Amiga 68k cpu

[roondar]

Originally Posted by Promilus

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

Which is why Commodore released the A1200 with 2MB at that time and not started out the Amiga with that much. Kinda my point really

[Daedalus]

Even when the CD32 came out a bit later, 2MB of RAM was still the single most expensive line on the BOM. RAM was getting cheaper, true, but it wasn't cheap.

[stevelord]

Originally Posted by Promilus

@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

I think you might be thinking I'm thinking of something different to what I'm thinking. I'm not talking about switching to a non-68000 CPU. I'm talking about switching from a 7mhz to a 14mhz crystal. By 1992 there were already 68000 accelerators like the Adspeed that did this. It'd require switching from the MC68000FN8 to the MC68000FN16 to run in spec but from everything I can tell it's the same chip just passing tests at higher frequencies. And there absolutely are fast RAM expansions for the A600, so it would've definitely been possible if there was an exploitable benefit.

Originally Posted by Promilus

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.

According to AmigaLove it might be more the other way around.

Originally Posted by Promilus

And - at that time - hardly any OCS game did require faster CPU in the first place.

Off the top of my head, 1992 releases that I know improve with a faster CPU:

• 4D Sports Driving
• Wing Commander
• Links
• Darkseed
• Robocop 3

Also Frontier, which while technically being a 93 release I erm... acquired in 92.

[Daedalus]

Originally Posted by stevelord

I think you might be thinking I'm thinking of something different to what I'm thinking. I'm not talking about switching to a non-68000 CPU. I'm talking about switching from a 7mhz to a 14mhz crystal. By 1992 there were already 68000 accelerators like the Adspeed that did this. It'd require switching from the MC68000FN8 to the MC68000FN16 to run in spec but from everything I can tell it's the same chip just passing tests at higher frequencies.

Running a 68000 at higher frequencies gives you next to no advantage due to the lack of cache. So not only do you need to add a faster and therefore more expensive CPU (yes, it may be the same chip, but chip companies charge more for the same chip that passes tests at a higher speed), but also cache RAM and the control logic to deal with it. And, of course, you'll need to turn the cache off for most games because they'll crap themselves otherwise. So when playing games (that's most of the time on a stock A600), all that's changed is that the more expensive CPU just twiddles its thumbs for twice as long as the 7MHz version.

[chb]

Originally Posted by Daedalus

Running a 68000 at higher frequencies gives you next to no advantage due to the lack of cache. So not only do you need to add a faster and therefore more expensive CPU (yes, it may be the same chip, but chip companies charge more for the same chip that passes tests at a higher speed), but also cache RAM and the control logic to deal with it. And, of course, you'll need to turn the cache off for most games because they'll crap themselves otherwise. So when playing games (that's most of the time on a stock A600), all that's changed is that the more expensive CPU just twiddles its thumbs for twice as long as the 7MHz version.

The Adspeed actually included 32k SRAM cache for that very reason - but that did not come for free, obviously. And, another crucial point: AFAIK it did not cache chip ram, as it could not tell whether the custom chips changed ram content. Which kind of makes it pointless for the A600. IMHO the Adspeed was only really useful for systems with slow mem/fake fast (or memory on the Zorro bus?), as real fast mem local to the CPU could be accessed at full speed even at 14 MHz.

BTW, if there's no DMA ongoing, a 14 MHz 68000 could access the ram at full speed (the 68000 uses only every second bus cycle at 7 MHz), but that if does some heavy lifting here for a chip ram based system like the A600... Still, it should give some speed increase during hblank/vblank and for mul/div/shift instructions. And I guess by 1992 the 16 MHz 68000 wasn't much more expensive than the 8 Mhz one.

[roondar]

Originally Posted by chb

BTW, if there's no DMA ongoing, a 14 MHz 68000 could access the ram at full speed (the 68000 uses only every second bus cycle at 7 MHz), but that if does some heavy lifting here for a chip ram based system like the A600... Still, it should give some speed increase during hblank/vblank and for mul/div/shift instructions. And I guess by 1992 the 16 MHz 68000 wasn't much more expensive than the 8 Mhz one.

The Amiga chipset (sadly) doesn't let the CPU access the bus 2 cycles in a row. Even AGA can't do that
(note for accuracy: the CPU can access any Chip RAM bus cycle, but it can never access two bus cycles in a row)

[Tigerskunk]

Hindsight - the thread.

