The Amiga: Why did it fail so hard in the United States?

[JamesD]

Again, I still go with the thesis that the Amiga did about as well as it was going to in the US given the clearly inevitable rise of PC Compatibles, which just happened to happen a little faster here than it did the rest of the world (other major markets like Europe, Japan, South America, etc., had fairly different trajectories, but still eventually ended up at the same place). There's no shame in being the second most popular computer platform for a time, no matter how brief, and no matter how distant of a second. With that said, it's clear that, if Commodore had executed better, the Amiga could have probably secured more marketshare in its prime leading up to the early 90s and lasted with at least some miniscule marketshare up to roughly the introduction of Windows 95. Again, Commodore's fate was inevitable, unless they pivoted to become a premiere PC Compatible maker, and even then, who knows if the Commodore name would have hindered them. As it was, only Apple and the Macintosh barely squeaked out the share it needed to survive up to the infamous Microsoft cash infusion, and then to go onto a corporate pivot to end all corporate pivots.

Well, the article you cited is largely based on recollection of Dr. Peter Kittle.

He worked for Commodore in Germany in support. Not even management.

In his own autobiography, comp.sys.amiga was even a favorite source for information.

He would have had no direct access to sales info. His own recollections indicate he didn't seem to know layoffs were coming.

[BillLoguidice]

Well, the article you cited is largely based on recollection of Dr. Peter Kittle.

He worked for Commodore in Germany in support. Not even management.

In his own autobiography, comp.sys.amiga was even a favorite source for information.

He would have had no direct access to sales info. His own recollections indicate he didn't seem to know layoffs were coming.

 

I didn't cite the article. It was someone else. But yes, I agree, that's not the best source.

[JamesD]

My info probably isn't anymore reliable.
You just have to remember that the US was way ahead of the rest of the world in what they were willing to spend.
When 8 bits were taking off in Europe, the US was already moving to 16 bits.
The VESA local buss and DOOM were pretty much the end for Amiga.
PCs were suddenly faster at graphics and could do something the Amiga couldn't... at least at the time.

[zzip]

 

I suggested that in the past too. Basically follow the Tandy model (which itself was modeled after the IBM PCjr AV system). Again, though, the problem with that is they would have to have started that more or less as soon as the Amiga came out, kind of undermining the purpose of the Amiga. The other issue, of course, is that that still would arguably only take you so far before a universal standard would develop, a la VGA. We can again look to how the Tandy standard eventually had to be deprecated, and that was one of the best selling clone systems for several years.

 

Outside of 20/20 hindsight, I think the most realistic path for Commodore would still have been to market the Amiga better, embrace its cross-compatibility more, and establish a higher percentage of marketshare. I still then see a surviving Commodore having to become a PC clone maker, with perhaps their name being held in more regard after managing the Amiga better. They could have leveraged that background/reputation when multimedia PCs started to become a thing.

 

Agreed, that's why I think Commodore/Amiga would have had a better shot at making sound cards than video. The Video standards were dictated by IBM and later VESA. While audio was open and Soundblaster became the defacto standard because they sold the most! At least that was the case until AC97 came along. Amiga was years ahead of Creative Labs.

[JamesD]

The DMA driven audio of the Amiga was largely possible due to the machine's memory architecture.
You had a large block of RAM dedicated to graphics and sound, and a non-segmented architecture.
I'm trying to figure out how that would work with the PC's memory architecture.
You had buss speed limitations, memory limitations, etc... with ISA.
I'm not sure if you could have something like that before the PCI buss was introduced. That was introduced in 1992.

[carlsson]

As reported by COMPUTE! in December 1986, the US home computer market consisted of 31% Commodore, 16% Apple, 14.7% Atari, 12% IBM, 9.5% Tandy, 1% Leading Edge, 15.8% other brands and 1% rounding error. I found another graph that suggests that out of those 31%, about 29% were C64 sales and then 1% each for C128 and Amiga 1000 at the end of 1986. By the end of 1987, the Amiga may have had up to 4% of the US market, at the same time nearly 2/3 of the sales came from IBM PC and compatibles.

 

Selling both the 68K based Amiga and ISA expansion cards with the chipset but no software that uses it probably had been pointless. Making the Amiga fully MS-DOS compliant with e.g. a 80286 and emulating CGA may have worked better, but then again besides the Tandy 1000 there were other brands like the earlier Sirius 1 / Victor 9000 (up to 800x400 resolution, 869K RAM, 600/1200K floppy disks etc) that only sold well in Europe due to the IBM PC got delayed.

 

Keatah previously mentioned that the limited audiovisual capacities of the PC prior to EGA & AdLib and even more so VGA & SoundBlaster, meant developers of professional software could put more efforts in making the programs usable than pretty. In case that is a common conclusion, I doubt the Amiga had fared any better than it did, unless Commodore had managed to keep the same price point for a mostly PC compatible as they did with the real Amiga, so you could've had a computer that both was dull at day (business) and fun at night (games) and somehow had convinced software houses to develop for their system, possibly in the same manner as games would be playable both on IBM and Tandy but with better graphics and sound on the latter.

