trend of "ARM"... will this replace all other micro-controller and ...

in

obsolete.

web-based

,

there

suffice,

expense cannot be justified.

8 bit chips will always be

Ow, that's nonsense. Engineers will ALWAYS be able to program or at least read assembly, even 100 years from now. I agree that in the very near future virtually all programming will be done in C (except on those cheap Chinese goods with $0.10 processors) but it won't lead to extinction of assembly programmers.

I wasn't presenting a 'meaning'.

True, but I never claimed it applied to years outside that graph.

The graph merely shows the trends in fairly recent years:

8-bitters far outsold 32-bitters, and sales were rising far faster.

This is contrary to the o.p.'s statement that things like ARM would replace

8-bitters.

I can't say what the trends are right now, but I would expect they haven't changed much.

Rickman asked if I extrapolate the stock market. If I could do that, I'd be rich! Lots of people do try of course, and we all hope stocks grow long term.

I agree it is much better to know the causes that create the effects.

The stock market is notoriously hard to predict because the causes are numerous and complex.

I'd guess proportions of processor sales are less so because wider data paths use more silicon and tend to cost more and manufacturers don't want to pay more than they need.

Of course, things change so maybe someone will find a killer app that raises demand for 32-bit chips by a factor of 10...

It was only a couple years ago that 8-bit processors passed 4-bit ones in volume (units sold). I think that 8-bit processors passed 4-bit ones in dollar sales a year or two before that.

When compared to 4 and 8 bit parts, unit sales of 32 bit parts has been and probably still is negligible. Dollar sales, OTOH, is a different matter.

--
Grant Edwards                   grante             Yow!  Where's th' DAFFY
                                  at               DUCK EXHIBIT??
                               visi.com

I don't think you understand my point. I am not talking about some frill that an engineer wants to add. I am saying that marketing always wants to put more into a product. If the incremental cost is very low, then those "frills" will start to be added.

Of course no one will use a 32 bit processor in a greeting card as you see them today. But when the cost is low enough to allow, you will see

32 bit processors in even disposable things like digital ink newspapers.

The issue in the cell phone is that the game takes up code storage space. That is an added cost unless you say "there is spare space". But at some point of cell phone development that game pushed the size of the flash up to the next notch or required another round of code reduction to make it all fit. Nothing is free, but often the cost is low enough.

--

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

I like your self assurance. RIGHT NOW it is far easier to hire a C programmer than an assembly programmer. I did not say they would become "extinct". I simply said they would become hard to hire as they are becomming NOW.

--

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

Especially since he doesn't know what I make a year...

--

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

Hi,

What's wrong with GCC?

It is absolutely everything but unreliable, the chance that bugs are found and eradicated is very high, because it is used much more than most commericial compilers (possibly except MS stuff, but I wouldn't call that better)

Another advantage of GCC is that it is portable between architectures, if you decide to switch to another processor or playform you can still use gcc. You'd have to buy and port all your code when using a CPU or platform dependent commercial compiler.

I can't think of any disadvantages of GCC except it is quite difficult to set up. Once setup it works like a charm though..

Wumpus

No, I don't think you understand MY point :) The marketroids - let's take toys for instance - have a range of trucks, say, nd they want to add one more product to that range. They come to the design people, who make a huge list of possible features. The engineers implement a prototype with some subset - perhaps a large subset - of those features. The marketing people then say "Great! Now make it for $x, and take out whatever features are required to achieve that!". This is not theory, it's bitter experience of how the mass-market consumer electronics industry works (except for very small companies). You simply cannot persuade marketing people to throw in extra features after the fact unless they are /literally/ free.

$0.10, by the way, is a massive cost factor in an electronic toy. I have spent upwards of two weeks trying to find a way to use two $0.0025 resistors instead of one $0.05 capacitor (in a toy with SRP $12.95), the end effect of which was to cut out the entire feature that required this capacitor, rather than spend $0.05. I can project from this that in a $79.95 microwave oven, a $0.10 BOM increase is still very significant, and days will cheerfully be spent to shave it out.

--

-- Lewin A.R.W. Edwards

Learn how to develop high-end embedded systems on a tight budget!

It's all about profit in high volume stuff (toys, greeting cards, etc), when 32 bit processors are the same price or cheaper, and are the same size or smaller, and use the same power or less, than 8 bit processors, then 8 bit processors will go away. Otherwise there will be niches that 8 bit processors will fill over 32 bits. (As there are still niches being filled by 4 bit processors, though I'm sure they are becoming more and more scarce.)

One thing I can think of that has worked this way is memory. Because of mass-production, it's cheaper to use a 32k RAM part than a 2k RAM part, assuming you can even find a 2k part (or even 32k for that matter!).

It could be processors will go the way of memory in price, but unless they can build 32 bits smaller, and have them use less power than 8 bits, there will probably be applications where spending *more* money for an 8 bit processor would be a requirement.

NASA, for instance, would probably place more emphasis on its current consumption budget over a processor's ability to play Tetris using its spare computational power. So even pricing won't necessarily kill the 8 bit processor.

-Zonn

-------------------------------------------------------- Zonn Moore Zektor, LLC

Remove the ".AOL" from the email address to reply.

Very little, however it is not the be-all-and-end-all.

It's rarely the best compiler for smaller micros. In my experience, on anything with (roughly) less than about four general-purpose registers capable of holding addresses or natural-sized integers, you're better off with a compiler that's been crafted with the constraints of the architecture in mind - there's too much juggling otherwise for it to be time- and space-efficient. (e.g. the gcc ports for HC11/HC12).