ARM wasn't even known to most people until like ten to fifteen years ago when it made some leeway in the mobile space.
And that was mostly due to its thermal and power saving capabilities.

Plus, RISC asm dialects were horrible to code in, while M68K is basically like coding in C.
And back then, coding in ASM was still important for a lot of tasks.

[chb]

Originally Posted by roondar

The Amiga chipset (sadly) doesn't let the CPU access the bus 2 cycles in a row. Even AGA can't do that
(note for accuracy: the CPU can access any Chip RAM bus cycle, but it can never access two bus cycles in a row)

Yes, you're of course right - but the RAM would be fast enough. Not a hardware guy, but I guess letting the CPU access every free bus cycle does seem to be a rather small modification, as blitter and bitplane already do that.

[ImmortalA1000]

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?

Lorraine/Amiga design was finished before ARM1 was in final prototype so no. The two options considered were Z8000 and 68000 and I suppose an 8086 or some other very exotic and expensive CPU of 1984 could have been used. They made the right choice, pretty sure the Z8000 wasn't as good value/performance of 68000.

Besides, the Amiga 1000 took it's strength from the custom chips running in parallel to the CPU via a shared memory bus anyway. In 1984 the 68000 was the best 16bit CPU for the price. The weirdo 6502 derived CPU of the SNES/Apple II GS was possibly an option but it was no match in power or value for money and I don't know when WDC invented it. The fact that Pilot Wings, a launch SNES title, had to have a DSP inside it due to the weak CPU is probably another win for "yes 68000 was the right choice for Amiga design"

What should have happened is the A500 been 14mhz and the A2000 20mhz 68000 or 14mhz 020 as standard etc, you can blame scumbag Irving 'I wish his parents wore a condom that day" Gould who is famously quoted as saying "7mhz 68000 is enough" in 1990 (New Computer Express weekly or possibly Popular Computing Weekly. I used to work in a Newsagents part time so I could read all the magazines!!).

I guess his annual million dollar bonus cheques he personally wrote himself from company profits had to be funded somehow

[roondar]

Unless you add Fast RAM, an Adspeed like cache or change the chipset, running the 68000 in the A500 faster won't make much of a difference (barring notable exceptions where lots of multiply/divide instructions are run, such as in 3D games).

As is, the Chip RAM speed is fully aligned with the 7Mhz CPU's access speed to that RAM. Making the CPU faster will not make the Chipset faster, which is what determines the CPU's maximum RAM access speed. In other words, no fast RAM is no speed advantage for a faster 68000 in most cases.

[Gorf]

> And - at that time - hardly any OCS game did require faster CPU in the first place.

what was first: the chicken or the egg?

releasing a computer with literally still the same specs 5 years later, was a recipe for disaster...

[Promilus]

A600 wasn't built to be expandable other than pcmcia, chipram and ide. That also means that async CPU access A3k style was too expensive and even should they opt for faster CPU and some 1MB of fast ram to keep upgraded CPU actually useful it will soon be less attractive. And while "SuperA600" would've been real and decent upgrade over stock a500 that'd also mean that in the same time there would've been 3 platforms to write games - 7MHz 68000 OCS with 512K CHIP and 512K slow at most (A500). 7MHz 68000 ECS with 1 or 2MB CHIP (A500+) and XXMHz 680x0 ECS with 1-2MB of CHIP, 1-2MB of FAST ("SuperA600"). Yeah, that makes perfect sense for game developers. And behind a corner A1200 with 2MB of CHIP and 0MB of fast ...

[sandruzzo]

A500 deserved even a little as 32k/64k of fastram and better bus for trapdoor ram..

[TCD]

Originally Posted by Gorf

releasing a computer with literally still the same specs 5 years later, was a recipe for disaster...

I've found an article (via Irving Gould's wiki page) that might shed some light as to why CBM did very little to keep up with other computers in the last years: https://www.commodore.ca/commodore-h...ing-commodore/

[DisasterIncarna]

Originally Posted by TCD

I've found an article (via Irving Gould's wiki page) that might shed some light as to why CBM did very little to keep up with other computers in the last years: https://www.commodore.ca/commodore-h...ing-commodore/

i always thought the issue regarding leaving commodore/amiga stagnant was part Irving but mostly Mehdi Ali, both wanting to just rebadge/resell the same old stuff because they wanted to earn big and spend little to nothing on R&D/etc.

