Hacker News Reader

The reason they didn’t do a full 640×480 was because they were focused on TV-compatible video hardware, not because an 8MHz 68000 couldn’t handle it.

Even without Commodore’s terrible corporate structure they would have had a rough time improving Amiga because they painted themselves into a corner with an expensive chipset that developers had to work with directly for good performance and that didn’t have a lot of room to grow in a compatible fashion—the first real revision to the Amiga was 1990’s A3000, the A500 was just a value-engineered A1000 while the A2000 was just an A1000 with internal slots. No real change for 5 years would have prevented market growth, even if corporate management had been at all competent.

They would have had a lot more leeway to make more interesting systems faster without the crushing weight of compatibility engineering. Apple was incredibly smart on that front with Macintosh—and the early introduction of Mac XL helped ensure developers didn’t ignore the “don’t touch the hardware directly” directive, at least for business software. That made the Mac II possible in 1987, and by the A3000’s release in 1990 Apple had an entire line of full 32-bit workstations that ran virtually all Mac software without weird caveats. And Windows 3.0 was also in the mix at that point on 386 and better PC hardware…

> The reason they didn’t do a full 640×480 was because they were focused on TV-compatible video hardware, not because an 8MHz 68000 couldn’t handle it.

The interleaved memory access of the CPU & Agnus means that the amount of data that can be pulled from the RAM is directly related to the bus speed.

This, combined with the horizontal line frequency and the bus width determines how many memory accesses there can be per line - to get a 640x480 progressive scan you'd have to double the amount of accesses per second and there's not enough access slots to do this at 7 MHz.

When A1200 came out (too little too late by the end of '92) with AGA the CPU speed was doubled to 14MHz with a 32-bit wide bus which finally allowed for higher resolution progressive modes, more colours etc.

Or you dual-port the VRAM. Or you have a division between RAM banks managed by the OS, e.g. “slow” and “fast.” Or you have the custom chips mediate access to the VRAM entirely, but offer a DMA feature for, say, block transfers of bits. Or…

There are lots of ways to support a full 640×480—or more—with an 8MHz 68000. Sun, Apollo, and HP all supported at least 1024×768 on their 68000/68010 workstations.

I agree. What seemed to be a game changer at the time and maybe was for a short period, the blitter and the bitmap concept, showed later on to be more of a burden than an asset.

The A3000 wasn't really much of a revision. True, it had a hack that made it possible for the CPU to access chip memory 32 bits at a time, but the chipset remained using a 16 bit data bus for all accesses. When running in the 640 pixel wide hires graphics mode using 4 bits per pixel, the memory bandwidth was completely saturated. The really sad part is that a motherboard the AGA chipset (which actually increased memory bandwidth for graphics, but not for the blitter and other parts of the chipset) was ready in an A3000 form factor about a year before the A4000 shipped. Heck, the A4000 couldn't ship in quantities to meet demand because Commodore was cash starved and couldn't afford to pay HP to make more chips. What a series of complete and total failures by management.

Yikes! So even the A3000 was hobbled by decisions made in 1983-85!

That limitation was not about frequency, it was about supporting widespread TV hardware. Flickering was inherent to how TVs worked at the time, but Amiga software did not sufficiently account for that. If they had used the interlaced modes correctly, i.e. as a minor tweak over simple scanline doubling rather than a true increase in vertical resolution, the flickering would've been a non-issue.

The Amiga only displays one interlace field per 1/30th of a second -- that is to say that it is effectively 15(ish)fps in interlaced mode. It's only using a traditional 200 line output, just phase shifted for the second field. It's basically a hack.

It flickers because the persistence of most CRT screens is less than 1/15th of a second, and so by the time you see the second field, the first field is already diminishing in brightness. IIRC Commodore released (or wanted to release?) a monitor that had a longer persistence.

> The Amiga only displays one interlace field per 1/60th of a second -- that is to say that it is effectively 30fps in interlaced mode. It's only using a traditional 200 line output, just phase shifted for the second field. It's basically a hack.

That's how TV's worked back in the day, yes. It would have led to flicker when you had lots of sharp changes in image content across scanlines, but not otherwise. E.g. filling the whole screen with a solid color, or even a smooth color gradient, would've worked quite fine in interlace modes.

Unfortunately, this inherent limitation of interlaced display went unrecognized in the Amiga dev community - everyone would've known that interlaced screens tended to flicker a lot but most devs didn't know why, or how to properly work around the issue.

That wasn't my experience at all. Everyone knew what caused flicker and how to work around it. For example, that's why Workbench 2.0 had a much lower contrast black on grey default color scheme, instead of Workbench 1's more garish white on blue. (Supposedly those were picked to look more readable on cheap monitors.)