It's also rarely the best compiler for very powerful micros - for example, Intel's own compiler eats it alive on high-end x86, in my experience, since Intel can afford to put a lot of chip-specific peephole optimisations into their own toolchain. (And do a hell of a lot of expensive dataflow analysis).

GCC is quick to port, generates tolerable code on most common 32-bit architectures, and comes with a lot of useful tools. It's capable of keeping most 32-bit RISCS fed and watered quite nicely.

Bugs in the core of GCC will be hunted down and slain quickly. Bugs in ports - whether ports to a new host system or back-ends for a new processor - might linger for some time.

See above. For most platforms it's a good solution. For some it isn't. For others it's good but there are better ones...

One size never fits all, at least not perfectly.

pete

--
pete@fenelon.com "there's no room for enigmas in built-up areas" HMHB

I understand. But an $80 microwave is not a $13 toy. The toy is planned with a well defined set of features that fit the price window. A microwave will have more flexibility to *plan in* features if the cost is not significant. How many microwaves do you find that *don't* have a menu of standard foods or other features that could be left out and still not impact the basic funtion of a microwave? The cheapest one I see in Walmart still has those features. I expect the couple of extra pads on the keypad alone increase the cost by $0.10.

But enough. There will always be products that do not have *any* use for a 32 processor no matter how small the incremental cost. Just as we still have 4 bit apps now when the cost difference is very slight. But clearly the trend will change as the cost of the 32 bit parts comes down. I expect the 8 bit parts will only dominate for a few more years.

--

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

32 bit processors, as used today, usually need quite a few support processors.

An intel based PC for example:

Base processor: A 32 bit (64 bit?) 80x86 derivative.

Support processors (of unknown bitsize)

Keyboard controller inside motherboard (usually hidden in a chipset now days). Keyboard controller inside keyboard. Monitor control (OSD, autosync, etc) Modem Display card (possibly more powerful than the base processor in many ways) Hard disk. CD rom. Printer. Newer speaker systems with digital volume controls and USB support. Mouse Network routers, switches, hubs, etc. Now days stupid stuff like fan controllers w/digital temperature readout...

Counting the main processor used to write this post, I just counted 24 working CPUs on my desktop. (Includes a couple of HP calculators, 3 remote controls, a wireless desk phone (1 CPU in the handset, 1 in the base), etc, etc...

It's incredible the number of places you find CPU's now days! Someone of 20 years ago might have made the argument: CPU's will never be as cheap as discrete logic, they'll always draw more current, and be more expensive. And they would have been wrong, so I know where you're going with this!

It's not hard to see why currently, small processors, outsell the larger ones by a good margin, at least qty wise (according to the Mouser distributor at my last job), I'm sure the gap is smaller dollar wise.

But then using the "discrete" .vs. "CPU" analogy you can predict that processors are going to get more and more powerful for the same price/performance ratio, and part of this is going to be an increase in the internal data path and register size.

Someday the cheapest, most energy efficient way to blink an LED, just may be with a 32 bit processor and a battery. And hopefully google will have archived your original post so you can say "I told you so!". ;-)

-Zonn

-------------------------------------------------------- Zonn Moore Zektor, LLC

Remove the ".AOL" from the email address to reply.

Maybe you had the wrong approach. You do not have to pay ARM anything upfront to do an ASIC, if you go to an ASIC vendor which has already licensed the core. You will normally pay a per die licence, but for a vendor like Atmel which has delivered tons of ARM ASIC, this is not very high. For the 50ku volume , you can look at Atmel AT91 std ARM products.

--
Best Regards
Ulf at atmel dot com
These comments are intended to be my own opinion and they
may, or may not be shared by my employer, Atmel Sweden.

Linux compiled on GCC + some apps = 1,7 MB. Linux compiled on proto Green Hills compiler + same apps = 1,2 MB.

If you can provide 30 % code reduction , it is always interesting.

--
Best Regards
Ulf at atmel dot com
These comments are intended to be my own opinion and they
may, or may not be shared by my employer, Atmel Sweden.

there

suffice,

to

Price for a chip is related to the chip size, and if the chip is pin limited then the 32 bit CPU has the same price as the 4 bit. If the design is core limited, then the extra die size of the 32 bit core will add cost to the die. The ARM7 core is but 0,7 mm2 in 0,18 micron and quite frequently the micros are around 25 mm2. I think that an ARM7 with 8 kB ROM , 1kB RAM in a 16 pin package easily should be sub $1 today.

for

--
Best Regards
Ulf at atmel dot com
These comments are intended to be my own opinion and they
may, or may not be shared by my employer, Atmel Sweden.

One problem with the ROM approach is that with ever decreasing technology the number of chip you have to order will increase significantly.

Assume a 10 mm2 chip on a 12" wafer in the future. The area of the wafer = 73000 mm2 and you may get maybe 6500 good chips out of the wafer. Each wafer lot = 24 wafers , so your minimum order is about 150,000 units...

We are not there yet, but ....

--
Best Regards
Ulf at atmel dot com
These comments are intended to be my own opinion and they
may, or may not be shared by my employer, Atmel Sweden.

Which is pretty much what I was saying, but you cut my text.

pete

--
pete@fenelon.com "there's no room for enigmas in built-up areas" HMHB