Amiga 68k cpu

[sandruzzo]

Was the 68000 the right choice for Amiga 500? Back in the day we had Arm processors that were far superior and I think even cheaper. Why didn't they use it?

[saimon69]

Were ARM processors around in 1986 when the 500 was designed? Thought the Archimedes came in 1987

[sandruzzo]

1985 arm1

[sandruzzo]

The ARM1 (Acorn RISC Machine 1) is Acorn Computers' first microprocessor design. The ARM1 was the initial result of the Advanced Research and Development division Acorn Computers formed in order to advance the development of their own RISC processor. The ARM instruction set design started in 1983. A reference model was written in BBC BASIC by Sophie Wilson and Steve Furber in just 808 lines of code. On April 26 1985, after 6 man-years of design effort, the first ARM processor prototype was delivered. The first batch of prototypes were functional and were shipped to customers in the form of evaluation systems. At that time the ARM1 was the simplest RISC processor produced.

The first prototype tested worked on the first try, this was despite the ammeter reading no power. The prototype test board designed was faulty with a short. The chip was entirely running off the leakage from the I/Os. Designed to run at 1 W, the chip averaged under 100 mW typical power.


Originally intended to perform at roughly 1.5 times performance of the VAX 11/780, the prototypes ended up achieveing between 2x to 4x the performance of the DEC VAX 11/780; this is roughly equivalent to 10 times that of that original IBM PC AT or that of the Motorola 68020 operating at 16.67 MHz.


https://en.wikichip.org/wiki/acorn/m...itectures/arm1

[alexh]

You couldn't buy ARM processors when the Amiga was designed which was 1984 if not earlier

[sandruzzo]

@alexh That's true. Think about Amiga 2000-2500-3000-4000 they could have been Arm based with 68k as retrocompatibility. They could have been a lot cheaper and a lot more powerfull

[Promilus]

68k was neat design with plenty of registers and nice ISA which made developing in asm good choice. ARM asm wasn't nearly as efficient. It took more instructions to get the same job done. And - as said - by the time A1000 was developed there was no ARM available yet.

Think about Amiga 2000-2500-3000-4000 they could have been Arm based with 68k as retrocompatibility

Double processors wouldn't be cheaper and emulating 68k on that ancient ARM is a joke. Even modern ARM processors reaching over 1GHz can hardly emulate 040 level of speed (JIT is much different approach and not exactly viable solution back then). Look in how many computer machines those ARM were used after Archimedes... Not many, right? Whole market was too stupid to realize how great architecture it was?

[chb]

Originally Posted by sandruzzo

Was the 68000 the right choice for Amiga 500? Back in the day we had Arm processors that were far superior and I think even cheaper. Why didn't they use it?

There's actually neither any indication that ARM processors were available to third parties in 1987 (and they weren't available for *anyone* in 1985), nor that they were cheaper (68000s were very cheap at that time, AFAIR in the single digit dollar range).

Furthermore, a lot of the ecosystem (compilers, tools, operating systems, libraries, applications) that was there for the 68000 was missing in the ARM's case. So even if Amiga/Commodore had had the possibility to use the ARM (which they most likely hadn't), the 68000 would still have been the safer (and saner) choice. Case in point: ARM never really took off on the desktop.

[sandruzzo]

@Promilus. That's true, I love 68k because it's very easy to programm, but sadly that architecture is gone, Arm and x86 are still here..

[stevelord]

68000 was absolutely the right choice for the A1000. At the time it was being designed the standard for home computers were still Z80 and 6502. Although the 68000 isn't exactly fast by even '91 standards from 1982-1985 (when the A1000 was being designed and built) the 68000 was a popular beast, as seen by it's adoption in Sun and Silicon Graphics' IRIS workstations. By 1984 the 68000 was a cost-effective mid-range 16-bit chip, which is why it was used in the Macintosh and ST too.

A better question would be whether 68000 was the right choice for CDTV, A500 plus, A600 and A1200. For the CDTV, A500+ and original "A300" (that became the A600) I'd say yes. For what became the A600 I'd have liked to have seen a faster 68000 or a 68EC020, with a 68030 in the A1200 (and 2mb fast RAM).

ARM just wasn't a realistic option. What would become chip licensing giant ARM plc came into existence as the result of an Apple/Acorn licensing deal in the late 80s that only made it to products in '92 (the Newton). DEC signed up in '94. Now if (and that's a big if that assumes competency in Commodore's management) Commodore when looking to replace 68k looked at ARM instead of PA-RISC for hombre that would've been interesting. The result would be the same as Commodore died before hombre materialised.

