When will the 8051 and othe 8-bits go away?

All those advances that make the 32-bit parts cheaper and lower power are also making the 8-bit parts cheaper and lower power.

8-bit parts are always going to be cheaper and lower power than 32-bit parts. 4-bit parts are always going to be cheaper than 8-bit parts.

I'd guess at least 10 or 20 years. Sales volume of 8-bit parts just passed 4-bit parts a couple years ago.

Why would they? 4-bit parts are still sold in huge volumes. They're not going away either.

No.

Even if in meant having to pay more for something and have the batteries last half as long? If you owned the company and putting an 8-bit part in a product could increase your profits by a million dollars, what you you do?

--
Grant Edwards                   grante             Yow!  I once decorated my
                                  at               apartment entirely in ten
                               visi.com            foot salad forks!!

Never. 16 bits is overkill for a talking barbie, keyboard controller, toaster, mouse, or low-end calculator. The only reason more toys aren't using 4-bit processors is because so much of the die is taken up by bonding pads that there is no cost savings to be had.

Pins? We don't need no steenking pins! Pry open a toy or a calculator and try to find the uC pins. :)

Not as long as there is a tenth of a penny to be saved by using a less capable part. That's a thousand dollars per million units.

Not really. We have reached the point where the die size is the same, because the bonding pads dominate.

Less than 5% of the market compared to 8-bits having 30-40%.

The place where 4-bit processors are still king is at the very low power end - watches. They want that tiny battery to run the device

27/7 for many years, and fewer transistors still means lower power. I don't see that ever changing.

Yes, I should have thought about that. Below a certain die size (for the active circuitry) the packaging cost dominates.

That has changed a lot. The last time I looked 8-bits were only slightly ahead. That probably was a couple years back...

Yup. Somebody recently told me that Swatch does their own uP designs now. Current draws down in the 10s of uA. Not sure what the clock rate is. Once upon a time 32KHz was common for watch stuff.

--
Grant Edwards                   grante             Yow!  Quick, sing me the
                                  at               BUDAPEST NATIONAL ANTHEM!!
                               visi.com

My predictions:

4-bits will only be used for extreme low power. 8-bits will never go away. Perfect for toys, keyboards, etc. 16-bits will not exist by the year 2020. 32-bits will be here forever. Great size for demanding embedded apps. 64-bits will not exist by the year 2040. 128-bits will not exist by the year 2060. 256-bits will become the top-end. 512-bits and higher will never happen. Further development will be in the direction of massive parallel processing on one die, followed by (unless it turns out to be impossible) quantum computers.

I don't see 16 bits surviving. It will be squeezed out by 8 and 32. I don't see 64 or 128 bits surviving. They will be squeezed out by

32 and 256.

Thanks for the article pointer, Guy. Very interesting.

Followup question:

Assuming the 8051 is the dominant 8-bit microcontroller for NEW DESIGNS(*), do you think it will continue to be, or is there a new rising star?

• If my assumption is wrong s/8051/dominant_mcu/

Cheers, Paul.

I agree with those two statements.

I think that 16 bit will be a small segment of the market but still worthwhile pursuing for some manufacturers. It may just be that the cores will be soft ones as part of a device with integral I/O or as part of an FPGA.

I agree with that statement.

This may be a bit on the early side. However, they will prove too big for most embedded applications and will only find a home in bigger processing solutions for a while.

I do not think that any bigger than 128 bit will gain any sort of a foothold for very long. I see the highest end becoming a much more multiply parallel architecture to deal with really huge applications (MPP on a big scale). These are likely to use mostly the 32 bit processors at ever increasing processing speeds. We may find that the highest end processing starts to move over to different technology (optical computing perhaps). In that sector I suppose that 32 bit wide optical busses will be all that they desire to handle.

Just an opinion for what it is worth.

--
********************************************************************
Paul E. Bennett ....................
Forth based HIDECS Consultancy .....
Mob: +44 (0)7811-639972
Tel: +44 (0)1235-811095
Going Forth Safely ....EBA. http://www.electric-boat-association.org.uk/********************************************************************

I think that's already pretty visible. Most of the traffic around here is about either 8 bit (PIC,AVR,8051) or 32-bit (ARM,PPC,SH,H8S) parts.

--
Grant Edwards                   grante             Yow!  LOOK!!! I'm WALKING
                                  at               in my SLEEP again!!
                               visi.com

The move to 64-bits is mostly due to the 4GB address limit on 32-bit machines that is a real barrier for large systems like database servers. 64-bits provides for 16EB of directly addressable memory.

Assuming RAMs double in capacity every 18 months it will take 48 years to build systems that hit the 16EB limit. That assumes it's even possible for the RAMs to continue to evolve that far.

I think 64-bits will be with us for 100 years if not longer. The trend now is for parallelism over speed. Disk drives are still the speed-bump for all high performance computing.

Not even Microsoft can drive demand more than 16EB.

Cheers, Paul.

Throw away items (e.g. muscial cards)...

