Ok, I drink the cool aid and accept that all 68000 systems are 16/32bit and also jump around happily that someone is working hard to give us 64bit execution capable m68k!! Seriously, this announcement is HUGE, it should have its own story!! :banana: :banana:
Well 68000 did have 32bit instructions
This operations "32bit-val" + "32bit-val" = "32bit-val" could be done with a single instruction.
So by this definition the 68000 was a 32bit CPU.
Also the registers of the 68000 are 32bit wide.
Having 32bit wide registers also would make it a 32bit CPU.
Later 680x0 CPU did then instructions faster.
This chart shows very nicely the improvement over time.
http://www.apollo-core.com/bringup/minibench.pdfWhile the 68030 was running AMIGA OS ver fast
You see very nicely how "weak" the 68030 design was compared with an 68040 or compared with the 68060.
For example doing this instruction:
ADD.l #$1234567,D0
The 68030 @ 25 MHz only reached about 4 Mips
The 68040 @ 25 Mhz reaches about 23 Mips
You see here clearly how the CPU was internally improved.
The 68060 if running @ 25 MHz would have reached 15 Mips here.
The 68060 was clearly the best 68K core but its design also had some weak spots.
Phoenix in the Vampire1 @ 25 MHz would reach 21 Mips
Phoenix underperforms here a little as we had to cut of an arm and a leg to make it fit the Vampitre1.
Phoenix in the Vampire2 @ 25 Mhz does reach 42 Mips
See see here clearly the evolution of the cores..
If you compare the 4 Mips of the 68030 with the 42 Mips of Phoenix you see that at the same clockrate the improved internal CPU architecture gives a 10 times speed up.
I think this example shows again how pointles comparing a Core by the clockrate is.
Of course a lot more then 25MHz can be reached....
I wonder if "grading" the core releases by their minibench scores would make sense.
Or maybe introducing a virtual clockrate like "performaning like a 68030 @ 400 MHz" would be usefull?