Fate of PIC32 If Microchip buys Atmel MPU business.

> Ulf Samuelsson wrote:
> 
>>Think NXP needs to rewrite their ads about fastest ARM flash MCU.

It may be in the semantics (just like Politics, and other spin....)

NXP claims :
"At 125 MHz, the NXP LPC2900 series
are the fastest ARM968 microcontrollers
available on the market,"

Atmel's AT91SAM9XE128 says ARM926EJ-S�

So, different cores, [and both companies can claim the
as fastest Flash ARM9xx, if they define XX to sufficent precision! ;) ]

-jg

> I'd certainly like to see the 8-bit AVR core replace the PIC core in
> low-end PICs. =A0>
> - Britt
> - Britt


Really?? You must be kidding. It looks like you never worked with LOW
END PICS; I would suggest you to look into them.

Do not make such malicious comments with out knowing anything about
them.


-dK

On Oct 6, 8:51=A0am, Dan Ash <dan_...@sbcglobal.net> wrote:
> Not that anyone can do anything but guess what might happen, but if
> Microchip buys Atmel's MPU business, what is the prognosis for the
> PIC32. =A0I bought some PIC32 tools and invested time learning the
> family, and now it may go away or new variant design activity may
> stop. =A0This is because Atmel has a popular ARM based family of MPUs
> and the MIPS based PIC32 has only been out a year. =A0Freescale sells
> Coldfire, PowerPC, and ARM based processors, but I am not sure
> Microchip will take on the same burden. =A0I guess we will all see how
> it plays out and whether Atmel will accept their offer.

My 2 cents;
Wrong subject, should rather be "Fate of AVR32..."
If Microchip is the one to lead, there is little chances for the AVR32
to make it to a long life. PIC32 and AVR32 are both rather new and
both play in similar leagues. Main difference, MIPS offers a roadmap
while with AVR32 everything needs to be developed by lots of
engineers. There is a variety of tools available for PIC32 (MIPS)
while hardly any tool suppliers are interested to invest in
development for AVR32 as long as Atmel provides free tools. So, you
are stuck with free tools and free support... Microchip is a company
known for making profit. It is one thing to support an 8-bit tool
chain for free and it is a completely different ball game to do it for
a 32-bit device.
Microchip has internal know-how of the PIC32, why would they give it
up in favor of an architecture that does not offer any significant
benefits over the own device but needs lots of money to be further
developed. AVR32 was announced as competitor to ARM7 and ARM9,
depending which device. That was before there were CortexM3 and
CortexR4, both make the AVR advantages void. They offer similar power,
R4 offers more performance, not even talking about ARM11, A8 and
whatever comes later. ARM will continue to develop the cores because
that is what they do and make their business with. But they develop
for many companies and the R&D money is rather well spent as the
investment serves many customers. Similar is true for MIPS, nothing
like that for AVR32.
IF, and that is a big IF, Microchip is successful in taking over
Atmel, I would not invest a single cent in further developing anything
with an AVR32.

Ulf, your comment about Microchip wanting to sell the ASIC business,
you know as well as I do they were not talking about the SAM7 and SAM9
but about the CAP9 and CAP7 and the like. You also know how the
microcontroller revenue within Atmel is divided, the largest share
going to AVR8 and already the second largest to ARM, not so much less
than the AVR business. You also know that the AVR32 business is off
the chart or better hardly visible on the chart, so do not imply that
the ARM devices might die and AVR32 get further money to do it's own
special thing that is doomed either way, whether taken over from
Microchip or not.

I wish Microchip luck in their attempt to take over, do some clean up
and make good business out of the remaining very good devices.

btw. I am absolutely sure that AVR8 and PIC10,12,14,.... will continue
to coexist. Too much money and most of all too much profit in there.

