places, but it was never officially released as CBM would have to pay royalties for each sold copy and CBM wanted to go cheap. Instead, they choose to steal ARexx and ship that instead without paying anything to its creator.... Oh well.

[pandy71]

Originally Posted by Bruce Abbott

Er, that's not quite right. Video RAM was memory mapped on CGA and EGA, not that it mattered much because you wouldn't want to use those cards for video work. And this is my point. The PC/AT had a 16 bit bus that could transfer data to/from the CPU at about the same rate as the Amiga, but the stock machine didn't have good enough graphics.

Thomas already wrote most of explanations

There is no video RAM accessible by CPU in PC until introducing UMA, it may be accessible indirectly or as RAM in some defined periods (usually during VBlank) or only indirectly by telling CRTC to alter video RAM state.
So Amiga is less restricted to PC when you have access to video RAM.
This was my point.

ISA is usually slower as it is being shared between other resources and those resources allocating fully ISA (not like in Amiga where some resources may access bus only in particular moments).
Classic example is floppy DMA access (to uPD765) when error occur...
Usually bus is congested fully by waiting on floppy response.
This freeze whole PC even in Windows XP times - luckily to PC floppy are no longer issue as even newest Windows will be frozen by this nasty heritage.

Originally Posted by Bruce Abbott

Theoretically you could plug in a card to do anything you wanted on the PC, which was often done when a specific need was identified. IBM introduced the Professional Graphics Adapter in 1984. It did 640×480 with 256 colors from a palette of 4,096 colors, and had an onboard CPU to accelerate drawing operations. But it was optimized for CAD, not animation. The 'card' consisted of 3 full length boards sandwiched together, and cost a whopping US$3000 (over twice the price of an A1000).

Yes, PGA is one of examples where IBM used general CPU instead dedicated HW graphic accelerators (such as uPD7220 or many others way faster)...
Anyway it shows technology available from Amige design times...

Originally Posted by Bruce Abbott

In 1985 the Mac was even worse, with a closed architecture and underperforming hardware - and it was twice the price of an A1000. But it did have a neat graphical interface and a serial port, so you could connect it to an external box that did video stuff - if anyone bothered to make such a thing.

Ah yes, you could hook it to some graphic terminal or decent computer like Amiga... S. Jobs was never technical guy (his visionary skills are highly overrated) and in Apple software was always cheaper than HW.

Originally Posted by Bruce Abbott

It was a big advantage for doing fast animation etc. at low cost, not so much for adding more powerful graphics hardware. But the A1000 did expose the entire CPU bus which was designed to be expandable. The disadvantage over the PC was that it didn't have bus slots built in. This oversight was corrected with the A2000.

Well you could integrate native Amiga capabilities with external graphics in many ways (at SW and/or HW level) - if you recall PC grpahic adapters plenty of them was equipped with something called 'Feature connector' so definitely problem could be solved even in not very expensive way.

Originally Posted by Bruce Abbott

The 'curse' as you say is that with excellent graphics hardware built in there was little incentive to make the OS hardware-agnostic. In contrast the PC was designed to take any graphics card you put in it, with the OS only requiring it to do basic text. There was no limit to what the PC could do with the right plug in cards. The Amiga wasn't limited in that way either, but it was perceived to be by most users because you couldn't just yank out the chipset and replace it with more advanced hardware.

Of course you could but at what cost and what you get in exchange...
I don't blame Amiga for having HW accelerated graphic and OS build around this functionality as for very long time this was advantage and many of us was educated by Amiga to expect more than just regular PC usage.

Originally Posted by Bruce Abbott

The other huge advantage the PC had was those 3 letters I, B, and M. The buisiness world went gaga for it from day one, and the by the time the Amiga arrived 'personal computer' meant IBM compatible. With dozens of manufactuers competing to produce the most advanced hardware at the lowest price there was no way any other platform could compete for long.

