Things the Amiga didn't get right from Day 1

[Thomas Richter]

Originally Posted by sandruzzo

@roondar

By heavly using copper, and dmas' cicles waste.. Think about if blitter could have been done this by itself..

As you wish:

http://eab.abime.net/attachment.php?...1&d=1595076698

The same with the blitter. Of course, then only 4 colors total, not using the copper. This is again simply the graphics.library VSprite engine.

[roondar]

Originally Posted by sandruzzo

I still think that some part of Amiga HW ARE unfinished. It miss so simple optimizations...

It's always simple to find optimisations after something is released (especially after living with it for 30+ years). It's not so simple to know them ahead of time.

I've never been convinced the Amiga was unfinished as such. Could things have been done better (now that we know what we know), sure... But that's not the same thing.

[sandruzzo]

Remember that Amiga was born as Computer, not a pure consolle, so A Little ram for only 68k would have done a lot of difference. Think about saving registers into stack...

[roondar]

Yes it would've made a small difference but it would also have meant making the motherboard quite a bit more complicated (on what bus do you want this memory?).

But more importantly, for 1985 the memory bandwidth offered by the Amiga was great considering the target market. It really wasn't a problem that needed fixing. Yes, later machines probably ought to have had fast memory as a standard. But for 1985 the 7MB/sec offered by the Amiga was no slough.

[sandruzzo]

@roondar

Just 32-64k same chip memory(only for 68k), but with all slots accessible to 68k. Huge difference

[roondar]

Originally Posted by sandruzzo

Just 32-64k same chip memory(only for 68k), but with all slots accessible to 68k. Huge difference

Yup, it's a notable difference in terms of motherboard layout for a small gain. Fast RAM is not panacea

[Thomas Richter]

Originally Posted by sandruzzo

@roondar

Just 32-64k same chip memory(only for 68k), but with all slots accessible to 68k. Huge difference

That was simply not an option with the technology from back then. You could either have chip memory, with the chipset stealing some cycles, or fast mem, with all cycles available. The chip memory bandwidth is the maximum bandwidth available with the technology available at its time.

The construction of a shared graphics/CPU memory is not so uncommon. It was the norm in the home compute area, and was pretty much identical to what the Atari 8bits did. Even nowadays, laptops with chipset graphics use a shared memory model, and it does have some advantages sometimes as graphics data does not have to be transported over PCIe.

[sandruzzo]

I remember when I plugged a590 to my A500: another computer! If you see Amiga as PC, yes fast memory is a must, as console, less important.

[roondar]

There's nothing stopping you from adding fast memory to an A500 with an A590, it has support for on-board fast ram IIRC

[sandruzzo]

Using bitplanes' DMA cicles for trap Door RAM, doesnt make sense as CPU' access to chip ram every other cicle. 68000 14mhz was there as 020 and 030 near to be released. Very pope coiche

[Thomas Richter]

Originally Posted by sandruzzo

Using bitplanes' DMA cicles for trap Door RAM, doesnt make sense as CPU' access to chip ram every other cicle. 68000 14mhz was there as 020 and 030 near to be released. Very pope coiche

Not quite sure what you want to say. Ranger Mem was a very poor choice. It was a "RAM expansion done as cheap as possible", combining the draw-backs of chip and fast ram in one go. They only made the refresh wider in Agnus, but they did not free the missing cycles on Agnus. The 1MB version of Agnus then fixed that by adding another bit to the DMA engines, allowing to extend chip memory.

[sandruzzo]

Better would have been to that RAM outside bitplanes'DMA cicles and let the other chipset and. Cpu able to access to It. Not a very expensive thing to do...

[roondar]

I think you're underestimating the cost of these things.

Basically, the CPU would need to have a certain amount of lines connected to the trapdoor. But, these lines are already connected to the side slot. You can't simply add some more traces and expect that to work. Having two connections to the CPU is not that easy. Furthermore, there are real world limitations (more so in the 1980's than now, but still) on how long traces to RAM can be. This is why the RAM is in the spot it is - close to both CPU and Agnus.

It obviously can be done, but it would almost certainly equire extra components or a complete redesign of the bus controller (i.e. Agnus). If the latter is too expensive (and it probably is as Commodore never went that route for any version of the Agnus/Alice chips), it may well need a new motherboard layout as the current one has the CPU close to the side slot / internal RAM and far away from the trapdoor slot.

Really, the A500 was a cost cutting measure. That's why it was cheap to buy and that was why it was successful. It never was about being the fastest system.

---
Edit: I may have misunderstood you. You might have meant that the CPU should be able to access parts of the internal RAM as if it were fast RAM.

That would still have pretty big problems to overcome. You're either looking at redesigning Agnus, or inserting additional bus logic between the CPU and RAM and also still connecting the CPU to Agnus (which would require a motherboard redesign).

Basically, none of what you propose is actually all that simple and I don't see it as being particularly cheap to do.

[sandruzzo]

@roondar

I mean to add a little extra ram, only for cpu, or just sell A500 with little fast ram as expansion card.

[roondar]

I understand what you mean, I'm trying to explain that building this onto the motherboard in the way you described it is not actually easy.

Now, if you simply want to have a standard "side slot" fast ram expansion included by default, that would be much more feasible. However, that would mean this card has just become mandatory.

[sandruzzo]

Think having cpu on that slot with "fast mem". You could upgrade it as nothing

at least have cpu able to access to all chip mem slots instead every otherwould have been great,s ince faster cpu were there.

