16-bit 6502 vs. ARM or MIPS?

[jmp(FFFA)]

So basically you (or David Cramer) is saying that all of the many advances in processor architecture that have occurred for the past 30-40 years are irrelevant, at least for the "hundreds of different products being made today."

Excuse me if I take that claim with a grain of salt and require further proof before I put my career on the line by telling my employer we should use a 65c02 (core or chip) in the company's next product. There are many factors to consider besides processor bandwidth.

[Arlet]

Quote:

He figures that if they went to 20nm geometry for the 65c02 like the newest Intel processors use, and put the memory and I/O onboard, they should be able to run at 10GHz.

Even if such high frequencies were possible (I doubt it), it is unlikely to have good Watt/MIPS, numbers compared to modern ARM or Intel processors.

[barrym95838]

Yeah, but wouldn't it be super-cool to lose a game of Space Invaders in 0.001 seconds?

Mike B.

[GARTHWILSON]

The fork and spoon are still relevant today, in spite of how many centuries they've been around. For the particular use they're put to, it's hard to improve on them, even if new foods come and go.

[sark02]

There comes a point when a preference becomes a fetish... and trying to compare a 40 year old micro design with a modern ARM is where that line is clearly crossed. I see Forth proponents often do the same(*).

Look, we all love the 6502 (that's why we're here), and it would be wrong to say it's completely obsolete and/or it's a peg that fits no hole, but this kind of proselytizing or claim of relevancy comes across as desperate.

Does the 6502 need to be generally relevant? Can't it just be something we choose to like, and something that still finds a commercial niche?

* Apologies to the Forth proponents, but I also spent a couple of decades reading comp.lang.forth, and the Denver baggage handling system just kept coming up as the, "see! SEE!" claim that Forth had a purpose. It's an interesting language, but it was voted off the island long ago.

[Dr Jefyll]

sark02 wrote:

There comes a point when a preference becomes a fetish... and trying to compare a 40 year old micro design with a modern ARM is where that line is clearly crossed.

[...] something that still finds a commercial niche?

To me, Garth's post claims relevancy for 65xx, nothing more -- a commercial niche, as you say, sark02. I'm glad to find agreement about that, yet on the other hand puzzled by the first part of the quotation, which (this is my first impression, at least) seems to be in reaction to someone stating that 65xx is equivalent to ARM, that they're interchangeable (which is ridiculous). Did someone seem to say that, or am I somehow getting the wrong spin from your post? Maybe there's some other explanation. I'll be happy for any clarification, especially as I suspect the conflict of opinions here will be revealed to be moderate to minimal.

cheers,
Jeff

[barrym95838]

sark02 wrote:

There comes a point when a preference becomes a fetish...

Guilty as charged! Can I still hang out with you guys, or do I need to slink off quietly into the "other" room?

Mike B.

[sark02]

barrym95838 wrote:

sark02 wrote:

There comes a point when a preference becomes a fetish...

Guilty as charged! Can I still hang out with you guys, or do I need to slink off quietly into the "other" room?

Mike B.

There's nothing wrong with having a fetish. Grab a beer, pull up a chair, and tell me all the ways you like to use the X register...

[sark02]

Dr Jefyll wrote:

sark02 wrote:

There comes a point when a preference becomes a fetish... and trying to compare a 40 year old micro design with a modern ARM is where that line is clearly crossed.

[...] something that still finds a commercial niche?

To me, Garth's post claims relevancy for 65xx, nothing more -- a commercial niche, as you say, sark02. I'm glad to find agreement about that, yet on the other hand puzzled by the first part of the quotation, which (this is my first impression, at least) seems to be in reaction to someone stating that 65xx is equivalent to ARM, that they're interchangeable (which is ridiculous). Did someone seem to say that, or am I somehow getting the wrong spin from your post? Maybe there's some other explanation. I'll be happy for any clarification, especially as I suspect the conflict of opinions here will be revealed to be moderate to minimal.

In the podcast that Garth linked to, Bill Mensch talks about running the '816 at 10GHz with everything on chip (all cached) - "we can run as fast or faster than any ARM processor or any Intel processor..." He's saying that their design is simple enough that it can be, and will always be able to be, targeted to any manufacturing process and run at extreme speeds. So far, so good. "What couldn't you do with that?". Well, for one, you couldn't add two 32-bit numbers together in a single instruction. Then he goes on about how he doesn't like Apple because they chose the 68000, then PowerPC and then Intel instead of staying with the 6502, which would have been 10GHz by now if it weren't for the "marketing and capital pressures of big business". He sounds like a crazy and slightly bitter old man yelling at the kids to get off his lawn.

The "fetish" point is when all your weaknesses are strengths and the advancements of other designs are irrelevant because they aren't needed if you really think about it... Don't need a 32-bit ADD when you can just jump to this handy function. Don't need hardware DIV as the software function has been available for 40 years.

"You could build a brand new 16-bit Atari [...] with state of the art technology [...] with the fun legacy stuff". Wait a second, there was a 32-bit Atari evolution back in the 80s... it was called the Commodore Amiga. Nowadays you can get a Raspberry Pi Zero for $5 with a 1GHz ARM, 512MB RAM and 3D graphics. Holding on to the past like this is just a little sad.

One of Bill's last statements is that the 6502, "Continues to be used by the people who want to use it". And that I agree with.