Yes but also primitivity of PC was beneficial - you could do things that in Amiga world demanded some skills, knowledge and cost more (even doing Autoconfig require programmable logic where PC bunch of regular TTL's is sufficient)

Originally Posted by Bruce Abbott

PCs eventually went UMA too, but with much more powerful CPUs etc. on dirt-cheap motherboards. I remember when the first 'SIS' brand motherboards with shared memory came out. Performance wasn't great compared to a good graphics card, but they were cheap and so quite popular. Over time they got so much better that only hardcore gamers need dedicated graphics cards today.

Yes, first approach toward UMA was Vesa Local Bus but thanks to faster DRAM's (SDRAM and newer) at some point you could share some RAM bandwidth with CPU and Video...

[ImmortalA1000]

Originally Posted by alexh

There is an era you're forgetting/don't know about that overlaps the Amiga era and is Apple Macs. The DTP and video editing software on Macs soon became significantly superior. Time meant everything. You could easily save the price differential of a Mac in a month by improved productivity.

Macintosh II with the upgraded graphics and faster VRAM needed for such work was about $6500 alone. Probably didn't come out in the UK until about 1988. 3 years might not sound like much but 3 years is the difference between C64 graphics and Amiga 1000 graphics There were alternatives before the Amiga 1000 PAL launch, Mac II didn't cause any revolution in digital art in the 80s at all, it's a bit more crap than Quantel, it's a bit cheaper for a full set up with digitizers, scanners and software, there were other alternatives too. But at this point we are talking the difference in cost between a BMW 5 series executive model or the cheapest BMW 3 series in a dealer showroom and 2 years after you could go down to the shops and get Amiga + Digi-view etc. When exactly did the first Mac II digitizers even come out, flatbed scanner is fine for photographs but as an artist you might need to grab a real scene, or even a model scene etc, not a photo. If you are going to take a photo then you need a decent camera (about £150 minimum in mid-late 1980s) and wait for rolls of film to be developed, hardly the creative users ideal way of working.

We are still talking about 1000% more expensive than a 1987/88 A500 1mb Digi-view+Digi-paint based setup anyway. Like I said it's not really a revolution if it was late to the party and did nothing for ordinary people who wanted to do creative things with their computer. The Amiga is the only real world creatively empowering computer of the 80s, by the time you could buy an affordable colour Mac we are talking mid 1990s, by which time Newtek had abandoned us.

Digi-view served it's purpose, without Amiga OR Newtek there would be no remotely affordable budding digital artists in the 80s, like I said a forgotten revolution, a match made in heaven. I wonder what Jay Miner made of Digi-view taking his 'left over' HAM mode and doing so much with it.

[Bruce Abbott]

Originally Posted by pandy71

Thomas already wrote most of explanations

There is no video RAM accessible by CPU in PC until introducing UMA, it may be accessible indirectly or as RAM in some defined periods (usually during VBlank) or only indirectly by telling CRTC to alter video RAM state.

This is not true. Video RAM in most cards was memory mapped, and the CPU could access it at any time. The original CGA card suffered from 'snow' in 80 column text mode if accessed during the display period because the CPU had priority. As the amount of RAM on the card increased they started implementing bank switching to avoid taking up too much of the memory map, but that is no different to EMS memory.

So Amiga is less restricted to PC when you have access to video RAM.

The Amiga's CPU also has limited access because display DMA has priority. ChipRAM is on a separate bus from the CPU, and access to it is controlled by Agnus. This is not that different to eg. a CGA card plugged into a PC.

The big difference between the PC and Amiga was that the PC could have any video card installed, or even none at all. The small amount of RAM on early display adapters meant that using it for other purposes was not reliable or useful. So by convention the RAM on the card was treated as separate even though it was part of the memory map.