[AmigaHope]

The joystick issue could've been solved easily by shipping 2-3 button joysticks stock with the system. They wouldn't even have to be good joysticks. They just needed to establish the standard. The only joysticks C= ever shipped with their system, going back to the A1000, were 1-button joysticks.

The A1000 shipped with a 2-button mouse. They could have shipped a 3-button mouse, but they shipped with a 2-button one. Afterwards all software that used a mouse assumed a 2-button mouse. Nobody supported a 1-button mouse, or ever wrote software that assumed you had only 1 button, moreover, almost no software supported a 3-button mouse.

Take the Sega Genesis/Megadrive. It shipped with a 4-button pad. Much, much later they shipped a 7-button pad but 90% of games only used 4 buttons even after that.

It required no change in hardware, it only required that initial establishing standard.

Regarding the color registers, etc. for 8-bit video, they really should've been forward thinking and left reserved bits for the future. No idea why they didn't do that, it would have been easy and had almost no overhead (this is a HUGE advantage of the bitplane model, that larger depths only required expanded registers rather than different pixel formats in memory).

Regarding a PSG, C= could have easily included a couple of SIDs and it would have sounded great. They had the advantage of owning the design and owning their own fab to produce them. It would have been cheap for them. Cheaper than buying even cheap PSGs from Yamaha or whoever. Imagine how awesome Amiga music would have sounded with SID + Paula. If they had scrapped some of their stupider projects like the C264 series they would have easily freed up fab capacity to do this, maybe even make a "Super SID" that had multiple SID circuits on the same die like Bob Yannes originally wanted. Check out what you can do when you combine SIDs:

[ Show youtube player ]

Regarding the CPU, maybe not the A1000 as parts that could go that fast were brand new back then, but the A500 could have included a 14Mhz 68000. The problem would be keeping it fed with RAM fast enough to keep it happy, which would be expensive, but you could add a small cache and then have the rest of it work with wait states. This would still speed up operations a lot.

Regarding MIDI ports, in hindsight it earned the ST a solid niche in musicians' studios but nobody knew that in 1985. The convenience factor vs a dongle is what sold it. It could've easily been added with just a few passive components and a switch, which would be cheap, but again I don't know if they could have predicted it.

Multithreaded Workbench would've been a huge plus, maybe with a more featureful details window like DOpus, but that would've actually cost development $$$ and a clear vision that wasn't totally apparent at the time. Norton Commander, probably the first of its kind, wasn't released until 1986, and DOpus which really refined it far and above Commander (to the point where Commander copied it) wasn't released until 1990.

[squidbass]

Regarding midi - even with a midi adaptor such as MidiMaster the amiga could not compete with the ST because the atari had Pro software such as Steinberg/Cubase, Creator/Notator. Those programs were used in just about every Recording Studio in the 80s.

Ironically there is an unreleased version of Cubase that runs on Amiga but it's buggy as hell.

Rgds,

[McTrinsic]

Hen and egg?

The ST had the Software you mention due to the MIDI interface that was available as standard.

[Thomas Richter]

Originally Posted by AmigaHope

The joystick issue could've been solved easily by shipping 2-3 button joysticks stock with the system. They wouldn't even have to be good joysticks. They just needed to establish the standard. The only joysticks C= ever shipped with their system, going back to the A1000, were 1-button joysticks.

CBM was, always, a cheap company. Including a joystick would have made the system more expensive. It was rather unclear to them, however, how to sell the machine. The original A1000 package was more that of a business machine. Mouse and keyboard included, joystick not included.

Originally Posted by AmigaHope

Regarding the color registers, etc. for 8-bit video, they really should've been forward thinking and left reserved bits for the future. No idea why they didn't do that, it would have been easy and had almost no overhead (this is a HUGE advantage of the bitplane model, that larger depths only required expanded registers rather than different pixel formats in memory).

The problem did not appear with OCS. It appeared with ECS where registers were added behind the color registers. But then again, those registers could have been considered private, and it wasn't the application to poke them, but rather the operating system, so anything could have been added later. Actually, AAA was designed to be only ECS compatible, and any extended registers would have been elsewhere.

Anyhow, it is not a showstopper, just a matter of an orthogonal register layout.

Originally Posted by AmigaHope

Regarding a PSG, C= could have easily included a couple of SIDs and it would have sounded great.

Different bus. You cannot easily ft a SID into an Amiga system. It would have required additional interface logic, or a redevelopment of SID, none of which CBM had resources for. Actually, the Amiga was not an in-house product for them in first place as CBM accquired the hardware design more or less in last minute, with the hardware already complete. Actually, it is an Atari-type design.

Originally Posted by AmigaHope

Regarding the CPU, maybe not the A1000 as parts that could go that fast were brand new back then, but the A500 could have included a 14Mhz 68000.

As said, this would have complicated the design as then Agnus has also to allocate DMA slots for the faster CPU, hence would have required a new Agnus design. However, CBM never "redesigned" much of the system. The development costs for such a system would be much higher than that of the "budget" A500 system which only collected some gates in advanced custom chips, without extending their capabilities.

Originally Posted by AmigaHope

Multithreaded Workbench would've been a huge plus, maybe with a more featureful details window like DOpus, but that would've actually cost development $$$ and a clear vision that wasn't totally apparent at the time.

But the workbench is multithreaded. You mean the "loading part"? Or the "copy part"? Program-wise, this is trivial enough, but it would have meant more ROM space, and that was already tight.