[Aslak3]

sark02 wrote:

... Nowadays you can get a Raspberry Pi Zero for $5 with a 1GHz ARM, 512MB RAM and 3D graphics. Holding on to the past like this is just a little sad.

See this thread - viewtopic.php?f=1&t=3300 - for the things that mater to folks. For me, all this is just:

A way to learn about current tech (things like: MCUs (AVR), Programmable Logic (VHDL), PCB design
The challenge of doing something, no matter how "pointless" it may be
Most important for brain health: an intellectual hobby

Like all hobbies, there is a big element of pointlessness about it. I personally, and please I'm not inviting a debate with anyone at all, think Garth overstates the commercial relevance of these processors. I don't care one way or another though and he is entitled to his view. If it wasn't used commercially in 20 years I think this forum and the people in it would still be here.

Sorry, am rambling. Hopefully this makes some sense. If not, then please carry on!

[joe7]

Thanks to Garth for the interview links.

sark02 wrote:

He sounds like a crazy and slightly bitter old man yelling at the kids to get off his lawn.

Wow, really?

[GARTHWILSON]

sark02 wrote:

"What couldn't you do with that?". Well, for one, you couldn't add two 32-bit numbers together in a single instruction.

Of all the products I've brought to market using PIC16's, I can count on the fingers of one knee the number of times I've needed a 32-bit add, regardless of how many instructions it takes. (I've needed 24-bit, but there was no need for it to be particularly fast.) Most additions and subtractions have been 8-bit, and a few 16-bit.

Quote:

Then he goes on about how he doesn't like Apple because they chose the 68000, then PowerPC and then Intel instead of staying with the 6502, which would have been 10GHz by now if it weren't for the "marketing and capital pressures of big business".

I took that to mean he lamented that Steve Jobs intentionally held the 6502's and 816's speed down well below their capability at the time just so they wouldn't make the 68K look bad. That's a legitimate beef. He's saying that even back then, the 65 family had the potential to go a lot further, but political factors held it back.

Quote:

The "fetish" point is when all your weaknesses are strengths and the advancements of other designs are irrelevant because they aren't needed if you really think about it... Don't need a 32-bit ADD when you can just jump to this handy function. Don't need hardware DIV as the software function has been available for 40 years.

And for many applications, you don't need them at all, in hardware or software.

Quote:

Nowadays you can get a Raspberry Pi Zero for $5 with a 1GHz ARM, 512MB RAM and 3D graphics. Holding on to the past like this is just a little sad.

The RPi has very little general-purpose I/O; and with the things it does have, you won't have any hope of every understanding every bit of all the software layers right down to the metal. Where will the next generation of engineers come from? They're all conditioned to have "all this stuff already done for you," including 3D color graphics (which turn me off, because of the rampant misapplications); but someone has to know the stuff down to the metal. Someone has to get their hands dirty. It's not the ARM that's preventing that. It's the RPi.

[Arlet]

Quote:

Of all the products I've brought to market using PIC16's, I can count on the fingers of one knee the number of times I've needed a 32-bit add

32 bit operations are used a lot in products you didn't work on.

Quote:

you won't have any hope of every understanding every bit of all the software layers right down to the metal. Where will the next generation of engineers come from? They're all conditioned to have "all this stuff already done for you

There's too much stuff now to understand every bit of everything. However, one engineer can still understand USB to the metal, where an other can understand 3D graphics, or floating point processors, or memory management, and so on. Using each other's efforts, they can built a complete working system in a reasonable time.

[jmp(FFFA)]

GARTHWILSON wrote:

The RPi has very little general-purpose I/O; and with the things it does have, you won't have any hope of every understanding every bit of all the software layers right down to the metal. Where will the next generation of engineers come from? They're all conditioned to have "all this stuff already done for you," including 3D color graphics (which turn me off, because of the rampant misapplications); but someone has to know the stuff down to the metal. Someone has to get their hands dirty. It's not the ARM that's preventing that. It's the RPi.

If you want I/O, buy a BeagleBone (http://beagleboard.org/bone) or any one of a number of other alternatives with more I/O than you can shake a stick at.

Also, it seems to me that you are lamenting the fact that most modern systems are too complex for any one human mind to digest in any reasonable period of time. Yet the alternative (of keeping all systems simple enough so that at least the best of us can fully comprehend them) would seriously limit human progress. I probably share many others feelings of empowerment in understanding in detail how a 6502 operates from the transistor level though the application software layer -- a feat I would have no hope of accomplishing on any modern system -- but that in no way makes it better than what forty years of evolution in the semiconductor industry has produced. Furthermore, any advantage I may obtain through my detailed understanding of how the hardware and software on a 6502 system operate is more than offset by the much greater potential offered by newer hardware.

I acknowledge that the 6502 is "good enough" for a handful of modern problem domains, but would argue that there are much better choices in terms of cost, power consumption, and yes, even simplicity (from a practical viewpoint).

[Aslak3]

jmp(FFFA) wrote:

I acknowledge that the 6502 is "good enough" for a handful of modern problem domains, but would argue that there are much better choices in terms of cost, power consumption, and yes, even simplicity (from a practical viewpoint).

And don't forget time-to-market. A case in point: I was surprised to see an LCD module (64x256 or so mono) come equipped with an ARM MCU. No doubt a 8 bit-er would have done the job quite nicely, but I'll bet they saved a great deal of effort getting the code written, debugged, etc having the "utility" of a 32 bit controller.