Email: Password: Remember Me | Create Account (Free)

Back to Subject List

Old thread has been locked -- no new posts accepted in this thread
Per Westermark
02/22/11 02:24
Read: 1206 times
Sweden


 
#181236 - ARM core already tiny enough that you gain no more
Responding to: Oliver Sedlacek's previous message
Oliver Sedlacek said:
Device cost is mostly proportional to silicon area, and the ARM devices probably do well because they use a comparatively advanced technology which results in a small die area. The NRE costs of a die shrink on a mature 8 bitter will probably never be recouped, and will therefore never happen.

The die size isn't really affected by the size of the core, i.e. 8-bit or 32-bit.

A ARM7EJ-S core in 0.13u is 0.45mm2 large. But the actual wafer size needs to be much larger, since the chip needs peripherials and most importantly it needs bonding pads and huge I/O transistors. The bonding pads and I/O transistors takes up a very significant percentage of the wafer size for a 0.13u microcontroller. That means that a number of thousand transistors more or less in the core doesn't actually matter. And you can't shrink the core any further and get any gain from getting smaller die sizes.

That is why 32-bit ARM processors are so tough competitors to 16-bit and 8-bit processors.

But the other important view, is that a generally low-tech core like the 8051 can be produced with very low-tech production equipment. So it is cheap to build a fab to make them. Because of improvements in technology, even smaller companies can afford fab capacity - or even afford building own fabs. When the 8051 got introduced, the fab cost was high. But today, universities can easily afford own production equipment that manages to build 8051 chips - and software to design new variants. That is a huge reason why it is hard to make real profits from the low-end 8051 chips. NXP can't produce them cheaper than anyone else. And countries with lower salaries do have great advantages on their side.

So while the ARM chips have a denser core in a more advanced technology, that technology doesn't give them a real gain in production cost. Instead, the manufacturers needs to kill of a lot of 8051 competition and use the more high-tech production technology to add in more peripherial functions. At 0.13 u, an extra UART and a matrix multiplexer to route the signals takes way less die space than what it would do for a 8051 designed in "classic" technologies.

So what we end up with from the major manufacturers are 8051 chips that do use new technology and draws hardly any current at all, or runs at very high speeds. Or cheap ARM chips with large amounts of peripherial extras.

List of 65 messages in thread
TopicAuthorDate
NXP suggests 32-bit ARM Cortex-M0 family for 8-bit replaceme        Jan Waclawek      02/21/11 04:03      
   Funny indeed!      Andy Neil      02/21/11 05:15      
      Rest of NXP's 8051 line to follow..?      Valentin Angelovski      02/21/11 05:37      
         comments      Erik Malund      02/22/11 08:23      
            I've been watching them for 20 years now ...      Richard Erlacher      02/22/11 08:40      
               So, what to do?      David Good      02/22/11 09:58      
                  Biting the ARM bullet      Andy Neil      02/22/11 10:27      
                  SST89E58      Jan Waclawek      02/22/11 11:15      
                  Well, if I had to do something ...      Richard Erlacher      02/22/11 21:32      
                     Linux?      Per Westermark      02/22/11 22:34      
                        Just a thought ...      Richard Erlacher      02/24/11 07:15      
                     Leaping to Linux would be ludicrous!      Andy Neil      02/23/11 00:17      
                        Unless...      Andy Neil      02/23/11 01:07      
                           The target wouldn't necessarily be the host      Richard Erlacher      02/24/11 07:26      
                              I don't ever build on target hardware unless target is a PC      Per Westermark      02/24/11 07:54      
                        Supplement - not replace      Andy Neil      02/23/11 01:21      
                           I can't disagree      Richard Erlacher      02/24/11 07:29      
                              Competition always needed      Per Westermark      02/24/11 08:05      
                              not everyone wants the 805x to survive      Andy Neil      02/24/11 08:29      
                                 newer '51 releases      Erik Malund      02/24/11 08:44      
                        It depends on your ultimate goal ...      Richard Erlacher      02/24/11 07:18      
                     Real cheap ARM eval boards      Oliver Sedlacek      02/23/11 02:19      
                        ADuC ARM      Jan Waclawek      02/23/11 02:23      
                           Nearly, ADuC702X      Oliver Sedlacek      02/23/11 04:11      
                        Yes! Lots of Real cheap ARM eval boards!!      Andy Neil      02/23/11 02:36      
                           Why go cheap ...      Christoph Franck      02/23/11 03:19      
                              Prototyping can interfere with extras        Per Westermark      02/23/11 03:30      
                                 "nfity" != "useful" or "helpful" (necessarily)      Andy Neil      02/23/11 03:59      
                                 That's often a problem with EvK's      Richard Erlacher      02/24/11 07:36      
                                    50/50 Agree/Disagree      Andy Neil      02/24/11 08:21      
                                       Perhaps you're right about the second point      Richard Erlacher      02/25/11 02:05      
      Cortex-M0s      Christoph Franck      02/21/11 06:00      
         "low end"      Andy Neil      02/21/11 06:18      
            How low is "low" ?      Andy Neil      02/21/11 06:22      
            money      Jan Waclawek      02/21/11 06:30      
               Depends on view      Per Westermark      02/21/11 07:07      
                  the small embedded view      Jan Waclawek      02/21/11 07:38      
                     yes      Per Westermark      02/21/11 08:22      
                        applications of low pin count      Jan Waclawek      02/21/11 08:35      
                           Either help with real-time or with wire count/length      Per Westermark      02/21/11 08:47      
               Money and technology      Oliver Sedlacek      02/22/11 01:23      
                  ARM core already tiny enough that you gain no more      Per Westermark      02/22/11 02:24      
                     Fab costs      Oliver Sedlacek      02/22/11 03:58      
                        Old fabs or old fab equipment      Per Westermark      02/22/11 04:57      
               Other Meanings      Andrew Ayre      02/24/11 03:14      
   Colonial English      Andy Neil      02/21/11 14:32      
      No new models      Per Westermark      02/21/11 15:07      
         End of the roadmap      Andy Neil      02/21/11 15:49      
   Anachronism      Andy Neil      02/21/11 15:50      
      Quite common to extend meaning of old terms      Per Westermark      02/21/11 16:35      
   a bit related      Erik Malund      02/24/11 09:23      
      Doesn't add any advantage so totally cornered      Per Westermark      02/24/11 09:34      
         16-bitters      Jan Waclawek      02/24/11 11:11      
            I think you missed the point      Per Westermark      02/24/11 12:28      
   NXP 8051s      Jim Granville      03/08/11 20:33      
      you are a bit slow      Erik Malund      03/09/11 06:23      
         you are a bit slow      Andy Neil      03/09/11 09:14      
            lots of power needed to swing 5V devices      Per Westermark      03/09/11 11:06      
               not just the swing      Erik Malund      03/09/11 11:58      
                  The area myth gets busted      Jim Granville      03/09/11 13:28      
                     Long time since chips started to get different scaling      Per Westermark      03/09/11 13:48      
         Wide Vcc is growing trend      Jim Granville      03/09/11 12:10      
            no such ceiling, just no avoidance      Erik Malund      03/09/11 12:49      
      3V3 or 5V      Per Westermark      03/09/11 06:41      
         Oxide thickness      Jim Granville      03/09/11 13:39      

Back to Subject List