[roondar]

I sometimes think people forget just how much of an upgrade the 68000 was vs the 6502's/z80's used in many other systems. I had a C64 before my Amiga and the CPU performance of the Amiga 500 that replaced it was miles ahead of it.

Just a simple example: I loved to dabble with fractal drawing back then. On my C64 I had written a program in Commodore Basic (plus an assembly extension to draw GFX in bitmap mode in a fast way) to draw a simple Mandelbrot set image. It took 3.5 hours to draw the image, even though it made use of vertical symmetry. Now, this wasn't an optimal piece of code - it used floating point number and ran in BASIC. But even so...

When I got the A500, I rewrote the program in AMOS without optimising any of it. Both versions ran with floating point numbers and the inner loop was basically the same. It now only took about 18 minutes instead (uncompiled), even though it now ran in 16 colours rather than just black & white.

True, part of that did have to do with the slow speed of bitmap drawing on the C64. And perhaps AMOS was simply more efficient than Commodore BASIC. But even when accounting for that, the difference was still huge. Big differences for CPU bound tasks weren't unusual either, many such things ran much faster on the Amiga than on the C64.

Another example would be how the Amiga versions of 3D games (which normally used the CPU for most things) almost always ran at much higher frame rates than their C64 counterparts, even though they tended to use filled polygons rather than wire frame models.

In retrospect it's always easy to point to things that can be improved. But back in 1985, the 68000 was a really, really big upgrade over what most people had access to.

[Promilus]

@stevelord - with hardly any way to add fast ram anything faster than 7MHz 68000 would be largely wasted on A600. It would have to follow A3k design which was rather expensive and not fit for something which was largely designed as a cost-reduced A500, right?


@roondar - C64 CBM Basic 2.0 interpreter is known to be rather slow in the first place. And by the time Amiga arrived there already was 65C816 which is essentially 16bit 6502. That one was a basis to Apple IIGs and Commodore 65 (which was never released). With all those features C65 had it would most likely make A500 eat dust (despite A500 still having performance advantage). Modified 65C816 powers SNES as well. While not perfectly suited for Amiga it was some worth noting option beside 68000. While we're at it there's also SuperH architecture but that one came out much too late to be used in Amiga.

[skan]

Originally Posted by Promilus

@stevelord - with hardly any way to add fast ram anything faster than 7MHz 68000 would be largely wasted on A600. It would have to follow A3k design which was rather expensive and not fit for something which was largely designed as a cost-reduced A500, right?


@roondar - C64 CBM Basic 2.0 interpreter is known to be rather slow in the first place. And by the time Amiga arrived there already was 65C816 which is essentially 16bit 6502. That one was a basis to Apple IIGs and Commodore 65 (which was never released). With all those features C65 had it would most likely make A500 eat dust (despite A500 still having performance advantage). Modified 65C816 powers SNES as well. While not perfectly suited for Amiga it was some worth noting option beside 68000. While we're at it there's also SuperH architecture but that one came out much too late to be used in Amiga.

Except the c65 was designed like 3 years after the 500 and always been an 8bit machine anyway (commodore would have never had used a 65c816, go check some interviews of Mensch to get the picture)

[roondar]

Originally Posted by Promilus

@roondar - C64 CBM Basic 2.0 interpreter is known to be rather slow in the first place. And by the time Amiga arrived there already was 65C816 which is essentially 16bit 6502. That one was a basis to Apple IIGs and Commodore 65 (which was never released).

Well, I do acknowledge that the relative speeds of the interpreters might play a role in part of the difference. But I also point to another example where the 68000 was much faster

Anyway, the 65C816 is a cool little CPU, sure. Kind of happy they didn't put it in the Amiga though. There is (sadly) no viable upgrade path from it, which the 68000 family did offer for the Amiga - allowing vastly more powerful CPU's to be added later on.

With all those features C65 had it would most likely make A500 eat dust (despite A500 still having performance advantage). Modified 65C816 powers SNES as well. While not perfectly suited for Amiga it was some worth noting option beside 68000. While we're at it there's also SuperH architecture but that one came out much too late to be used in Amiga.

