Should the A1200 have used a 28Mhz 68000 processor instead?

[Samurai_Crow]

Originally Posted by d4rk3lf

I am a complete hardware noob...

Would it be possible for Commodore at that time to develop custom chip that would help 3D stuff run much faster on (slower) Motorola processors? Something better then Akkiko. Maybe something like PS 1 had.

Amiga always had that multitask multiprocessor advantage, and some custom stuff that could push 3D, could be at that time something revolutionary.

Not really, other than the planned AAA chunky modes, Fast RAM and an '030 or better.

[roondar]

Originally Posted by d4rk3lf

I am a complete hardware noob...

Would it be possible for Commodore at that time to develop custom chip that would help 3D stuff run much faster on (slower) Motorola processors? Something better then Akkiko. Maybe something like PS 1 had.

Amiga always had that multitask multiprocessor advantage, and some custom stuff that could push 3D, could be at that time something revolutionary.

Ironically, the best way Commodore could have quickly improved Amiga 3D oomph would've been to make sure the CPU had a faster bus to graphics memory.

Had Commodore released a system where the CPU had similar bandwidth to Chip RAM as Lisa had, it would have been far more capable of 3D games like Doom/Doom II, especially with an accelerator added.

Not a very Amiga-like solution, but would've worked a treat in the short term. Chunky GFX would've been great too, but overall less important.

[redblade]

Originally Posted by roondar

Ironically, the best way Commodore could have quickly improved Amiga 3D oomph would've been to make sure the CPU had a faster bus to graphics memory.

Had Commodore released a system where the CPU had similar bandwidth to Chip RAM as Lisa had, it would have been far more capable of 3D games like Doom/Doom II, especially with an accelerator added.

Not a very Amiga-like solution, but would've worked a treat in the short term. Chunky GFX would've been great too, but overall less important.

How do you calculate the Read/Write bandwidth speed to Fast/Chip RAM?? What was the bandwidth speed?

Thanks

[Foebane]

We can speculate all we like, but one thing would've killed the Amiga stone dead even if Commodore HADN'T gone bust:

The Sony PlayStation

Honestly, the Amiga was being hammered by the PC and consoles of the time, and the PlayStation absolutely hammered the PC and consoles of the time. Simples. Put simply, in 1994 and early 1995, there was no PC that could stand up to the technical prowess of the Sony console, so how on earth would the Amiga have been expected to compete?

Face it, it's the march of progress, and these pieces of hardware had their time. All we can do is dream how things could've been better (and boy, they certainly could have been instead of the business-obsessed, money-grabbing, inept-minded and short-sighted Commodore execs that ruled the roost after they kicked out Tramiel), and, short of inventing a time machine and going back to the 80s and making everything right with the Amiga, there's not a lot we can do about it.

In any case, the Amiga would've had a natural end to the series anyway, if Commodore hadn't gone bust. They released the original hardware (OCS), a step-up (AGA) and they would've released their ultra-advanced hardware... and then that would've been it. And I heard this from Dan Wood of Kooky Tech, so ask him.

[Marle]

Originally Posted by Foebane

We can speculate all we like, but one thing would've killed the Amiga stone dead even if Commodore HADN'T gone bust:

The Sony PlayStation

*this* and also the period of 1994-1997 was the time when so many families bought their first Windows PC (not ourselves, we didn't have the money)

None of this casts shade on what a great machine the Amiga is and there were some real hardware pushers and good games that came out in the latter period in late '93 / '94.

For Commodore and the Amiga to have had a fighting chance of keeping a pace the A1200 should have been out by October of 1990 to compete comparably well with the Mega Drive and SNES - AGA may not have compared to the SNES's graphics but with a couple of years on the market, devs would have been more familiar with AGA and working around some of its limitations just as they did with OCS/ECS. After that Dr Ed Hepler's Hombre based CD console in 1994 may have kept a pace with the PSX. And all the above somewhat depends on Commodore management not being total cockwombles.

[Amigajay]

A Commodore console wouldn’t/couldn’t have competed with the PSX, but a Hombre Amiga computer would have not been in competition with it, it always confuses me when people say the PS1 would killed everything at that time, simply not true as obviously PC’s and Mac’s lived through that era, and if the Amiga fanbase and Commodore kept going they would have too.

[zardoz]

Originally Posted by Amigajay

Mac’s lived through that era...

Apple had a pretty tough time of it in the mid 90s too, no idea if they were on the brink. Was it the iMac and the iPod that turned it round for them?

It's probably useful to compare Apple's and Commodore's fates - niche computer makers squashed by the PC juggernaut on one side and Japanese consoles on the other. Is it that Apple cultivated a brand? It's not tech. or engineering that got Apple through but marketing.

An earlier suggestion for the A500 / A600 to go downmarket is the only way I can see C surviving that, if they had carved themselves out a niche market as the cheap / first / bedroom computer.

[Amigajay]

Originally Posted by zardoz

Apple had a pretty tough time of it in the mid 90s too, no idea if they were on the brink. Was it the iMac and the iPod that turned it round for them?