[TCD]

Originally Posted by DisasterIncarna

i always thought the issue regarding leaving commodore/amiga stagnant was part Irving but mostly Mehdi Ali, both wanting to just rebadge/resell the same old stuff because they wanted to earn big and spend little to nothing on R&D/etc.

Pretty much that. Just found it interesting that Gould had no interest in or knowledge about the product he sold and could fire Ali if he didn't agree. Not saying that Ali didn't agree with how it went down, but no magical wonder hardware would have saved anything given the people in charge of CBM.

[ImmortalA1000]

Originally Posted by roondar

Unless you add Fast RAM, an Adspeed like cache or change the chipset, running the 68000 in the A500 faster won't make much of a difference (barring notable exceptions where lots of multiply/divide instructions are run, such as in 3D games).

As is, the Chip RAM speed is fully aligned with the 7Mhz CPU's access speed to that RAM. Making the CPU faster will not make the Chipset faster, which is what determines the CPU's maximum RAM access speed. In other words, no fast RAM is no speed advantage for a faster 68000 in most cases.

Well I should have added specifically that it would be designed around a 14mhz asynchronous CPU bus+7mhz Chip RAM done properly not just slap in a 14mhz CPU type affair but you should assume I meant that. Commodore had Jay Miner with them until 1989/1990? They never used his skills. They also never tried to keep the genius of RJ Mical/Dave Needle, more fool them.

Almost all of the cost cutting of A1000 to A500 is purely down to replacing a complex daughter board full of 192k (?) protected Kickstart RAM with a single ROM chip on a single motherboard and the falling price of 512k worth of DRAM chips required for A500 in 1987 vs price of 512k+192k of DRAM required for 512k A1000 in 1985. I see no world class cutting edge engineering talent at Commodore during 1985-1994, so maybe it would be impossible anyway

[ImmortalA1000]

Originally Posted by TCD

Pretty much that. Just found it interesting that Gould had no interest in or knowledge about the product he sold and could fire Ali if he didn't agree. Not saying that Ali didn't agree with how it went down, but no magical wonder hardware would have saved anything given the people in charge of CBM.

Gould was cutting himself million dollar bonus cheques out of C= profits annually, he knew enough to know better spec=smaller bonus cheques. He only bought Amiga because Jack wanted it, childish little twat that he was.

Medhi Ali may have been paid too much and a clueless idiot with the personality of turd that wont flush but Irving Gould is the one who refused to sacrifice his bonuses and ultimately forced out the genius of Jack.

EVERY reason C= failed is ultimately down to Irving Gould getting his claws into C= during the disastrous and dirty Texas Instr' calculator war. So either you blame the scumbag at TI who decided to start that calculator war or Irving for what he did with his new toy for a tiny outlay of cash.

[roondar]

Originally Posted by sandruzzo

A500 deserved even a little as 32k/64k of fastram and better bus for trapdoor ram..

Adding a small amount of Fast RAM tot the design is something that on paper sounds like it makes a lot of sense. But despite me not being a hardware guy (so I have no idea how to make that work), it seems to me it's not actually trivial to do this and still allow a Fast RAM/CPU slot on the side of the machine as we have now.

The way I understand it you can't just randomly make an extra bus and those pins have to be connected somehow. So it sounds to me like you'd need potentially expensive bus arbitration.

Of course, I know nothing about hardware, so maybe I'm wrong

Originally Posted by ImmortalA1000

Well I should have added specifically that it would be designed around a 14mhz asynchronous CPU bus+7mhz Chip RAM done properly not just slap in a 14mhz CPU type affair but you should assume I meant that.

Ah, I see. I'm no HW guy, but my guess it that this wouldn't have been so easy to do back in 1983/1984. You got to remember, when the Amiga was first released, it was actually really quite fast as far as consumer hardware went. The 7MHz 68000 was a massive upgrade compared to 8 bit systems and it ran circles around the common 8088 based XT's. It competed favourably with 286 based systems (certainly in terms of GFX abilities) at a much lower cost.

My problem probably is that I just don't see the Amiga as being too slow or poorly designed when you place it in it's time frame

[hooverphonique]

Originally Posted by roondar

As is, the Chip RAM speed is fully aligned with the 7Mhz CPU's access speed to that RAM. Making the CPU faster will not make the Chipset faster, which is what determines the CPU's maximum RAM access speed. In other words, no fast RAM is no speed advantage for a faster 68000 in most cases.

And this "feature" is exactly what Mathesar's 14MHz (68000) A500 accelerator uses to "disable" itself - turn off access to its onboard fastram.