In contrast the Amiga had a large amount of 'video' RAM built in, and no other RAM, so the OS was 'hard wired' to use it. Later on when accelerator cards became popular, efforts were made to move the OS stuff out of ChipRAM. However location 4 is execbase, which is the only fixed address in the OS that all libraries are referenced from. So every Amiga has to have at least a bit of ChipRAM accessible by the CPU, especially on startup when other RAM may not be switched in.

That 'hard wiring' of the OS to use the onboard chipset and RAM became a big problem for switching to a plug in card. Graphics library had a number of functions that assumed a particular architecture, and the OS called some of those functions internally, making it difficult to redirect to a different architecture via the API. And of course coders who bypassed the OS assumed a particular hardware that was directly accessed by the application. This meant that changing the video hardware was much harder than on the PC. But it was a software problem, not hardware.

ISA is usually slower as it is being shared between other resources and those resources allocating fully ISA (not like in Amiga where some resources may access bus only in particular moments).
Classic example is floppy DMA access (to uPD765) when error occur...
Usually bus is congested fully by waiting on floppy response.

You think that's bad - on the PCJr and JX (which I had) there was no DMA controller so PIO mode was used. During disk I/O everything was locked out. If you typed on the keyboard keystrokes would get lost.

On a standard PC disk DMA is interleaved with CPU activity with very little overheard since the disk only transfers a data byte every 16us or so. AFAIK the reason for the machine 'freezing' during disk errors is simply poor implementation. DOS was not multitasking so it was OK to 'freeze' while handling disk errors, and it could safely busy-wait on the controller chip when recalibrating etc.

The Amiga puts a lot of effort into keeping multitasking going during disk activities, even though it needs more CPU intervention to step the heads etc. That's why you may hear the step rate go down when the CPU is heavily loaded by other tasks.

S. Jobs was never technical guy (his visionary skills are highly overrated) and in Apple software was always cheaper than HW.

Yes, and Macs were slowed down by it. Furthermore many of their machines were not that fast. The original Macintosh had a 68000 clocked at 7.8336 MHz, but effectively ran at 6MHz, 25% slower than the Amiga. But their QuickDraw graphics library was highly optimized and provided excellent virtualization. Apple strongly discouraged coders from accessing the hardware directly, making it easier to develop more advanced video systems.

BTW while it's true that Jobs was not a hardware guy, his visionary skills are not overrated. The problem with hardware guys is they are always thinking about the hardware as an end in itself rather than a means to an end. Computer hardware engineers are the worst because they are designing general purpose hardware for someone else to develop stuff on, not application specific products. OS engineers have the same problem. Steve Jobs saved Apple by giving them his innovative NextStep OS, after they failed to finish their Copland behemoth. It wasn't skilled coders they lacked, but vision.

Well you could integrate native Amiga capabilities with external graphics in many ways (at SW and/or HW level) - if you recall PC grpahic adapters plenty of them was equipped with something called 'Feature connector' so definitely problem could be solved even in not very expensive way.

Sure you could, in fact the A3000 did that with its built-in flicker fixer. I had a Picasso-II card on my A3000, which passed through the ficker-fixed native video to get seamless operation on a single monitor.

[pandy71]

Originally Posted by Bruce Abbott

This is not true. Video RAM in most cards was memory mapped, and the CPU could access it at any time. The original CGA card suffered from 'snow' in 80 column text mode if accessed during the display period because the CPU had priority. As the amount of RAM on the card increased they started implementing bank switching to avoid taking up too much of the memory map, but that is no different to EMS memory.

Yes, graphic local video memory has allocated defined address window, it is not physically located on CPU side, there is no synchronization between video and CPU, some cards provide limited access but most of them separate CPU from video RAM as much as possible.
More fancier graphic controller more CPU access is limited.

Originally Posted by Bruce Abbott

The Amiga's CPU also has limited access because display DMA has priority. ChipRAM is on a separate bus from the CPU, and access to it is controlled by Agnus. This is not that different to eg. a CGA card plugged into a PC.

