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

Every gate will consume power, so you want as few of them as possible. Also, in the 70s feature size was enormous to what we can get nowadays, so more gates were more expensive. (today cost is often determined by pin count). I don't think you could build a microcontroller with fewer transistors than a loop counter solution. Just think of what you'd need: ALU, ROM, memory interface, address generator (think branches)...

Reply to
steven
Loading thread data ...

In my experience it like this- people learn a micro at collage or where ever. They know this micro and that makes it easy for them to use. They use it in every application, regardless of whether it is suitable or not. I you want to see this in action ask some one what their favorite micro is and tell them it is crap. They immediately launch into a full on emotional assault on you. I have worked in Asia and have seen the 8088/8051 used in almost everything. The most amazing one is in a controller application such as a telephone system or PABX, it is clearly not suited to this. But for the above reason they continue to use it. The development platform is also cheap. Smart micro houses provide cheap development kits and software.

Reply to
kerlin88

Guy Mac> Not really. We have reached the point where the die size

Depends on the process. To make inexpensive chips, you have to use an old process, not the latest. In other words, in early 2005 inexpensive chips are made in 0.5 micron or larger, not 130 nm or 90 mn. There are several reasons for this:

1) Mask costs for smaller processes are *much* higher. This isn't likely to change much in the future, so 180 nm and smaller mask sets are likely to always be very expensive (hundreds of thousands of dollars). Even though you can amortize the mask cost over a very large number of units, it still ends up contributing a significant amount to the per-chip cost. 2) Wafer starts in older processes are cheaper. This does shift over time, so when 65 nm becomes common, 180 nm will probably be a lot cheaper than it is today.

In a 0.5 micron process, you're a lot less likely to be pad limited, so 4-bit and 8-bit processors with similar amounts of memory and the same pin count are likely to have different die sizes.

Reply to
Eric Smith

If you're referring to the size of the integer registers, I don't see much justification for 128-bit and 256-bit processors. I think the changes to high-end processor architecture between now and 2060 are much more likely to increase parallelism (in various forms) rather than simply make registers and ALUs wider.

For floating point, 128-bit registers are probably reasonable, but 256-bit is questionable. Some current 64-bit processors have 128-bit floating point registers.

Certainly the use of 128-bit and 256-bit hardware data busses is worthwhile for high-end systems, and that's already done today on 32-bit and

64-bit processors.
Reply to
Eric Smith

Actually, it seems like never, since he didn't say that.

Reply to
Eric Smith

I just thought of a LARGE volume embedded DSP market: Cell phones (there's also mp3 players such as the ipod, but that's minor compared to cellphones). Do any of them use some other microcontroller in addition to the DSP (perhaps for a very low power mode, though the DSP's I've seen can operate over quite a range of MIPS and currents)?

-----

formatting link

Reply to
Ben Bradley

It's interesting to compare those numbers with these just quoted by Philips.

formatting link

This has the 8 & 16 bit shares pretty much reversed (& more credible) than the figures above !?

These inlude forecasts for 2009, and clearly, 8 bits is not going away, but neither is the 16 bit sector going to collapse. These figures are also by revenue; do them by volume, and the 8 bit will continue to be way out in front.

-jg

Reply to
Jim Granville

You generally cannot prove a negative statement.

Reply to
Bryan Hackney

About 1972 I built a PABX around the 8008. I would have killed for the 8088 or 8051. When I left due to a disagreement with the President the 8080 was about to come out, and I was planning a system using two of them. It would have been more than adequate. One was to handle communications, and the other was the executive. And that was a full featured PABX, with all the frills, such as call waiting, forwarding, busy waiting, whatnot.

Please include adequate context in your posts. Look around for information on how to handle the insipid google usenet interface.

--
"When I want to make a man look like an idiot, I quote him."
                                               -- Mark Twain
Reply to
CBFalconer

Are you quibbling over the exact wording? Because he did say, many years ago, something about 640K being enough for anybody.

Bill

Reply to
William Meyer

Computer manufacturers have known this for quite a few decades. Many had educational institution prices for the hardware and the system was often delivered loaded with all kinds of software, which for a commercial customer would have costed a lot. By creating a pool of graduates familiar with their system, this also helped in selling similar systems to commercial customers, when there was a pool of trained professionals when otherwise it was hard to find professionals. In a few years, these people might also be selecting the hardware.

Later on, the microprocessor vendors eagerly donated Intel Intellecs and Motorola Exorcisers with some local media visibility. Other processor vendors that did not donate development systems, had later on very hard to get their processors into products designed by the graduated students, even if the technical merits were better.

This is really not a new thing, since all vendors try to get a good visibility in educational institutions, thus some institutions have rules what donations they can accept and independently decide what systems to buy, in order to avoid any unfair competition advantages.

In a growing economy with limited resources, it is tempting to get all the donations that you can get.

Paul

Reply to
Paul Keinanen

I take the bit-ness of a CPU in terms of addressing - bus sizes and such are irrelevant as 256-bit wide buses and 128-bit SIMD registers are already in widespread use today and there is no reason to believe we won't end up with over 1024-bit wide (internal) buses in the near future.

I agree with the argument 16-bit is being squeezed by both 8 and

32-bits, but they won't disappear from niches - just like there will always remain niches for 4 and 8-bits for cost, power or pin reasons.

Definitely.