It's probably useful to compare Apple's and Commodore's fates - niche computer makers squashed by the PC juggernaut on one side and Japanese consoles on the other. Is it that Apple cultivated a brand? It's not tech. or engineering that got Apple through but marketing.

An earlier suggestion for the A500 / A600 to go downmarket is the only way I can see C surviving that, if they had carved themselves out a niche market as the cheap / first / bedroom computer.

Ipod, Mac was always a niche market, but its something that kept them afloat, something Commodore could have it managed better.

The suggestion was the exactly the thing Commodore should have made with the A1200, made it a cheap slim A500 instead (ditching CDTV, A500+, A600 and A1200 AGA) and held out for Hombre instead of rushing AGA into the market they already owned and in turn self combusting in on itself thanks to Commodore confusing the very consumers they were trying to sell to.

Mac’s had the DTP market, PCs had the higher end mainstream computer gaming, the Amiga had the lower end computer market, there was no reason why the Amiga couldn’t have continued on this tangent. Even the budget console market was still selling Megadrives and Snes into 1998 even with the so called unstoppable PS1, thats the good thing about markets, there is more than one level.

[roondar]

Originally Posted by redblade

How do you calculate the Read/Write bandwidth speed to Fast/Chip RAM?? What was the bandwidth speed?

Thanks

The TL;DR version is: the chipset gets ~7MB/second (3.54*16/8=~7), the CPU gets ~3.5MB/second, unless the machine is an A3000 or an AGA machine, in which case the CPU gets ~7MB/second.

Lisa is a special case and gets ~28MB/second - which means that AGA machines can access chip memory speed much faster than OCS & ECS but sadly only for the display.

(and my point about bandwidth was that given the choice, I'd rather have seen a non-chunky Amiga with ~28MB/second Chip RAM bandwidth for all devices than a chunky one with ~7MB/second)

The total bandwidth is shared, so any access done by either the chipset or the CPU potentially slows the other down (i.e. the A500 has about 7MB/second of bandwidth in total, of which the CPU can get about 3.5MB/second).

Now follows the long version
---------
I'll explain it for Chip RAM, because the answer for Fast RAM is "it all depends on what kind of Fast RAM". For Chip RAM, however, the results are known

We need to know only a couple of things:

• Clock speed of memory bus
• Access width of memory from the perspective of the chipset
• Access width of memory from the perspective of the CPU

So, let's fill in these numbers

For all Amiga's, the speed of the chip memory bus is the same: about 3.54MHz (you might have heard it's 7MHz, but this is not actually true - a 7MHz memory bus@16 bits would allow for double the memory speed the Amiga actually has)

For OCS Amiga's and all but one of the ECS Amiga's, the access width for both chipset and CPU is 16 bits
For AGA Amiga's and the Amiga 3000, the access width from the perspective of the CPU is 32 bits
For AGA Amiga's, the chipset accesses the bus as 16 bits wide - apart from Lisa, which accesses it as either 16, 32 or 64 bits wide

In all these cases, the CPU is further limited in maximum bandwidth - it generally doesn't get more than 1/2 of the total slots the bus offers.

Given these numbers, we can calculate the total Chip RAM bandwidth.

Do note, however, that the bandwidth as seen by the CPU and as seen by the chipset can be different. The total available bandwidth doesn't change, but the ability to access it does.

A500/A1000/A2000/A600
Total Chip RAM bandwidth = 3.54*16/8 = ~7MB/second
Max CPU bandwidth = 3.54*16/8/2      = ~3.54MB/second*

A3000
Total Chip RAM bandwidth = 3.54*16/8 = ~7MB/second
Max CPU bandwidth = 3.54*32/8/2      = ~7MB/second**

A1200/A4000/CD32
Total Chip RAM bandwidth (all except Lisa)= 3.54*16/8 =  ~7MB/second
Total Chip RAM bandwidth (Lisa only)= 3.54*64/8       = ~28MB/second***
Max CPU bandwidth = 3.54*32/8/2                       =  ~7MB/second**

*) Note that OCS/ECS 16 bit Chip RAM cannot exceed ~3.5MB/sec due to the fetches of data being limited by Agnus to 16 bits a go at 3.54MHz, regardless of CPU speed.

**) 32 bit Chip RAM access is limited to ~7MB/second instead of 14MB/second because 32 bit Chip RAM incurs a wait state (presumably from Agnus/Alice)

***) Lisa can access the Chip RAM in a more complicated way, which allows a quadrupling of standard bandwidth but requires the chip to fetch 64 bits every time

I hope this helps.

[meynaf]

Originally Posted by roondar

*) Note that OCS/ECS 16 bit Chip RAM cannot exceed ~3.5MB/sec due to the fetches of data being limited by Agnus to 16 bits a go at 3.54MHz, regardless of CPU speed.

16 bits a go at 3.54Mhz is 7MB/s, not 3.5MB/s

In fact the situation seems simpler than that.
The bus runs @3.5Mhz.
However half the bus clocks are reserved for video output, regardless if they're used or not.
Therefore at 16 bits we're limited to 3.5/2Mhz *2 bytes transfer speed, i.e. 3.5MB/s.
And at 32 bits it's simply twice that amount.
Of course Alice/Lisa can get all cycles and can do 7MB/s in 16 bits, 14MB/s in 32 bits, 28MB/s in 64 bits (actually double CAS).