The problem with running Workbench in interlaced mode is that it just gave you tiny text and icons: the ROM only had a single bitmap font in 8 and 9 pixel height size, and adding custom fonts to your boot disk would've taken up a whole lot of space.

(Similar to running unscaled High-DPI today on systems that weren't designed for it out of the box.)

True, but 640x200 was an annoyingly low vertical resolution.

No, it's one field per 1/60 seconds, giving 30 frames per second.

Could you elaborate on the "giant tax dodge" angle?

https://www.commodore.ca/commodore-history/the-crazy-story-o...

I don't support this version. For me much more looks like, Commodore/Atari tops were very afraid of being accused of having a monopoly on PC market, but they couldn't create some semi-competitor like AMD that years, and did not have good enough control, to make open source design like IBM PC.

So back in the day, Commodore flew me out to Frankfurt Germany to interview me for the role of CTO of Commodore[1]. We had a wide ranging discussion about the role and Commodore's strength as a brand, etc etc. I wanted to dig into the financials though because even then, I had a reasonably good understanding that corporations are organized around their core mission, they invest in that mission to the exclusion of things that are "not core."

As we got to discussing things, the structure of holding companies holding holding companies and subsidiaries in the Netherlands licensing technology out to other subsidiaries, it was obvious that Commodore invested most heavily in creating a corporate structure that was "perfectly legal" and could argue away any tax liability based on any level of income. The CEO at the time, Rattigan, was a finance guy (in my opinion) not a computer guy. He liked the C64/C128 "better" because they had better margins and less of a support burden than the Amiga did. (emphasis mine)

I left unconvinced they were ever going to be serious about building computers, very much in the consumer/toy mindset of maximum margin/maximum volume. I politely let them know I would not be interested in pursuing the discussions further.

[1] Fun fact, the guy that took the job lives about 1/4 mile from me and we're friends :-)

Sun surpassed Commodore in revenue first a couple of years after the Amiga launched, I think? Though with the management chaos at Commodore maybe the Sun market cap had already surpassed them?

At the point Sun hit revenue parity Sun was one of the fastest growing tech companies, and Commodore had never reclaimed the peak revenue it reaches in 1984 (when it exceeded a billion USD), and so I'm not surprised it didn't seem attractive by the time it might have been financially possible.

At the time (early 90's) Commodore was very much struggling. Since I knew their hardware team and had a lot of respect for them, and they were 68K based and seriously considering doing UNIX workstation level like things, my pitch to the corp dev folks was for a "modest" investment, Sun could produce an economical workstation that leaned in on hardware acceleration (important note: Andy Bechtolsheim was a huge "don't bother with custom chips, CPUs will get fast enough" person) and gave Sun a way to create a flanking attack on Windows. Here's an OS that doesn't have the baggage of SunOS, can compete head to head with DOS, and cause Microsoft pain trying to swat it away.

Apparently it wasn't compelling enough of a pitch :-) But then again it was one of those times where I got a lesson in how folks cannot see what they cannot see.

Thanks for the added detail. It makes for another interesting "what if" in the very, very many missed opportunities in the Commodore saga...

They did do a workstation level Amiga - the A4000.

https://en.wikipedia.org/wiki/Amiga_4000

It was just a little too late as the workstation market was drying up and they arrived with a Sun 3/50 competitor in a market that Sun had already started to walk away from. You could buy alternate CPU for the A4000 as it had the CPU installed on a daughter card and I think you could get a PowerPC setup in that chassis.

I am aware :-). It was the direction Amiga (although not Commodore) was headed. The key is that Commodore didn't see it as strategic nor could they justify re-organizing around it once the end of the runway was pretty clear.

> I tried to get Sun to buy them but alas

Commodore had major management rot and corporate issues. Maybe that rot would have reached Sun, which would have been detrimental to the future of Sun at that time.

I think that without Sun, the world of computing today might have looked quite different.

Commodore was also surprisingly big, and Sun grew surprisingly fast, and I don't think that was a good match, as much as it'd have been intriguing.

Commodore reached a billion USD revenue in 1984. Throughout most of the rest of the 1980s, its revenue was in the $800+ million range.

March 30th 1986, SF Examiner reported Sun's revenue as $115.2m.

By 1988, however, Sun passed $1bn in revenue, according to The Press Democrat, who also reported their 1987 revenue as $537.5m.

At some point in that interval it'd probably have been possible for Sun to get a good price for Commodore, but I think at that point Commodore would've already seemed like a bad deal to them given their growth trajectory. A couple of years later Commodore was in terminal decline, and Sun had kept growing and while Commodore at that point might have been small fry enough to be more viable to pick up without changing Sun too much, I'm guessing they didn't seem very relevant to Sun at that point.