CPU and DMA cycles are interleaved so unless graphic saturate limited RAM bandwidth then CPU may access memory (code or video data).
By design this fundamentally different approach than on PC. (UMA vs NUMA).

Originally Posted by Bruce Abbott

The big difference between the PC and Amiga was that the PC could have any video card installed, or even none at all. The small amount of RAM on early display adapters meant that using it for other purposes was not reliable or useful. So by convention the RAM on the card was treated as separate even though it was part of the memory map.

Not sure how to understand above - technically you don't need to use Amiga chipset graphic too... and RAM on PC graphic board was not used by anything else (till modern chipsets and modern graphic cards where some software allow to use such RAM as for example RAM DISK - not highly popular solution on PC but possible)

Originally Posted by Bruce Abbott

In contrast the Amiga had a large amount of 'video' RAM built in, and no other RAM, so the OS was 'hard wired' to use it. Later on when accelerator cards became popular, efforts were made to move the OS stuff out of ChipRAM. However location 4 is execbase, which is the only fixed address in the OS that all libraries are referenced from. So every Amiga has to have at least a bit of ChipRAM accessible by the CPU, especially on startup when other RAM may not be switched in.

Yes, but i don't think this is something very bad when compared to PC.

Originally Posted by Bruce Abbott

That 'hard wiring' of the OS to use the onboard chipset and RAM became a big problem for switching to a plug in card. Graphics library had a number of functions that assumed a particular architecture, and the OS called some of those functions internally, making it difficult to redirect to a different architecture via the API. And of course coders who bypassed the OS assumed a particular hardware that was directly accessed by the application. This meant that changing the video hardware was much harder than on the PC. But it was a software problem, not hardware.

I disagree - see no big problem in designing separate HW path to combine both worlds. I also understand that OS friendly applications usually operating at some HAL so eventually rewriting graphic.library is possible.

Originally Posted by Bruce Abbott

You think that's bad - on the PCJr and JX (which I had) there was no DMA controller so PIO mode was used. During disk I/O everything was locked out. If you typed on the keyboard keystrokes would get lost.

On a standard PC disk DMA is interleaved with CPU activity with very little overheard since the disk only transfers a data byte every 16us or so. AFAIK the reason for the machine 'freezing' during disk errors is simply poor implementation. DOS was not multitasking so it was OK to 'freeze' while handling disk errors, and it could safely busy-wait on the controller chip when recalibrating etc.

DOS is not a problem, problem is NT - don't follow MS Windows too closely so not sure when MS removed support for floppies (it is removed?) but it was very bad to have frozen modern multitasking system (XP for sure, not sure about Vista/Win7).

Originally Posted by Bruce Abbott

The Amiga puts a lot of effort into keeping multitasking going during disk activities, even though it needs more CPU intervention to step the heads etc. That's why you may hear the step rate go down when the CPU is heavily loaded by other tasks.

Well this is obvious but fdd step itself is not CPU consuming, overall CPU load is high and as such overall slowdown.

Originally Posted by Bruce Abbott

Yes, and Macs were slowed down by it. Furthermore many of their machines were not that fast. The original Macintosh had a 68000 clocked at 7.8336 MHz, but effectively ran at 6MHz, 25% slower than the Amiga. But their QuickDraw graphics library was highly optimized and provided excellent virtualization. Apple strongly discouraged coders from accessing the hardware directly, making it easier to develop more advanced video systems.

Yes, QuickDraw was impressive piece of software and Apple was not only one that discouraged HW bit banging.

Originally Posted by Bruce Abbott

BTW while it's true that Jobs was not a hardware guy, his visionary skills are not overrated. The problem with hardware guys is they are always thinking about the hardware as an end in itself rather than a means to an end. Computer hardware engineers are the worst because they are designing general purpose hardware for someone else to develop stuff on, not application specific products. OS engineers have the same problem. Steve Jobs saved Apple by giving them his innovative NextStep OS, after they failed to finish their Copland behemoth. It wasn't skilled coders they lacked, but vision.