[Keatah]

Yes yes. Now Hear this!! Doom caused the Amiga's doom!!

Edited February 6, 2017 by Keatah

[Nebulon]

Yes yes. Now Hear this!! Doom caused the Amiga's doom!!

 

True dat.

 

Jay Miner actually warned of this back in 1989. His concerns about the industry shift from planar to pixel-driven graphics systems fell on deaf ears with the execs at Commodore.

 

Of course, in the same talk he mentions that C= never listened to what he said anyway.

 

You know, like why listen to the advice of the industry leader in hardware-accelerated graphics at the time...

Edited February 6, 2017 by Nebulon

[Keatah]

Figures. Planar graphics came about because of memory bandwidth limitations. And despite custom hardware, I can't see planar operating as fast or being as easy to program for as pixel graphics.

Edited February 6, 2017 by Keatah

[JamesD]

As reported by COMPUTE! in December 1986, the US home computer market consisted of 31% Commodore, 16% Apple, 14.7% Atari, 12% IBM, 9.5% Tandy, 1% Leading Edge, 15.8% other brands and 1% rounding error. I found another graph that suggests that out of those 31%, about 29% were C64 sales and then 1% each for C128 and Amiga 1000 at the end of 1986. By the end of 1987, the Amiga may have had up to 4% of the US market, at the same time nearly 2/3 of the sales came from IBM PC and compatibles.

...

If memory serves, the 5% number was a year or two after the release of the Amiga 500 in 1987, but not in the 90s.

I think market share was more like 3% in the 90s even though sales topped 1 million units in a year during that period.

 

[BillyHW]

Can you guys explain the difference between planar and pixel graphics?

[D.Daniels]

Can you guys explain the difference between planar and pixel graphics?

 

In computer graphics, planar is the method of representing pixel colours with several bitplanes of RAM

 

now can you explain to me, what the hell that all means in laymen terms

Edited February 7, 2017 by D.Daniels

[JamesD]

Pixel graphics isn't the proper term, since planar graphics are also pixel graphics.
Chunky pixels has been commonly used... not sure that's the best term.

Planer graphics was a way of providing more colors and saving memory.

With using a single BYTE to represent pixels with chunky pixel hardware, you can have:
1 bit per pixel = 8 pixels per byte for 2 colors
2 bits per pixel = 4 pixels per byte for 4 colors
4 bits per pixel = 2 pixels per byte for 16 colors

8 bits per pixel =1 pixel per byte for 256 colors



Using planar graphics, the hardware doesn't take the bits to represent one pixel from a single BYTE. It takes them from separate blocks of RAM. called bitplanes
1 bitplane can use each bit to represent a pixel for each bit in 2 colors
2 bitplanes can use a bit from each plain to represent 4 colors
3 bitplanes .... 8 colors < not possible with chunky method,
4 bitplanes ... 16 colors. Still uses 4 bits per pixel, but each bit comes from a different block of RAM (bitplane)
5 bitplanes ... 32 colors < not possible with chunky method

 

If we compare 16 color modes:

Non planer:
<BYTE 0, bit 0,1, 2, and 3 supplies bits for 1st pixel color, bit 4, 5 ,6, and 7 supplies bits for 2nd pixel color> ,<BYTE 1, bit 0,1, 2, and 3 supplies bits for 3rd pixel color, bit 4, 5 ,6, and 7 supplies bits for 4th pixel color>, <BYTE 2, bit 0,1, 2, and 3 supplies bits for 5th pixel color, bit 4, 5 ,6, and 7 supplies bits for 6th pixel color> ...

Bitplanes
<bitplane 0 bit 0, supplies bit 0 of 1st pixel color >, < bitplane 0 bit 1, supplies bit 0 of 2nd pixel color >....
<bitplane 1 bit 0, supplies bit 1 of 1st pixel color >, < bitplane 1 bit 1, supplies bit 1 of 2nd pixel color >....
<bitplane 2 bit 0, supplies bit 2 of 1st pixel color >, < bitplane 2 bit 1, supplies bit 2 of 2nd pixel color >....
<bitplane 3 bit 0, supplies bit 3 of 1st pixel color >, < bitplane 3 bit 1, supplies bit 3 of 2nd pixel color >....
etc...

[Keatah]

https://en.m.wikipedia.org/wiki/Planar_(computer_graphics)

[D.Daniels]

eh, i think it means rather than using a byte to create a whole cake

 

you create in ram layers of the cake and join them together or highlight the creamy middle bit

Edited February 7, 2017 by D.Daniels

[tschak909]

While yes, planar graphics were chosen for memory and memory bandwidth limitations, Planar graphics did have some advantages, one of the biggest being that fixed viewports for certain planes allowed for limited color backgrounds to easily be made (which could easily be overcome with copper changing the palette per scanline)... The plane viewports can be moved at different speeds, and thus parallax scrolling becomes a piece of cake, much easier to do than in a chunky format. Keep in mind, that the original "ask" was to make cartoon-realistic animation.

 

-Thom

[Keatah]

The point is to store each layer of the cake in a separate memory chip.

And the hardware reads how many ever layers (chips) you want at a time.

