Comment by mrandish | Hacker News Reader

Comment by mrandish | Hacker News Reader

The Amiga 1200 was arguably the last and best of the golden era "console computers", (meaning all-in-ones with the CPU inside the keyboard and could display on a standard analog TV). This style of home computer largely reigned the consumer market between 1980 and 1990, beginning with 8-bit machines like the Commodore 64, Atari 400/800 and Apple II, evolving to 16-bit with the Amiga 500 and Atari 1040STE and ending with the 32-bit Amiga 1200.

The Amiga 1200 was based on a Motorola 68020, had a hardware-based bit blitter and could display 320x240 in 256 colors, basically the holy grail of analog gaming graphics modes. It was the only widely distributed home computer capable of faithfully emulating the vast majority of 1980-90 dedicated arcade machines with few compromises. I have all the machines mentioned above in pristine working condition and still enjoy playing around with them.

It was a special era in computer history because every platform had its own unique hardware and operating system providing different capabilities and trade-offs. This gives each platform an unmistakable (and strongly-opinionated) 'personality'.

The 68020 is probably the last great CPU to write pure assembly in, too. The instruction set is rich enough to make it straightforward to write by hand - unlike RISC-ier chips, C is almost superfluous for many tasks - but simple and clean enough to keep in your head (unlike x86). It cleans up some of the rough edges from the 68000, and adds some great features and a ton of speed. But: no MMU, no complex pipeline or branch prediction, and the cache is instruction-only, meaning you rarely need to think about it to get optimum performance.

I disagree that RISC chips are especially hard to write for by hand - plenty of people did exactly that for the Acorn Archimedes. (And writing by hand can still be worthwhile even when you're lacking fancy CISC instructions, because you can design bespoke, custom-use "calling conventions" that make way better use of the register file. This is hard to do in C even when theoretically possible-- via fancy, non-standard asm intrinsics-- because you don't know beforehand how the compiler will choose to allocate registers.) The MC68k architecture proved a dead end, and this is ultimately what would have doomed the Amiga installed base, even if Commodore hadn't been mismanaged to the extent it was.

The ARM chips were brilliant and definitely the most pleasant of the RISC chips to handwrite for, particularly with a nice assembler that handles the quirks of dealing with constants for you.

While the 68k ISA was a dead-end, that's mostly because Motorola was all-in on PowerPC; AFAIK there's no fundamental reason they couldn't have pulled an Intel and thrown transistors at it until it could keep up despite the decode complexity.

The biggest problem I have with RISC assembly is that only being able to access memory through dedicated load/store instructions tends to make the code more complex with additional register shuffling.

FYI, the 68060 was fast, providing a good upgrade path from the 68040s found in a4000s.

Near the end, after the initial bankruptcy, PowerPC boards, some with 060s and PPC chips both, hit the scene.

Entire SDKs were introduced, and many bits of software were ported. This was a good bridge, and as Apple made this sort of jump, eg, they didn't die when switching from 68k to ppc, I don't think a healthy Commodore would have either.

So I disagree. The Amiga could have transitioned too.

I think its too much to lump all RISC chips in the same bucket.

I have written some assembler for MIPS and Alpha AXP back in the day. The mips assembler was quite straight forward and reasonable but I found the alpha to be very hard.

Another thing that made this era special is that these were meant to be mass market devices in which it was kind of assumed that most owners would want to learn basic programming (usually in BASIC, no pun intended) to get the most out of their devices. The machines came with manuals including programming tutorials.

> The Amiga 1200 was arguably the last and best of the golden era "console computers", (meaning all-in-ones with the CPU inside the keyboard and could display on a standard analog TV).

Surely if "last and best" is your criterion, one can hardly discount the Atari Falcon. (And the earliest models in the Acorn Archimedes line were definitely "all-in-ones" integrating a keyboard although they might have lacked in-built TV out.)

Surely you're not trying to restart the A1200 vs. Falcon debate? We all know the Amiga was superior anyway. ;)

>the Atari Falcon

Was just an Atari. Braindead architecture throughout.

It is remarkable how the 68030 still had to wait for its turn to access RAM. Like the ST, Falcon had no concept of FAST RAM.

Falcon had a cool DSP chip in it, but that's about it.

The Acorn A3010 was the only "home" computer, targeted at the non-educational market, that Acorn ever made. It is distinguished by being the only Acorn computer with green (not red or grey) function keys, and also by being the only Acorn computer to which Econet could not be fitted. And it had a TV-out.

At the time it was launched, 1992, the ARM2 processor inside it was considerably more powerful than workstations that could be bought for ten times the price. It was insanely powerful, but hamstrung by not having any form of graphics acceleration. (The CPU was so fast it was just assumed you'd do everything in the CPU.)

It's a shame, really. RISC OS running on ARM deserved to be the British standard of computing, but Acorn's failure in the US market meant they could not compete with the sheer PR power of the large American competition.

Photos and details: http://chrisacorns.computinghistory.org.uk/Computers/A3010.h...

As someone who back then actually used those machines together pc and macs of the same era, thosw performance claims was nonsense.

Maybe it was the OS that caused it, but working on acron was painfully slow.

(Developed a CAD software for it, moved to PC very quickly)

There must be some rose-tinted glasses in the retrospective pipeline here. Even assuming a generous IPC advantage, a 12Mhz ARM wasn’t likely to be competitive with a contemporaneous SPARCstation @ 40Mhz or an Intel 486DX2 @ 50Mhz.

>meaning all-in-ones with the CPU inside the keyboard

I keep looking at my mechnical keyboard and thinking it ought to be possible to squeeze a NUC like computer in there.

Reply on news.ycombinator.com