@Hypex

Quote:

Hypex wrote: @bison [quote]Compared to the next big thing, x86 or 80x86 at the time, was terrible. Now if we compare like for like we stick to CPUs below the 80386. It was based on the 16-bit 8086. It had segmented address space, so you had to fiddle with 16-bit offsets and shifting them left to break the 16-bit limit.

The 8086 had multiple segments registers available, which allows it to break the 16-bit limit. You needed the shifting (and masking) only if you wanted to address any memory location having the absolute 1MB address.
Quote:

I don't think so. There was already talk I read of Jay Miner discussing chunky but planar was better for memory conservation.

Unfortunately bitplanes had a big drawback too: they wasted a lot of memory bandwidth, compared to an equivalent packed/chunky mode. And Amigas hadn't that much memory bandwidth...
Quote:

And Commodore like to cut costs. And besides, I see no reason why the pixel width (set at 1) couldn't have been increased to 2,4, and 8 bits for one plane to do chunky later. So increased pixel width as opposed to depth. Same colours, same space required, different packing model. But, looking back to the C16/+4, in multicolour mode they were chunky.

The problem is that the Blitter is bitplane-based, so it doesn't work with packed/chunky modes. You need a proper Blitter in that case, which is also more complex.

Am I the only person that remembers that even the very 1st Amiga wasn't restricted to bit plane gfx? :)
The a1k had zorro slots remember? :)

@fishy_fis: but the o.s. supported only bitplanes...

@cdimauro

Well, in OS3.1 there is some support for chunky GFX (WriteChunkyPixels in graphics.library), but that was just before Commodore banruptcy.

@fishy_fis

Quote:

The a1k had zorro slots remember? :)

Where?

@pavlor

Quote:

pavlor wrote: @cdimauro Well, in OS3.1 there is some support for chunky GFX (WriteChunkyPixels in graphics.library), but that was just before Commodore banruptcy.

But fishy_fis talked about the very first Amiga 1000.

@pavlor

I guess "could have" is more accurate (via expansion port).

To be honest its been so long I'd forgotten it wasn't that way by default.

@cdimauro

Quote:

But fishy_fis talked about the very first Amiga 1000.

"Even very first" to be more precise - that may mean any Amiga since A1000. Semantics again.

@pavlor

Quote:

pavlor wrote: @cdimauro Quote: But fishy_fis talked about the very first Amiga 1000. "Even very first" to be more precise - that may mean any Amiga since A1000. Semantics again.

Yes, semantics AND history AND technical facts: I fail to see how an Amiga 1000, 500, etc. "wasn't restricted to bit plane gfx" (cit.).

Can you enlighten me?

BTW, the Amiga o.s. 3.1 was the first of the post-Commodore era...

@cdimauro

Quote:

BTW, the Amiga o.s. 3.1 was the first of the post-Commodore era...

Never heard about CD32?

@cdimauro

Quote:

Anyway, I have nothing to measure the scalability of each processor family, but looking at the 68060 and compared to the Pentium I feel that Motorola had its hard time to manage such complex ISA.

Motorola went with PPC because it was cheaper to make and
offered better performance per price.Apple advertised it as :
"You can have more speed for less money".Motorola had to
worry about IT/Military/Industry markets too , unlike IBM , so
they didn't mind slower but more quality PPC CPUs.IBM was
the one pushing for higher frequencies.Although it was joint
venture both Motorola and IBM had their own PPC CPUs made
by them.Problem with 68k and PPC was that there was no
chance for them to compete with Taiwan and China's companies.

In the end Apple bought FoxConn (mispelled probably) , I remember
then I sold refurbished motherboards and nobody wanted to buy
those and all saw them as inferior (one of the few Chinese mother-
board manufacturers at the time) so they went bankrupt.

I found them quite reliable , much more then some Epox mobos.

@pavlor

The 68040 can be overclocked to 50MHz and has been on third party Mac asccelerators.

@Aslak3

And, btw, Coldfire would tend to disprove the idea that the 68000 was not scalable.

