I believe calling the 6809 a 16-bit machine is exaggerating somewhat. It had an
8-bit ALU and an 8-bit data bus, and an 8-bit almost everything else. Even the
6801 had instructions that would concatenate the A and B registers so a single
instruction would operate on the pair, and the fact that the 680x series had a
16-bit index register was more of a hindrance than a help as far as performance
was concerned. The instruction set was very nice, though. Although the
"internal architecture" supported 16-bit operations somewhat, it's a bit of a
reach to compare it with, say, the i8088, which, though I hate the segmented
effective address computation scheme, is, in fact, a 16-bit architecture, as it
has a 16-bit register SET, a 16-bit ALU, and 16-bit instructions that operate on
16-bit operands.
The 6502 had advantages over the MOT-style processors primarily BECAUSE of its
8-bit index registers. Because the index registers were a single byte, the
state machine that controls the physical address computation doesn't have to
decide to skip a cycle when the MS byte of the index is zero, nor does it have
to decide what else to do with the upper byte, since there isn't one. Its 8-bit
character is what makes it faster than the MOT equivalents. It does 8-bit
things VERY fast, so long as they are scoped into table sizes easily handlled
with the 8-bit index registers. Any 4 or 8 bit processor can process 16-bit
data, given enough time and resources. Some do it more elegantly than others,
and I freely admit that the 6809 does it more elegantly. It just doesn't do the
8-bit stuff as fast as the 650x family.
The fact is, the 650x family was a smaller chip than the MOT, Intel, or Zilog
parts of the same class. That suggests it had fewer transistors, since the
number of transistors would increase the chip size. Keep in mind, when
considering this contrast, that the 650x's general purpose registers could be
viewed to reside in the "zero-page" since the access to that region was faster
than to the rest of memory. Having larger registers and more of them, as did
the 8080 and Z80 certainly would make the chip larger and the transistor count
larger.
It has a pipelined data bus, which, in reality, is just double-buffer for the
benefit of the instruction decoder, so it can decode the opcode WHILE the
opcode's first operand, or the next opcode, is being fetched. While this
doesn't speed up an opcode's execution in a test setting, it does speed up a
sequential execution as you encounter when running code. If you look at what
appears on the 6502's pins as it executes various instructions, you'll see that.
(this is documented in MOS Technology's 6500-Series Hardware Manual, so you
don't have to dig out the logic analyzer.)
Dick
----- Original Message -----
From: "Eric Chomko" <chomko_at_greenbelt.com>
To: <classiccmp_at_classiccmp.org>
Sent: Friday, May 04, 2001 10:59 AM
Subject: Re: How many transistors in the 6502 processor?
> Sipke de Wal wrote:
>
> > With hindsight one can consider the 6502 to be the only 8-bit RISC CPU
> > It had a reduced number of registers compared with the 6800 and this and
> > other logic-reduction simplified the design so it could execute code a lot
> > more efficiently as compared with the 6800. Also the 256-bytes Zero-page
> > could be regarded as an (extended) RISC-like registerset of the CPU.
> >
>
> Many looked at the zero-page and the 8 bit stack pointer as shortcomings.
> It was the X and Y index registers both 8 bit, that made the chip interesting.
> The X/Y pair made memory-madpped video graphics easier to implement.
>
>
> >
> > I remember a magazine (BYTE?) describe the 6502 as a true RISC-chip
> > but I don't thing the designers had RISC-CISC philosofies in their heads
> > while working it out.
> >
>
> RISC per se didn't come out until the 80s.
>
> >
> > Only 6809 bas a "better" chip but that should have been a true 16-bit design
> > It came way to late to make a large impact. Only the COCO used it in a
> > homecomputer.
> >
>
> The 6809 was fine as a 8 bit chip with a 16 bit internal architecture, and
many
>
> home computers used the 6809 CPU not just the CoCo.
>
> SWPTC made one. The operating system OS/9 was built around that chip.
> Viirtually every manufacturer of the SS-50 bus had a 6809-based system, and
> that
> would be around a half dozen.
>
> The CoCo may be the most well known but not the only.
>
> Eric
>
>
> >
> > Sipke de Wal
> > ----------------------------------------------------------------------
> > http://xgistor.ath.cx
> > -----------------------------------------------------------------------
> >
> > ----- Original Message -----
> > From: ajp166 <ajp166_at_bellatlantic.net>
> > To: <classiccmp_at_classiccmp.org>
> > Sent: Friday, May 04, 2001 1:03 PM
> > Subject: Re: How many transistors in the 6502 processor?
> >
> > > From: Brian Chase <bdc_at_world.std.com>
> > >
> > > >Does anyone know how many transistors made up the 6502? These days with
> > > >Intel's boasting of the number of transistors their latest processors
> > > use,
> > > >it'd be interesting to know what we used to get by using. What, it
> > > can't
> > > >have been more than a few thousand, right?
> > >
> > >
> > > Memory says it was one of the lower transistor count cpus, very efficient
> > > design.
> > >
> > > >And then it'd be rather fun to implement your very own 6502 using 74*
> > > >series logic chips.
> > >
> > >
> > > I'd bet it would be fairly high chip count. IT would be interesting to
> > > see how fast
> > > you cound make it go.
> > >
> > > Allison
> > >
> > >
>
>
Received on Fri May 04 2001 - 13:15:04 BST