[Foebane]

Originally Posted by Amigajay

A Commodore console wouldn’t/couldn’t have competed with the PSX, but a Hombre Amiga computer would have not been in competition with it, it always confuses me when people say the PS1 would killed everything at that time, simply not true as obviously PC’s and Mac’s lived through that era, and if the Amiga fanbase and Commodore kept going they would have too.

The way things turned out, we can only read this and cry.

[roondar]

Originally Posted by meynaf

16 bits a go at 3.54Mhz is 7MB/s, not 3.5MB/s

In fact the situation seems simpler than that.
The bus runs @3.5Mhz.
However half the bus clocks are reserved for video output, regardless if they're used or not.
Therefore at 16 bits we're limited to 3.5/2Mhz *2 bytes transfer speed, i.e. 3.5MB/s.
And at 32 bits it's simply twice that amount.
Of course Alice/Lisa can get all cycles and can do 7MB/s in 16 bits, 14MB/s in 32 bits, 28MB/s in 64 bits (actually double CAS).

Indeed, which is what I meant to say - the bus does 3.54MHz/16bits, which would be 7MB/sec. But the CPU only gets half of it. That wasn't very clear from my text. Mainly because I screwed up dividing by two . Your explanation gets that across better, due to actually doing that .

In any case, the rest of my figures where at least correct

[meynaf]

Yup, anyway one sure thing is that chipmem is deadly slow. 50Mhz '030 has to wait ~28 clocks for a single access :/

[coder76]

Originally Posted by meynaf

Yup, anyway one sure thing is that chipmem is deadly slow. 50Mhz '030 has to wait ~28 clocks for a single access :/

Yes, but all is not lost here, after writing into chip ram, the 68030 can continue executing code in parallell in fast ram, provided no more memory accesses are done during these 28 cycles. This means you can fetch data from memory into registers before doing the chip ram write and then continue calculating stuff in registers only after a chip write. This usually works rather well with a 68030, as some instructions are still rather slow, but a 68040/68060 starts to be too fast to use the idle time well, and there is a limit how much stuff you can read into unused registers before a chip write.

[redblade]

@roondar. Thanks for that long explanation it helped a lot.

Originally Posted by meynaf

16 bits a go at 3.54Mhz is 7MB/s, not 3.5MB/s

In fact the situation seems simpler than that.
The bus runs @3.5Mhz.
However half the bus clocks are reserved for video output, regardless if they're used or not.
Therefore at 16 bits we're limited to 3.5/2Mhz *2 bytes transfer speed, i.e. 3.5MB/s.
And at 32 bits it's simply twice that amount.
Of course Alice/Lisa can get all cycles and can do 7MB/s in 16 bits, 14MB/s in 32 bits, 28MB/s in 64 bits (actually double CAS).

Now if some could mention a 28MB/s 64bit demo that I can watch even a 32bits one
Would be great thanks

[Bruce Abbott]

Originally Posted by redblade

Now if some could mention a 28MB/s 64bit demo that I can watch even a 32bits one
Would be great thanks

Meh. Give me an 8 bit demo that does things which should be impossible on a machine with no Blitter or Copper or sprites.

[ Show youtube player ]

[meynaf]

Originally Posted by redblade

Now if some could mention a 28MB/s 64bit demo that I can watch even a 32bits one
Would be great thanks

Easy. Just setup your WB to 1280x512 in 256c and you'll have 64-bit data access for sure

[Gorf]

Here is a nice an very detailed Blog about what it takes to speed up an Atari ST from 8MHz to 16MHz (and beyond) - not just the CPU but also access to the "ST RAM" (= our ChipRAM).

Very interesting as he tries many ways and documents not only success but also failure. It is not that easy ;-)

The date is being clocked into the shifter at double speed, so the display simply runs out of data each scan line. Apart from that, the GLUE and MMU control the video syncs so they become unsuable also. The only way it could work is to re-build the video syncs, and build a new shifter than can take increased data input speeds, but still output at the correct speeds.


Same would be true of course for the Amiga Custom Chips ...

but I guess Commodore could have done that, if they would have realized early enough that the Amiga is a "computer platform" and not just a one-hit wonder...

I think a real (switchable) 14MHz mode could/should be done by the time of ECS.

[meynaf]

Originally Posted by Gorf

Same would be true of course for the Amiga Custom Chips ...

but I guess Commodore could have done that, if they would have realized early enough that the Amiga is a "computer platform" and not just a one-hit wonder...

I think a real (switchable) 14MHz mode could/should be done by the time of ECS.

But the Atari ST hardware is simple and look at how complicated it ended up. For Amiga it would just be a nightmare.

[graffias79]

I think technology moved so fast throughout the 1990s that you needed to be able to rapid-fire release hardware upgrades to keep up. We went from 25MHz 386s and 68030s at the beginning of the decade to 1GHz Athlons chips at the end. Commodore never stood a chance.