@cdimauro

Yes the '286 was competitive with the 68000, but not the 68020.
And the '286 was a lousy processor for a multi-tasking OS.
Windows 3.0/3.1 was only recommended for '386 and above.
Even the 68000 was a better choice for more advanced OS'.

@iggy

Quote:

The 68040 can be overclocked to 50MHz and has been on third party Mac asccelerators.

Am486 DX2 has up to 80 MHz and DX4 up to 120 MHz - on the same motherboard.

@pavlor

Quote:

pavlor wrote: @iggy Quote: The 68040 can be overclocked to 50MHz and has been on third party Mac asccelerators. Am486 DX2 has up to 80 MHz and DX4 up to 120 MHz - on the same motherboard.

Yes, up until a few years ago, I had a couple of those boards based on VESA expansion slots.
A bit kludgey, but cool for their time.

@cdimauro

Quote:

You needed the shifting (and masking) only if you wanted to address any memory location having the absolute 1MB address.

I had this old PC book once that taught how to do it in assembler. Can't find it now. But I couldn't get my head around it. It just didn't make sense as it was moving an address four bits to the left. I couldn't see how shifting it by a nibble was useful and looked like it would corrupt. A byte, maybe, a word, sure. But mostly I probably wondered why the straight address wasn't just plonked into a register.

Quote:

Unfortunately bitplanes had a big drawback too: they wasted a lot of memory bandwidth,

They would have as 16-bits or pixels would have been read in at one time which meant four reads from differnet locations for 16 colours. 8 bytes total. For a depth of 8 it's 16 bytes total. So the same amount read in if it was chunky which would have been in sequencial order. So I think it could have worked.

Quote:

The problem is that the Blitter is bitplane-based,

It is yes. It would have had to be updated. So that it would write bit patterns across to match a pallete index for lines, which it could almost do as it had bit pattern support. And images would need a bitmask to represent multiples of packed data going across with an extended width to acount for the extra data. But still being able to specify X,Y and W,H of destination would have helped to do a warped blit. Which would have helped with texture mapped games doing it by hand. And optimised that scale routine in the OS.

fishy_fis 

Quote:

Am I the only person that remembers that even the very 1st Amiga wasn't restricted to bit plane gfx? :)

No but Amiga graphic cards didn't have a real Amiga graphic chip on them. They tended to be PC chips with a nice Amiga boot rom on a video card. AFAIK they couldn't be copper controlled. And only real OS software would work transparently.

Last edited by Hypex on 28-Aug-2015 at 05:46 PM.

@pavlor

Quote:

pavlor wrote: @cdimauro Quote: BTW, the Amiga o.s. 3.1 was the first of the post-Commodore era... Never heard about CD32?

Was its Kickstart marketed as 3.1? Honestly I don't remember.

Anyway, it doesn't change the picture. I fail to see how an Amiga 1000, 500, etc. EXCEPT CD32 "wasn't restricted to bit plane gfx" (cit.).

@blizz1220

Quote:

blizz1220 wrote: @cdimauro Quote: Anyway, I have nothing to measure the scalability of each processor family, but looking at the 68060 and compared to the Pentium I feel that Motorola had its hard time to manage such complex ISA. Motorola went with PPC because it was cheaper to make and offered better performance per price.Apple advertised it as : "You can have more speed for less money".Motorola had to worry about IT/Military/Industry markets too , unlike IBM , so they didn't mind slower but more quality PPC CPUs.IBM was the one pushing for higher frequencies.Although it was joint venture both Motorola and IBM had their own PPC CPUs made by them.

I know, and as I said, before the end of 90s RISCs were cheaper, easier to produce, and better performers. It made sense to invest on the them.

But then the picture changed...
Quote:

Problem with 68k and PPC was that there was no chance for them to compete with Taiwan and China's companies. In the end Apple bought FoxConn (mispelled probably) , I remember then I sold refurbished motherboards and nobody wanted to buy those and all saw them as inferior (one of the few Chinese mother- board manufacturers at the time) so they went bankrupt. I found them quite reliable , much more then some Epox mobos.

I can understand it for motherboards, but why do you think the same for processors?