 

With pixel matrix addressable graphics, you must read ALL the pixel information for each pixel, this can mean several locations need to be read. With all the delays that add up.

 

Think serial vs parallel.

[BillyHW]

Seems to me like one method is slightly better at some things, and the other method is slightly better at some other things, but in the grand scheme of things there's not really a huge difference either way.

[D.Daniels]

Seems to me like one method is slightly better at some things, and the other method is slightly better at some other things, but in the grand scheme of things there's not really a huge difference either way.

 

less memory used, so planar was better for older computers and systems I think their saying...............may take us as many years as the release of duke nukem forever to comprehend

Edited February 7, 2017 by D.Daniels

[TMR]

less memory used, so planar was better for older computers and systems I think their saying

For 16-bit systems it's more efficient; 6 bitplanes for a 320x200 display is a little shy of 48K so a serious ask for 8-bits (where the CPU can for the most part only see 64K at a time) and a larger overhead was required for anything using a "chunky" option. It possibly looked more "future proof" back in the day as well, when the Amiga went from ECS to AGA they simply gave the hardware a couple of extra bitplanes to up the 64 EHB colour display to 256. In theory at least, that scales up forever so programmers on theoretical future Amigas could have stored 128, 512 or 2,048 colour images efficiently.

 

The down side is how much data juggling is required to update a bitplane-based display; with a 256 colour image on an AGA Amiga it's necessary to write to all eight bitplanes (one bit in the same place on each is used to give eight bits per pixel) to set one pixel to a specific value whilst the equivalent on 256 colour VGA where the display uses one byte per pixel is a single write. With the same processing power attached to both video systems the planar version comes out significantly slower and, whilst that can be worked around in 2D using the hardware sprites and scrolling, when the focus shifted to 3D and games like Wolfenstein 3D the Amiga was already in trouble before the difference in CPU speed was considered.

[Keatah]

Speed of memory was the determining factor whether planar or pixel/matrix method was used.

[Keatah]

Speed of memory access determined whether planar or pixel/matrix method was used.

 

[zzip]

For 16-bit systems it's more efficient; 6 bitplanes for a 320x200 display is a little shy of 48K so a serious ask for 8-bits (where the CPU can for the most part only see 64K at a time) and a larger overhead was required for anything using a "chunky" option. It possibly looked more "future proof" back in the day as well, when the Amiga went from ECS to AGA they simply gave the hardware a couple of extra bitplanes to up the 64 EHB colour display to 256. In theory at least, that scales up forever so programmers on theoretical future Amigas could have stored 128, 512 or 2,048 colour images efficiently.

 

The down side is how much data juggling is required to update a bitplane-based display; with a 256 colour image on an AGA Amiga it's necessary to write to all eight bitplanes (one bit in the same place on each is used to give eight bits per pixel) to set one pixel to a specific value whilst the equivalent on 256 colour VGA where the display uses one byte per pixel is a single write. With the same processing power attached to both video systems the planar version comes out significantly slower and, whilst that can be worked around in 2D using the hardware sprites and scrolling, when the focus shifted to 3D and games like Wolfenstein 3D the Amiga was already in trouble before the difference in CPU speed was considered.

 

I can see how it benefited the Amiga with 5 bitplanes. I still don't understand how it benefits the ST, which had 1, 2 or 4 bits per pixel, all of which could be packed in a single byte for the same memory usage as planar.

[JamesD]

Speed of memory was the determining factor whether planar or pixel/matrix method was used.

A 16 color mode using chunky graphics requires 4 memory accesses for 8 pixels.

The planar version requires 4 memory accesses for 8 pixels.

*edit*

Unless you are talking about drawing lines, setting pixels, chunky to planar...

Edited February 7, 2017 by JamesD

[JamesD]

For 16-bit systems it's more efficient; 6 bitplanes for a 320x200 display is a little shy of 48K so a serious ask for 8-bits (where the CPU can for the most part only see 64K at a time) and a larger overhead was required for anything using a "chunky" option. It possibly looked more "future proof" back in the day as well, when the Amiga went from ECS to AGA they simply gave the hardware a couple of extra bitplanes to up the 64 EHB colour display to 256. In theory at least, that scales up forever so programmers on theoretical future Amigas could have stored 128, 512 or 2,048 colour images efficiently.

 

The down side is how much data juggling is required to update a bitplane-based display; with a 256 colour image on an AGA Amiga it's necessary to write to all eight bitplanes (one bit in the same place on each is used to give eight bits per pixel) to set one pixel to a specific value whilst the equivalent on 256 colour VGA where the display uses one byte per pixel is a single write. With the same processing power attached to both video systems the planar version comes out significantly slower and, whilst that can be worked around in 2D using the hardware sprites and scrolling, when the focus shifted to 3D and games like Wolfenstein 3D the Amiga was already in trouble before the difference in CPU speed was considered.

That's why a hardware blitter is important. Sadly, the Amiga blitter was pretty limited.

A generic set or get pixel would have solved the problem on all models.

*edit*

Well, that doesn't completely solve chunky to planar.

Edited February 7, 2017 by JamesD