Well... don;t get me wrong but Jobs is overrated... how someone can be visionary by refusing user to get color desktop only because printers are black&white

Originally Posted by Bruce Abbott

Sure you could, in fact the A3000 did that with its built-in flicker fixer. I had a Picasso-II card on my A3000, which passed through the ficker-fixed native video to get seamless operation on a single monitor.

I have no experience with Picasso in A3000 so i assume this is similar solution to Voodoo bypass (VGA In, VGA Out combined at analog level and eventually synced by H and V sync). But in case Amiga you could in thory integrate legacy chipset in single IC and output it in digital form or trough data channel (DMA video over PCI to graphic RAM and overlay it on top of RTG screen) or digital video combine trough for example Feature Connector with RTG video, definitely this would provide legacy compatibility and open way for RTG. It was possible in 92..95 - in fact Lisa was HP made due of CGS limitations - single step to combine all PAULA, ALICE and LISA with some of AKIKO (CIA's) elements into single IC capable to output video data over PCI.

[Bruce Abbott]

Originally Posted by pandy71

Yes, graphic local video memory has allocated defined address window, it is not physically located on CPU side, there is no synchronization between video and CPU, some cards provide limited access but most of them separate CPU from video RAM as much as possible.
More fancier graphic controller more CPU access is limited.

I'm talking about CGA etc., not modern graphics cards. ChipRAM on the Amiga is also 'not on CPU side'. Run hires 16 colors in OCS and the CPU is locked out of the ChipRAM bus during the entire active display period. Other things like the system ROM, CIA chips and Zorro cards are connected directly to the CPU bus so they are not affected.

The CPU and graphics system on the Amiga both run off the same master clock. But so did CGA. Check out this photo of an original CGA card - what do you not see? That's right, a crystal. All timing was derived from the motherboard clock.

CPU and DMA cycles are interleaved so unless graphic saturate limited RAM bandwidth then CPU may access memory (code or video data).
By design this fundamentally different approach than on PC. (UMA vs NUMA).

This also happens on CGA when in graphics modes.

The 68000 has the advantage of not using the first half of each bus cycle, so it 'interleaves' well most of the time. However some instructions use a few extra clocks and then it has to wait. 68020 and above don't have this 'dead' time so they have to wait on every other cycle, which is why the A1200 goes twice as fast when FastRAM is installed. Most accelerator cards are asynchronous, so if you have an accelerated Amiga it probably isn't in any way 'unified'.

Memory on the Amiga isn't unified in the same way as it is on a modern PC. Chip RAM is behind Agnus, on a separate bus to FastRAM. If you plug in more RAM it will be on one bus or the other, and can't straddle them. You can't go into the BIOS and say, I want this much to be Chip RAM and the rest Fast RAM.

Not sure how to understand above - technically you don't need to use Amiga chipset graphic too... and RAM on PC graphic board was not used by anything else (till modern chipsets and modern graphic cards where some software allow to use such RAM as for example RAM DISK - not highly popular solution on PC but possible)

This is purely by convention. There is no technical reason that memory on a CGA card couldn't have been used as general purpose RAM. Difference on the Amiga was you didn't have any other RAM so there was no choice.

I disagree - see no big problem in designing separate HW path to combine both worlds. I also understand that OS friendly applications usually operating at some HAL so eventually rewriting graphic.library is possible.

Yes, but it's not an easy task. When Amiga OS was originally developed no thought was put into the possibility of different graphics hardware, and much was done inside the ROM that wasn't exposed to the API.

To make matters worse a lot of things were not 'finished', temping developers to create hacky workarounds that are very 'brittle'. All this can be fixed, and hopefully will be in an upcoming version of AmigaOS 3, but it won't be easy. Some apps are bound to break.