If you read up on the actual features (and specifically how fast/slow they were) of the C65* in the hardware documentation, you might be surprised at how little some aspects were improved over the C64. In particular, the HW Sprites were essentially exactly the same as on the C64 - including the colour/resolution limits. On top of that the C65's DMA chip couldn't do the equivalent of Blitter Objects (it doesn't appear to offer shifting), so objects would have to be drawn using the CPU.

Edit: I went and checked the C65 documentation because I was starting to doubt myself. So yeah, the idea was indeed to allow the DMAgic chip in the C65 to allow for Blitter Objects, though as far as I can see now the cookie-cut operation used for these on the Amiga would require 2 passes on the C65 hardware because the HW only supports a single source/destination.

Would be interesting to see how fast such an approach would really be. The HW docs of the C65 don't really go into detail about speed much.

All that said, I've always wanted a C65 as I was a massive C64 fan back in the day

*) Not the MEGA65 remake, which fixes many of the issues of the VIC-III

[mousehouse]

Originally Posted by stevelord

A better question would be whether 68000 was the right choice for CDTV, A500 plus, A600 and A1200. For the CDTV, A500+ and original "A300" (that became the A600) I'd say yes. For what became the A600 I'd have liked to have seen a faster 68000 or a 68EC020, with a 68030 in the A1200 (and 2mb fast RAM).

I remember the continuous disappointment I felt with the stock CPUs in the Amigas released after the 3000.

[Daedalus]

The low-end Amigas (A500, 600 and 1200) were built to be as cheap as possible, though they were still expensive for many. Adding faster CPUs would only have increased the final sale price; adding fast RAM even moreso. People seem to forget that RAM was expensive back then.

[dreadnought]

Originally Posted by Daedalus

People seem to forget that RAM was expensive back then.

That's the key problem with a lot of these modern takes (not only in the Amiga scene, I've seen variations of them in most of the retro hardware ones). With decades of hindsight it's easy to go into smh mode and question many decisions of the original designers/companies, but given the realities of the time they mostly tried to do their best and often succeeded.

[roondar]

Yup, modern takes kind of forget the cost of components and the compromises that need to be made. For instance, since the 68020 was launched in 1984, technically the Amiga could've been a 32 bit system with 32 bit custom hardware based on the 68020 as well (and I've seen some argue for this). And they could've had it start with say 2MB of RAM. But doing so would've made the system insanely expensive to produce and hence sell.

Which isn't very useful when you're trying to design a mass market product.

All computer systems are ultimately compromises after all

[stevelord]

Originally Posted by Promilus

@stevelord - with hardly any way to add fast ram anything faster than 7MHz 68000 would be largely wasted on A600. It would have to follow A3k design which was rather expensive and not fit for something which was largely designed as a cost-reduced A500, right?

I was specifically thinking about work like Mathezar's 14mhz accelerator. Admittedly his 2.4x speed comes from zero waitstate RAM which isn't what I'd expect without. Using a 14mhz 68k CPU and making it available to the 1mb expansion slot would've given the A600 at least some reason to exist. The reason the A600 was so botched was because the A300 was meant to be a cost-reduced A500 and instead of making it better than an A500 they made it worse (for the time) and it sold at higher prices than the 500+.

Originally Posted by Daedalus

The low-end Amigas (A500, 600 and 1200) were built to be as cheap as possible, though they were still expensive for many. Adding faster CPUs would only have increased the final sale price; adding fast RAM even moreso. People seem to forget that RAM was expensive back then.

You're that RAM was expensive. I don't think an ECS system needed 2Mb as nice as that baseline might've been. The PCMCIA slot and an IDE interface weren't exactly in-line with a cost-reduced approach though. Most choices made on the A600 were downright boneheaded considering it ran against the CDTV and cheaper A500+. There was plenty else wrong with the A600 before we get to 68k though.

[Promilus]

@dreadnaught - maybe yes, but between A1000 and early A500 Agnus which can have ONLY 512KB of chip ram and ECS super fat agnus that's just 1 address line more and demux on board. Nobody will convince me it just had to be cut down like that (512K/1M/2M versions of chips).

@roondar - by the time A1200 was released RAM price dropped significantly.

@stevelord - I know all about A300 and how it became A600, switching to different CPU and memory design would increase the cost and introduce yet another delays and by the time that new and revised A600 would hit the market it would be something closer to A1200 with price but still closer to A500 with performance. And that one was fairly easily turbocharged with either CPU slot or edge connector upgrades. And - at that time - hardly any OCS game did require faster CPU in the first place.