Sounds very unlikely. The demand of 64-bit is driven by the need to address large virtual memories, have over 4GB of RAM and the affordability of it. 64-bits is just about enough to use as a single virtual address space for most desktop PCs in the world (250 million PCs each using 100GB of harddisc space), so a 64-bit virtual addresses will be sufficient for desktop systems this century.

Having 2^64 bytes of RAM is going to be unlikely unless there are are several major breakthroughs in the number of atoms needed to store a bit. If current DRAM densities grew by a factor of 2 every 2 years (it is extremely optimistic to assume this rate is sustainable indefinitely), it would take until 2070 before you could have a DIMM with 2^64 bytes of memory.

I don't believe any of these will ever happen. 96-bit addressing might be adopted but even that is exceedingly unlikely - getting anywhere near that amount of RAM would imply single bit electron or photon based storage. More likely is that the concept of "addressing" will change radically before we reach the 64-bit limit.

Wilco

Reply to
Wilco Dijkstra

About 70% use ARM cores with one or two DSPs on the same chip. Some high-end phones integrate two ARM cores and two DSPs.

Stephen

-- Stephen Pelc, snipped-for-privacy@INVALID.mpeltd.demon.co.uk MicroProcessor Engineering Ltd - More Real, Less Time

133 Hill Lane, Southampton SO15 5AF, England tel: +44 (0)23 8063 1441, fax: +44 (0)23 8033 9691 web:
formatting link
- free VFX Forth downloads
Reply to
Stephen Pelc

AFAIK, cellular phones all have a "general-purpose" CPU core. According articles I've read, ARM is the most popular CPU core for cellular phones.

--
Grant Edwards                   grante             Yow!  Isn't this my STOP?!
                                  at               
 Click to see the full signature
Reply to
Grant Edwards

formatting link

Reply to
Guy Macon

"He didn't say that" is informal shorthand for "there isn't a shred of evidence that he ever said that, the first "quotes" are from long after he was supposed to have said that, he denies ever saying that, many people have seached and searched to try to find the source of the "quote" and failed, and the 640K limit was not a microsoft decision but rather an Intel/IBM decision."

Reply to
Guy Macon

Evidence, please.

Reply to
Guy Macon

As Bryan says, you can't prove a negative. So, the nay-sayers don't have to do anything: they have the logic on their side. The aye-sayers OTOH may be able to prove they're right, but not necessarily so: the problem can be undicidable. Sitting comfortable on their ass the nay-sayers will demand definite proof, in the form of a written and signed document or a voice recording, and when such a document is found, they'll ask more proof: prove that it's really BG's signature, prove that it's his voice. Bottom line, neither one can prove he said it or didn't.

Time for another approach: why was video memory placed at the top of a 1MB memory map, with a RAM limit of 640K? It was all the 8088 could do. So that was what Microsoft had to work with in the first place. At that time nobody (not IBM nor BG) seemed to see the wall ahead. I've always found that IBM made a very poor choice at that time by choosing for the 8-bit 8088 with it's 64K segments which has bugged developers for many years. IMO the 68000 was a far better uP. It had a flat memory map of

16 MB, a set of general-purpose 32-bit registers and a neater instruction set.

Steven

Reply to
steven

There are probably many reasons why IBM chose to use the 8088 instead of the 68000. I always felt that their decision was poor. At the time the first IBM PC's came out I was running a BBC-B micro with 5.25" floppies. Shortly after I aquired an Atari-ST. It was only a couple of years later that I swapped the ST for the Atari-Mega-ST to which I added a laser printer. I have never managed to find software that performed like that I used on the Mega-ST in the PC world (I was running quite a full featured DTP system which produced all my documentation, I also ran GemForth on that and was able to communicated between the Atari's and BBC Micro. My first PC was actually a Zenith laptop (because I was after an easily portable solution at the time) and used MPE's PowerForth and cross-compiler tools to develop systems for the nuclear industry.

Systems I was developing in that period of time used the 6809/6309 processors. I did have a 68000 to play with but never needed the power in the applications I was dealing with at the time.

I expect that IBM felt that the price of the 8088 was a better deal than what Motorola were offering the 68000 for at the time. I guess the 68000 was quite new at the time they were developing the design. By the time the PC came about I had already been developing systems with 6800/6809 for a while and had dealt with some mini-computer systems (GA SPC12, SPC16 and Prime family). Perhaps there may be others who know what was in the minds of the PC developers at the time.

--
********************************************************************
Paul E. Bennett ....................
 Click to see the full signature
Reply to
Paul E. Bennett

Ah, the 6809! Most beautiful CISC uP ever! and OS-9 was far better than DOS too. I remember a column in BYTE, around '86 IIRC, where the columnist (Pournelle?) compared DOS to OS-9 as Sylvester Stallone compared to Robert De Niro. I've always liked that one. Pity that Microware just let the 8-bit OS-9 die. Why didn't they make it public domain, like TurboPower, who donated their prize-winning Delphi collections to SourceForge?

BTW, if an 8088 is called a 16-bit controller, then the 6809 is one too! I recall Motorola had a 6829 paged MMU which extended the memory map to 2MB, had 4 task mappings and could be cascaded to 8 devices, for 32 tasks. A real beauty, at least on paper, because I've never seen a design withe the

6809/6829.

An aside: does any1 know if the complete Radio Shack Color Computer has been recreated in an FPGA?

Steven

Reply to
steven

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.