An Schwob

"-jg" <Jim.Granville@gmail.com> skrev i meddelandet
news:5817c044-6186-4a04-9b39-f7e7b87e6c3e@o40g2000prn.googlegroups.com...
> Ulf Samuelsson wrote:
>
>> "Jim Granville" <no.spam@designtools.maps.co.nz> skrev i meddelandet
>> news:48ea5a43@clear.net.nz...
>> > Dan Ash wrote:
>> >> Not that anyone can do anything but guess what might happen, but if
>> >> Microchip buys Atmel's MPU business, what is the prognosis for the
>> >> PIC32.  I bought some PIC32 tools and invested time learning the
>> >> family, and now it may go away or new variant design activity may
>> >> stop.  This is because Atmel has a popular ARM based family of MPUs
>> >> and the MIPS based PIC32 has only been out a year.  Freescale sells
>> >> Coldfire, PowerPC, and ARM based processors, but I am not sure
>> >> Microchip will take on the same burden.  I guess we will all see how
>> >> it plays out and whether Atmel will accept their offer.
>> >
>> > This is a longer term question. Once silicon is fully released, is it
>> > rare to pull the plug. Unless it is so niche, that they cannot afford
>> > a small run, or so bug ridden, that it is not really released....
>> >
>> > What does happen, is the next generation development goes onto the
>> > 'back
>> > burner', and the device's larger customers are quietly steered onto
>> > something else. Then, the price climbs, to the threshold of pain for
>> > other customers.... :)
>> > All that might take 2-3 years.
>> >
>> > I did note that NXP have just released a 125MHz ARM9. Nice looking
>> > peripherals.
>> >
>> > Atmel also floated an ARM9 FLASH, but quietly said no more,
>> > so looks like that needed a respin.
>> >
>>
>> Yes, there was some issues, which should be fixed by now.
>>
>> You can order AT91SAM9XE-EK kits,
>> At least in early september they were in stock without any significant
>> leadtime.
>> Production volume is coming out before end of the year.
>>
>> 200 MHz ARM926EJ-S with plenty of peripherals.
>> Basically it is a SAM9260 with 128/256/512 kB (128 bit) Flash
>> and 16/32/32 kB SRAM.
>> The core has 16 kB of Instruction Cache and 8 kB of Datacache.
>
> and the FLASH memory Bandwidth is  ??

It has a 128 bit flash memory running at 1,8V.
The 1.8V operation limits the flash to 17 MHz giving 272 MB/s.
If you run the CPU in Thumb Mode, you will be able to top out at 136
native MIPS when not using the cache.

Running in 32 bit ARM mode will give you 68 native MIPS max from flash.
When you run from the cache, or from the ITCM you have the full 200 MIPS
in 32 bit mode.

Most applications will use a mix so maybe 150-160 MIPS as an average should
be OK.


NXP typically uses a 24 MHz flash at 3,3V (power hungry) on their ARM7s.
Assuming the same flash technology is used, we get:
24 MHz * 16 bytes = 384 MB/s.

In order to sustain 125 native MIPS, you need 4 * 125 = 500 MB/s.
In ARM mode, the part will max run 384/4 = 96 native MIPS.
In Thumb mode, the part has bandwidth for 384/2 = 192 MIPS,
but obviously it can only do max 125 Thumb MIPS.


>
>> Will run the internal bus at 100 MHz and can access SDRAM at 100 MHz.
>
> So can the peripherals Clock at 100MHz, or 200MHz ?
> (eg timers, SPI ...)

The Bus clock is 100 MHz, and all peripherals derive their speed from
this clock, so max 50 Mbps on the SPI.

>
>> Think NXP needs to rewrite their ads about fastest ARM flash MCU.
>
> Perhaps their peripherals are faster ;)

Doubt that, without knowing anything, I would guess the 968 bus
would run at 125/2 = 62,5 MHz.


>
> -jg


-- 
Best Regards,
Ulf Samuelsson
ulf@a-t-m-e-l.com
This message is intended to be my own personal view and it
may or may not be shared by my employer Atmel Nordic AB

"Jim Granville" <no.spam@designtools.maps.co.nz> skrev i meddelandet
news:48ef20cc$1@clear.net.nz...
>> Ulf Samuelsson wrote:
>>
>>>Think NXP needs to rewrite their ads about fastest ARM flash MCU.
>
> It may be in the semantics (just like Politics, and other spin....)
>
> NXP claims :
> "At 125 MHz, the NXP LPC2900 series
> are the fastest ARM968 microcontrollers
> available on the market,"
>
> Atmel's AT91SAM9XE128 says ARM926EJ-ST
>
> So, different cores, [and both companies can claim the
> as fastest Flash ARM9xx, if they define XX to sufficent precision! ;) ]
>
> -jg
>


In 1983, I was in Japan, and enjoyed the view from the worlds highest Tokyo
Tower :-)

-- 
Best Regards,
Ulf Samuelsson
ulf@a-t-m-e-l.com
This message is intended to be my own personal view and it
may or may not be shared by my employer Atmel Nordic AB

Ulf Samuelsson wrote:

>>Perhaps their peripherals are faster ;)
> 
> 
> Doubt that, without knowing anything, I would guess the 968 bus
> would run at 125/2 = 62,5 MHz.

The NXP data is rather 'new', but it does say this
in the user manual

ARM968E-S processor running at frequencies of up to 125 MHz maximum.
Multi-layer AHB system bus at 125 MHz with four separate layers.

I could not find a specific Timer MAX relevent to the 125MHz parts,
but on the older parts I did find Timer_Max = CPU_max.

