eslapion wrote:
bloodline wrote:
But the 68K isn't the x86... the 68k is a fantastic chip, full of features and a nearly orthoganal design... while these make it a dream to program... they also hinder it's maximum speed. Intel got lucky (sort of) that the x86 is a very simple processor, from a CISC point of view, it's very easy to bolt it's ISA onto a nice fast RISC type core, it has very rigid addressing modes and rules as to what instruction can be used and when... these things lend themselves to modern CPU design... The 68k simply isn't that simple... it's too versatile... if you try and make it more suitable for hight speeds you end up with the mostly incompatible coldfire... and if you push it harder, you'll end up with something that pretty much looks like an x86... in almost every way... and you'll be wishing you'd started from scratch (or used a MIPS :-) ).
Something I don't understand is... why not just clock the damn thing faster?
That faster you clock it, the more electricity it needs, the hotter it gets and due to propegation delays, parts of the chip go out of sync... all of these things end in a crash.
To solve the power problems you try and use a smaller manufacturing process, this requires a chip redesign so that the smaller components don't interfere with each other.
Look at the first Pentiums. They began running at 60MHz powered on 5V, then they took the same architecture and used better transistors to make the thing run at 75 MHz and quickly pushed it to 133MHz.
When designing a chip, you try to plan how fast you want the clock to run. The Pentium was released at 60Mhz (at the very begining), but the design was made with the idea of a faster clock rate... as production improved (as chip design and manufacturing process never stop delvelopment), the chip will (as predicted by the designers) be able to clock faster.
A year of research in new materials... still no redesign and it went up to 200MHz.
Plently of redesigns are made, a chip is never just designed and then made... the designers work constantly to improved the layout and adapt the design to new processes and materials.
Why not do the same with the 68060? They could have done that easily, I am quite sure.
If you wanted to make an 68060 now, you would have to find an old Fab that could manufacture such an old design, then get the masks from Freescale, but all you'll end up with is probably a few 40Mhz 68060 chips... until the engineers that you are paying managed to get yealds up and iron out any bugs in the process so that you can get them to clock up to 50... or even 80MHz (at a push).
If you wanted to build it on a new process... 65nm? Then you would have to hire a team of chip designers and buy time in a test Fab. You'd probably have to design the chip from scratch as the last revision is 8 or 9 generations old. This would cost millions... now given that the 68k architecture is not suited to High speed operation so you couldn't really take the "Pentium" approach of a RISC 68k style core with a nice decoder in front (well you proably could but it would be a expensive and time consuming to get right... perhaps would could depreciate most of the 68k addresing modes, but then why usea 68k?), and the fact that you don't own any patents on modern cpu chip design... you would probably end up with a chip only two or three times faster than the existing 060... you'd need a massive die size, it would consume a large amount of power, it would have no Vector unit and a very old CPU bus (so no modern support chips)... coupled with the fact you don't have anyone to sell it to... all you end up with is a very large bill., and something that if it would work at all, would be slower than UAE with a JIT running on a PC that cost £500.
This is why Motorola dropped the 68k in favour of the PPC, all those years ago. There was no gaurenteed market for continuing the 68k... but with Apple and IBM on board, they knew they could sell the PPC (to apple and IBM) which was better suited to high performance computing (smaller die, lower power consumption, and 64bit ), and thus cheaper to get to market.
