• Please review our updated Terms and Rules here

Most advanced 8-bit CPU ever...... 6809? 65816? Z280? eZ80? AVR? Opinions?

L

lowen

Veteran Member
Joined
Jan 16, 2014
Messages
1,684
Location
Western North Carolina, USA
Seeing the other thread dealing with the 6809, 65802, and 65816, and not wanting to hijack that thread, I couldn't help but wonder which 8-bit CPU really would be considered the most 'advanced.' No, I'm not wanting to start any flames, either; I'm genuinely curious, and if it can be shown that the 65816, for instance, is demonstrably the most or one of the most advanced CPUs, that's wonderful. I'm just curious; nothing more.

There are 8-bit CPUs still in production, of course, including the Z80, Z180, eZ80 (in several forms such as eZ80F91), AVR (again in several forms with ATMega328 being probably the most popular). For that matter, there is the 68008, but is that really an 8-bit CPU? Is the 8088 really '8-bit?' I'm only going to consider the 'bittedness' of the basic instruction set for my purposes: that is, are the instructions byte-oriented or are they 16-bit (or 32-bit) word oriented? Both the 68008 and the 8088 have a basic instruction opcode size larger than 8-bits, right?

What is considered 'advanced' for the purposes of this discussion? Well, a multiply instruction of some sort is probably the minimum. But what else makes a CPU 'advanced?'

The Z80 is definitely not 'advanced' these days, nor would the 6502 or 8080 be considered 'advanced.' The 6809 can, as well as the 65816; both of those families are quite advanced. I couldn't tell for sure, but is the 65816 still in production? That would be fantastic if so.

I'm going to advocate the Z280 as one of the most advanced 8-bit CPUs ever, based on a number of factors:
1.) 24-bit addressing through a paged MMU;
2.) 8 and 16 bit multiply and divide (yes, a 'DIVW DEHL,BC' is a legal instruction: this divides the 32-bit long in the register quad DEHL by the 16 bit word in register pair BC, leaving the quotient in HL and the remainder in DE);
3.) Many addressing modes, including the 'program counter relative' mode that allows a fully relocatable program to be written, including subroutines (relative calls, and relative jumps with displacements to anywhere in the 64K logical address space);
4.) The ability, in the 16-bit Z-BUS bus interface mode, to have 'co-processors' (Zilog's term is EPU) similar to the coupling of the 8086/8087 or 8086/8089; the 65816 has a similar facility;
5.) A 256-byte line-oriented on-chip cache that is available for instructions only, data only, or instructions plus data. It can even be turned off with the 256 bytes allocated as lines in memory.
6.) Multiprocessing (yes, you could build a dual, triple, or more Z280 machine, and the processor has hooks to support global versus local buses);
7.) A massive instruction set with many useful instructions (computed CALL, for instance: 'CALL (HL)' is the RETurnable version of 'JP (HL)'), although the instruction set could not be accused of being very 'regular;'
8.) 4 on-chip DMA channels with 24-bit addressing, on-chip counter/timer units, and an on-chip UART.

I'm sure there are others, but these are the ones that jump out at me. The very advanced Z280 was so far ahead of its time that many bugs exist in early masks and revs, but later revs seem to have fixed many of these.

Anyone have other opinions? If you count only bus width, the 68008 would likely be the most advanced 8-bit (data bus) CPU of all time, but even though the bus width of the Pentium is 64-bits no one calls it a 64-bit CPU; the Z280 works with 8-bit instructions (and instruction prefixes) even though one of the bus modes is a 16-bit Z-BUS one (it can also operated in a Z80-compatible manner with an 8-bit data bus).

Now, in my opinion the Z280 is far more advanced than the HD64180/Z180, even though the Z180 is still being made, and I would say that the Z280 is more 'advanced' than the still-in-production eZ80, even though the eZ80 is far faster and has its own advances, such as the ADL addressing mode suffixes (actually implemented a a prefix at the machine code level).

Too bad Zilog stopped production in 1996. Right now there are several eBay auctions for Z280's; many of these probably came out of some of the point-of-sale and other embedded systems that used these grand old chips to good effect.
 
Last edited:
Have you considered the MEGA65 in your survey? Hypervisor, virtual memory, 28 bit address space...

You can do some wonderful stuff in a modern FPGA nowadays.
 
Last edited:
Chuck(G) said:
Have you considered the MEGA65 in your survey? Hypervisor, virtual memory, 28 bit address space...
Click to expand...

No, I hadn't, primarily because I didn't know it even existed. That's one reason I started the thread; I know that there are far more architectures out that that the few I've been exposed to. Hmm, looks cool.

Virtual memory is definitely an 'advanced' feature. The Z280 has hooks for this with page invalidation bits in the PMMU with a general access violation trap available, but it's pretty primitive by, say, 386 standards.

Hypervisor capability, aka hardware virtualization, is pretty advanced, too.

And, yes, I'll count a VHDL core as a CPU......

(EDIT: the actual core for the Mega65 is an advancement of the 4502 used in the C65. The new core is referred to as the GS4510 in the document I'm reading.....)
 
Last edited:
I did some work with an 80251 core microcontroller 20-some years ago. Somewhat of an acquired taste to it but not bad once I got used to it. Up to 256KB address space (as implemented, up to 16MB in theory), 8-bit and 16-bit multiplication and division, some 16-bit and 32-bit operations, useful addressing modes. Not making any claim that it's better or worse than others, just that I found it to be reasonably advanced for for what I needed to do with it at the time.