Lots of simple low speed applications don't require anything else e.g. small security alarm systems, remote temp sensor to serial, house heating controller.

As others have yes and no, either as soft core or for some embedded apps where there is a cost advantage (on PCB layout) or other advantage.

Yes..

Above this becomes the problem of NOT processing power but for an embedded app that is not a 'PC' like appliance, making the boards small enough with all the front side and other bus issues and the number of tracks and pins and bus widths using so much board space.

The amount of interconnects on a 32 bit system and more so on a high speed 64bit system are a pain already.

The amount of chippery connector sizes and debugging is only worthwhile for dedicated parts of systems (high end graphic controllers) and specialised systems that are more akin to existing mainframes now.

More likely to have distributed processing with higher level commands to multiple processing and I/O processing units. Like having an I/O processor handling the disk drives and off loading I/O delays to an I/O processor which deals with the file system, formatting, auto-insert detection etc.. rather than tying up the main processing parts of the system.

With higher speed networking already having quite a bit of the processing inside the chip, more of the network stack will exist in the network processing unit, which hands off complete web pages, files to the operating system on the main processor, rather than packets.

This is what a lot of mainframes do now and have done for a while (remembering a Cray system that years ago had a VAX cluster as its front end I/O processor).

For the majority of embedded systems (standalone SMALL boxes not 'PC') I don't see 256 for a long time yet due to the number of interconnects, unless the system RAM and CPU are on one die and smaller bus exist externally (Transputer).

Where 16 bits may well survive is in next generation of things like keyboard and touch screen where less transistors can mean less cost and power. Possibly a keyboard that has a multi-lingual changeable graphic layout and works using 16 bit (or more) scan codes to get away from the ASCII constraint we currently have.

The main constraints on any system are the interconnects and I/O bottlenecks.

--
Paul Carpenter          | paul@pcserviceselectronics.co.uk
    PC Services
              GNU H8 & mailing list info
             For those web sites you hate

You seem to be unclear on what '8 bit micro' means. If they go into

8 pin devices, how does that mean '8-bit micros have faded away' ?

The answer to this question, is the three P's : PinCount PicoJoules Price

In sheer volumes, 8 bit uC will dominate 32 bit ones for years to come. Only if you mapped something like shipped-code-size, would they start to come close.

-jg

8-bit micros will survive as long as they are cheaper to produce than 16-bit micros.

8051 is dominant only because there are 7 different makers. Its a very old technology on the verge of being pushed aside. There are plenty of new 8-bit players waiting to replace the 8051. Atmel AVR, TI 430, and god forbid the 18 series PIC.

[snip] :The place where 4-bit processors are still king is at the very low :power end - watches. They want that tiny battery to run the device :27/7 for many years, and fewer transistors still means lower power. :I don't see that ever changing. : 27/7 ? I think your 4-bit watch needs a new battery :-)

But when we have a huge address space, it gets used in a sparse manner, for virtual files, etc. No normal user wull use 16EB of RAM, for sure. That's not to say they won't use the address space.

"Paul Marciano" wrote: [snip] :The move to 64-bits is mostly due to the 4GB address limit on 32-bit :machines that is a real barrier for large systems like database :servers. 64-bits provides for 16EB of directly addressable memory. : :Assuming RAMs double in capacity every 18 months it will take 48 years :to build systems that hit the 16EB limit. That assumes it's even :possible for the RAMs to continue to evolve that far. : :I think 64-bits will be with us for 100 years if not longer. The trend :now is for parallelism over speed. Disk drives are still the :speed-bump for all high performance computing. : :Not even Microsoft can drive demand more than 16EB. : :Cheers, :Paul.

:Guy Macon wrote: :> Less than 5% of the market compared to 8-bits having 30-40%. :>

: :Thanks for the article pointer, Guy. Very interesting. : :Followup question: : :Assuming the 8051 is the dominant 8-bit microcontroller for NEW :DESIGNS(*), do you think it will continue to be, or is there a new :rising star? : :* If my assumption is wrong s/8051/dominant_mcu/ : :Cheers, :Paul.

First, it depends on what "dominant" means. If you count number of units shipped, Elan/EMC, WinBond and SunPlus are way ahead of

8051, Pic, ARM, etc. If you count number of engineers working on a chip, it's the opposite.

In my opinion, for another chip to displace the 8051, it would have to have:

[1] Good, free tools. (C Compiler, Simulator/degugger, macro assembler, RTOS, etc.) [2] A bunch of variants (lots of analog I/O, USB, low voltage...) [3] good cost/performance. [4] All patents expired.

I think that being able to have an instruction that has two 64-bit source addresses and one 64-bit destination address embedded in it will eventually drive the extreme high end to 256 bit instructions and a 64-bit address space. I also see the current trend found in MMX/SSE driving future processors towards 256 bits; a SIMD extension that deals with horizontal and vertical adjacent pixels in 32-bit color is likely to need more than 128 bits.

Ah. I forgot that some of you are still living on Sol-3/Terra. Out here on Alcyone-9, 27/7 is correct.

:)