SEGA Genesis VS Super Nintendo

Not having ever used either one or watched Ben's video, I went with the SEGA Genesis since it has a 68000 CPU.  I have very fond memories of programming and writing compilers for the 68000.  It's a very clean 32-bit architecture and... wait a minute... didn't you say this was a battle of 16-bit consoles?  Why in the world is a 32-bit 68000 considered a 16-bit machine?  Sure, it has an external 16-bit data bus but the internal architecture is all 32-bit.  That would be like saying an Intel 8088 is an 8-bit architecture rather than a 16-bit CPU with an 8-bit data bus.  Grumble grumble whine moan insinuate.

If you believe Wiki it seems its not sure if its 16 or 32 bit.

http://en.wikipedia.org/wiki/Motorola_68000

mcb1, frankly it sounded clear to me that is a 16 bit with internal 32 bit architecture. Despite the wiki I remember this CPU. It was the "revolution" as I have used before the 6502 processor.

Wiki says:

The Motorola 68000 ("'sixty-eight-thousand'"; also called the m68k or Motorola 68k, "sixty-eight-kay") is a 16/32-bit^[1] CISC microprocessor core designed and marketed by Motorola Semiconductor Products Sector (now Freescale Semiconductor). Introduced in 1979 with HMOS technology as the first member of the successful 32-bit m68k family of microprocessors, it is generally software forward compatible with the rest of the line despite being limited to a 16-bit wide external bus.^[2] After 35 years in production, the 68000 architecture is still in use.

mcb1 wrote:

 

If you believe Wiki it seems its not sure if its 16 or 32 bit.

http://en.wikipedia.org/wiki/Motorola_68000

The bit size of a CPU architecture is usually the size of data registers and data ALU.  All the 68000 data and address registers are 32-bit, and arithmetic is all 32-bit.  So internally the 68000 is definitely a 32-bit machine.  A specific chip implementation may have different external data and address bus widths.  The 68020 has 32-bit address and data buses, the 68000 has 24-bit address and 16-bit data buses, and the 68008 has 20-bit address and 8-bit data buses.  Part of the bus width decision is package pin limitations.

If you're planning to support an architecture for a long time it's good to design the architecture for future technology.  The 680X0 architects expected to have future implementations with 32-bit external buses even though they could only do 16-bit data with the first version.

IBM did this with the 32-bit IBM 360.  The first implementations included 8-bit and 16-bit versions which performed 32-bit operations with multiple cycles through 8- or 16-bit ALUs.  I think these used logic made from discrete transistors, or else small-scale integration ICs.

Guys

I was convinced from the start ...

I came into microcontrollers late, so I never did play with these, but it was certainly the day when the hardware and software limitations ensured the designers had to be clever.

So maybe we should report element14Dave poll as false or misleading??

Mark

How did you get this about IBM? Were you a CE?

Not sure if that is the scientific definition, but if I were to use a "32-bit" processor I would expect to be able to:

• Store 32 bit values in a single register and be able to use more than 1 32 bit register at a time.
• Perform basic 32 bit operations such as add, shifting, moving, and multiply and store the result in a single register. (add eax,edx or mov eax,edx)

For a 32 bit "system" I would expect to be able to read/write 32 bit values to memory in a single cycle and/or a single instruction.  I would also expect to be able to address somewhere in the vicinity of 2^32 bytes of memory.  I understand there is some overhead I need to take into account in those numbers but you get the point.

If I had a processor with only a 32 bit accumulator I would not call it 32 bit.  A system with a 32 bit processor and 16 bit bus in my book is not 32 bit.

I would however consider the "system" to be 32 bit if the processor is only 16 bit but there is an instruction set to move 32bits at a time across the memory bus in a single cycle.  I find memory moves, especially to a video buffer, are the most critical for gaming.

That's my definition.  Based on what I see in the videos both systems are 16 bit by my definition, which has not authority what so ever.

Clem Martins wrote:

 

How did you get this about IBM? Were you a CE?

I was an assistant professor for six years.  I taught two levels of computer architecture.  People like to malign IBM as a bunch of salesmen, but in fact IBM was the first to develop a lot of key computer technology.  For example, the IBM 360/91 (shipped in 1967) pioneered many pipeline speed-up techniques like out-of-order execution and register renaming.  A couple years ago I attended an ARM Techcon session on the then-new Cortex-A15 and its pipelining tricks.  It was pretty much the same material as my own lectures circa 1990.

Every your post I read I should add about 10 years to your experience. Just to be curious, how was the life with babbage ?

Joey Thompson wrote:

 

Not sure if that is the scientific definition, but if I were to use a "32-bit" processor I would expect to be able to:

• Store 32 bit values in a single register and be able to use more than 1 32 bit register at a time.
• Perform basic 32 bit operations such as add, shifting, moving, and multiply and store the result in a single register. (add eax,edx or mov eax,edx)

 

For a 32 bit "system" I would expect to be able to read/write 32 bit values to memory in a single cycle and/or a single instruction...

 

That's my definition.  Based on what I see in the videos both systems are 16 bit by my definition, which has not authority what so ever.

That's a good definition.  I also note that the SEGA Genesis case says "16-Bit" in large letters, so it's definitely declaring itself to be a 16-bit system even though it has a 32-bit processor.

OTOH, the original IBM PC considered itself a 16-bit computer even though it was an 8-bit system.  This distinguished the IBM PC from the popular Z-80 systems of the day.  The original Apple Macintosh considered itself a 32-bit computer even though it was a 16-bit system.  This distinguished the Mac from the IBM PC/XT/AT.  So the definitions are somewhat flexible, especially when marketing gets involved

The BeagleBone uses 16-bit wide memory, but it's clearly a 32-bit computer.  OTOH, that 16-bit wide memory is DDR, so you are getting 32 bits transfered per complete clock cycle.  But then does a 32-bit DDR give you a 64-bit system?  At a certain point the debate gets rather silly IMO

Enrico Miglino wrote:

 

Every your post I read I should add about 10 years to your experience.  Just to be curious, how was the life with Babbage?

Chuck was a good guy, but he spent all his time in his machine shop trying to make gears that meshed properly.  And people here complain about having difficulty soldering

(I once saw Babbage's lathe at the Science Museum in London.)

Enrico Miglino wrote:

 

Every your post I read I should add about 10 years to your experience.  Just to be curious, how was the life with Babbage?

Chuck was a good guy, but he spent all his time in his machine shop trying to make gears that meshed properly.  And people here complain about having difficulty soldering

(I once saw Babbage's lathe at the Science Museum in London.)

Oh John Beetem

That explains much on your posts and blogs here. IBM used to show much about the 360 series until a Michigan CS department took the information and made a model 25 run like a higher one. They quietly hide all material to make that impossible in the future.

Clem

Clem Martins wrote:

 

Oh John Beetem

That explains much on your posts and blogs here. IBM used to show much about the 360 series until a Michigan CS department took the information and made a model 25 run like a higher one. They quietly hide all material to make that impossible in the future.

Sounds like microcode-fu.  Those early 360s were microcoded, and sometimes you'd sell a slow model where the micro-engine was exactly the same as the fast model, but the slow microcode had extra NOPs to slow it down so that people would spend more for the fast model.  If you knew how to replace the microcode, you could get the fast model for the slow model price.

Yes, all you need to do is get that big 5 1/4 inch floppy copy that was loaded into ROS. A funny thing I watch was a POR for a 370/158 that needed a foot on the breaker to come up! The CE just "did not get around" to a fix. LOL

I think we could have a great conversation on the early mainframe days when GE and Burroughs were common along with IBM.

Clem