8051 cores seem to live on forever. The 80251 core seemed to have a short life. I don't know if there are any out there anymore.
 
I must play pedant and point out that the 65816 and of course 65c02 are very much in current production; the 6502/65c02 probably being the most prolific microprocessor of all time. (Maybe surpassed by ARM eventually at the rate we're going, but on the other hand, maybe not).
 
I recall the PC simulator done by Valentin with his souped-up 8052 (80 MHz?). It was interesting and he did an amazingly good job.

But, what the heck, talking about 8 bit CPUs is a bit silly in today's world of cheap silicon. A 72MHz 32-bit ARM Cortex MCU is probably just as cheap as any 8 bit version--and many times faster. PIC32MM MCUs, the same--dirt cheap--although I think that MC shot themselves in the foot with their "Harmony" IDE.

We live in a world of RPi and similar gear that will run rings around any legacy system. There's no fighting progress.
 
32 bits don't always mean faster. There are plenty of cases where at the same clock speed, an 8 bit processor outperforms a 32 bit one.
 
KC9UDX said:
32 bits don't always mean faster. There are plenty of cases where at the same clock speed, an 8 bit processor outperforms a 32 bit one.
Click to expand...

Sure, just point me at a 250MHz 8-bit processor and we'll run them side-by-side :) The other thing is that modern cheap MCUs have a huge variety of builtin peripherals. Not so much for 8 bitters.
 
Chuck(G) said:
Sure, just point me at a 250MHz 8-bit processor and we'll run them side-by-side :) The other thing is that modern cheap MCUs have a huge variety of builtin peripherals. Not so much for 8 bitters.
Click to expand...
As far as I know, you can order 65c02 and possibly 65816 at that speed. But not in a package by itself, and not in small quantities.

If I weren't using a slow old small "phone" right now, I could probably cite examples of propietary industrial control chips with 65c02 cores inside running at high clock rates. They can be found in the most surprising places.
 
I can think of several 4-bit MCUs; the Rockwell PPS-4; National Semi COP400 (used well through the 1990s), TMS1000,....and a few Japanese chips.

My point was that modern CPU/MCUs often have advanced features (ARM brags that the Cortex MCUs often have a smaller code footprint than most 8 bit MCUs). An FPGA implementation is never going to run faster than a contemporaneous dedicated silicon version. Many have master GPIO ports for interfacing onto, say, ISA bus.
 
Chuck(G) said:
I can think of several 4-bit MCUs; the Rockwell PPS-4; National Semi COP400 (used well through the 1990s), TMS1000,....and a few Japanese chips.

My point was that modern CPU/MCUs often have advanced features (ARM brags that the Cortex MCUs often have a smaller code footprint than most 8 bit MCUs). An FPGA implementation is never going to run faster than a contemporaneous dedicated silicon version. Many have master GPIO ports for interfacing onto, say, ISA bus.
Click to expand...

The 65xx licensed IP running at high clock rates is not FPGA as far as I know.
 
Well, while 4 bitters are interesting and all, the question was about 8 bitters.

And while ARM and others are interesting and advanced, well, they're also not 8 bitters.

TMS7000 would be interesting and on-topic; TMS1000 not so much.

65C02, 65C816, Z80, Z180 are all on current production. In fact, if you have a DVD burner, there's about a 50% chance it has a Z80 of some type. Doesn't need to be 250MHz; speed alone doesn't indicate it's advanced. The SOC-Z80 runs as if it were a 200MHz Z80; that alone doesn't make it advanced.
And popularity isn't everything; the single most popular 8-bit CPU of all time would be the PIC; the PIC16 has shipped over 7 billion units.
 
Last edited:
It's okay playing it your way with me. But what do you mean by "advanced"? (Much less what you mean by "8 bit")

So, are you looking for advanced features, advanced performance or what? Large ISA? WCS? VLIW? Vector facility?
 
KC9UDX said:
the 6502/65c02 probably being the most prolific microprocessor of all time.
Click to expand...

Not even close. The Z80 probably outsold the 6502 10-to-1 and was far more popular in industrial equipment, and even that is nothing compared to the number of 8051s out there. The 6502 is little more than an also-ran outside the late 70s/early 80s home computer market.
 
FishFinger said:
The 6502 is little more than an also-ran outside the late 70s/early 80s home computer market.
Click to expand...
It sounds like you're forgetting that the 65C02 is still in production. Well, the Z80 is too, but I bet that the 65C02 outsells the Z80 by a large margin. Mostly as cores inside something you would never guess had a 65C02 in it. Or, of course, if you have an implanted defibrillator. It'll have a 65C02 CPU, still the only CPU certified for body implantation.

On the other hand, the 65C02 can't beat the ARM in sales numbers. But it's not an invalid comparision - the wdc site says that about 6 billion 65xx processors have been shipped, and they sell several hundred millions per year. I believe the Z80 doesn't outsell that, unless somebody has some evidence I haven't found yet. But ARM Holding claims to have sold about 15 billion processors in total, but more imporantly, at least 3 billion ARM processors are sold every year now. That's somewhere around three times x86. And about an order of magnitude more than for 65xx.

Well, about advanced 8-bitters.. the 6809, and even more so the Hitachi 6309 are nice, but they are not in production. Sometimes they're claimed to be 8/16-bit processors, not pure 8-bitters. But then again that discussion will end up nowhere, before we know it we'll start discussing what the 68k is. Again.
 
You must log in or register to reply here.
Back
Top