The important to realize is that this is purely a software issue. Whether the hardware is 'unified' or not is irrelevant, just as it is in PCs. Unfortunately attempts by 3rd parties to make Amiga graphics hardware agnostic were very hacky, and now we are stuck with them (CyberGrafX, Picasso 96).

[pandy71]

Originally Posted by Bruce Abbott

I'm talking about CGA etc., not modern graphics cards. ChipRAM on the Amiga is also 'not on CPU side'. Run hires 16 colors in OCS and the CPU is locked out of the ChipRAM bus during the entire active display period. Other things like the system ROM, CIA chips and Zorro cards are connected directly to the CPU bus so they are not affected.

The CPU and graphics system on the Amiga both run off the same master clock. But so did CGA. Check out this photo of an original CGA card - what do you not see? That's right, a crystal. All timing was derived from the motherboard clock.

Once again - CGA run asynchronously from main CPU - it use MB XTAL but beside this there no synchronization between CPU and 6845 - you can query Status register at $3DA and check status bits but not much above this - there is no HSync, VSync information so CPU is not aware which line, which part of line is active - CPU may only check status bit - in theory you can reprogram 8254 to generate IRQ and be more or less on safe side of timing but still this is only indirect way. This is something fundamentally different than Amiga design.
And yes i agree that UMA is limited only to chipset part but this is more than 90% of standard Amiga configurations as such.

Originally Posted by Bruce Abbott

This also happens on CGA when in graphics modes.

Not same as on Amiga.

Originally Posted by Bruce Abbott

The 68000 has the advantage of not using the first half of each bus cycle, so it 'interleaves' well most of the time. However some instructions use a few extra clocks and then it has to wait. 68020 and above don't have this 'dead' time so they have to wait on every other cycle, which is why the A1200 goes twice as fast when FastRAM is installed. Most accelerator cards are asynchronous, so if you have an accelerated Amiga it probably isn't in any way 'unified'.

Memory on the Amiga isn't unified in the same way as it is on a modern PC. Chip RAM is behind Agnus, on a separate bus to FastRAM. If you plug in more RAM it will be on one bus or the other, and can't straddle them. You can't go into the BIOS and say, I want this much to be Chip RAM and the rest Fast RAM.

True, never claimed anything opposite.

Originally Posted by Bruce Abbott

This is purely by convention. There is no technical reason that memory on a CGA card couldn't have been used as general purpose RAM. Difference on the Amiga was you didn't have any other RAM so there was no choice.

Yes, it can be used but usually latency will be higher than MB RAM due ISA.

Originally Posted by Bruce Abbott

Yes, but it's not an easy task. When Amiga OS was originally developed no thought was put into the possibility of different graphics hardware, and much was done inside the ROM that wasn't exposed to the API.

To make matters worse a lot of things were not 'finished', temping developers to create hacky workarounds that are very 'brittle'. All this can be fixed, and hopefully will be in an upcoming version of AmigaOS 3, but it won't be easy. Some apps are bound to break.

Have no clue about Amiga OS developers thoughts - i can imagine ROM was limited size, time was also tight, as Amiga was closed architecture till A2000 then no need of external graphic too. Probably it was then sane approach to solve most of the problems in such way.
This is why i think keeping legacy separated is probably best solution - new application should sue newgraphic.library where legacy application can use old graphic.library and perhaps do happy bit banging as glue logic will combine legacy layer with new one.

New CPU ISA seem to be perfect sharp transition point- sadly Commodore died and as such EOT

Originally Posted by Bruce Abbott

The important to realize is that this is purely a software issue. Whether the hardware is 'unified' or not is irrelevant, just as it is in PCs. Unfortunately attempts by 3rd parties to make Amiga graphics hardware agnostic were very hacky, and now we are stuck with them (CyberGrafX, Picasso 96).

Agree but technically... this could be solved quite easy if Commodore was interested...