Another 'annoyance' with first generation 32 bit uC, is the restriction
that the CPU clock cannot be slower (power saving) than the peripheral 
clock.

ie Designers might want 10ns edge precision, but only need 5 MIPS
to keep up with that.

I think that is slowly improving, but such details often are
missed from the data sheets...

-jg

> "An Schwob in the USA" <schwobus@aol.com> skrev i meddelandet 
> news:570f96a9-4a71-4d3d-9ef1-
> 06ea1efd47dd@r38g2000prr.googlegroups.com...
> On Oct 6, 8:51 am, Dan Ash <dan_...@sbcglobal.net> wrote:

> Microchip has internal know-how of the PIC32, why would they give it
> up in favor of an architecture that does not offer any significant
> benefits over the own device but needs lots of money to be further
> developed.

There are advantages of having a proprietary core,
as well as there are advantages of having a "std" core.
So after any takeover, there will be two "std" cores, MIPS and ARM,
and one proprietary core, AVR32.

I see people asking for ARM, I do not see anyone asking for MIPS
(actually I now remember one).

> Ulf, your comment about Microchip wanting to sell the ASIC business,
> you know as well as I do they were not talking about the SAM7 and SAM9
> but about the CAP9 and CAP7 and the like.

Same building, same office, same cores, same peripherals.
Even though they are formally two groups, the interaction
is so high, that I doubt that there is *any* value in an ASIC division
which does not have access to AT91 resources.

> You also know how the microcontroller revenue within Atmel is divided,
> the largest share going to AVR8 and already the second largest to ARM,
> not so much less than the AVR business.
> You also know that the AVR32 business is off the chart or better hardly
> visible on the chart, so do not imply that the ARM devices might die and
> AVR32 get further money to do it's own
> special thing that is doomed either way, whether taken over from
> Microchip or not.

That would make several Fortune 500 companies annoyed.


An Schwob



-- 
-- 
Best Regards,
Ulf Samuelsson
ulf@a-t-m-e-l.com
This message is intended to be my own personal view and it
may or may not be shared by my employer Atmel Nordic AB

I don't have any good inside info on this but IMHO the pic32 and avr32 will
both survive for some time.  I see the move as a good one by MCHP and they
will be in control in the merged venture.  Customers will want both, so
PIC32 will have to survive even though it is less mature than AVR32.

With that said, AVR32 might be the main engine that is pushed going
forward.  Does anyone have data on relative performance of these two
processors?  I think that this might be the difference if AVR32 is equal
and the tools are better.   If not, PIC32 has common peripherals which is a
big boost for those that migrate.

Renesas manages them all still... and finally convergence after many
years.

Comments?



>Not that anyone can do anything but guess what might happen, but if
>Microchip buys Atmel's MPU business, what is the prognosis for the
>PIC32.  I bought some PIC32 tools and invested time learning the
>family, and now it may go away or new variant design activity may
>stop.  This is because Atmel has a popular ARM based family of MPUs
>and the MIPS based PIC32 has only been out a year.  Freescale sells
>Coldfire, PowerPC, and ARM based processors, but I am not sure
>Microchip will take on the same burden.  I guess we will all see how
>it plays out and whether Atmel will accept their offer.
>

On Oct 10, 9:31=A0am, dk <mcdivyaka...@gmail.com> wrote:
> > I'd certainly like to see the 8-bit AVR core replace the PIC core in
> > low-end PICs. =A0>
> > - Britt
> > - Britt
>
> Really?? You must be kidding. It looks like you never worked with LOW
> END PICS; I would suggest you to look into them.
>
> Do not make such malicious comments with out knowing anything about
> them.
>
> -dK

I have worked quite a bit with PIC 12C509's and PIC12F629's, writing
PIC assembly code software for a patented product that has shipped in
a few 100k units.  So my dislike of low-end PIC "architecture" is
based on experience.

- Britt

On Oct 28, 9:41=A0pm, britt.snodgr...@gmail.com wrote:
> On Oct 10, 9:31=A0am, dk <mcdivyaka...@gmail.com> wrote:
>
> > > I'd certainly like to see the 8-bit AVR core replace the PIC core in
> > > low-end PICs. =A0>
> > > - Britt
> > > - Britt
>
> > Really?? You must be kidding. It looks like you never worked with LOW
> > END PICS; I would suggest you to look into them.
>
> > Do not make such malicious comments with out knowing anything about
> > them.
>
> > -dK
>
> I have worked quite a bit with PIC 12C509's and PIC12F629's, writing
> PIC assembly code software for a patented product that has shipped in
> a few 100k units. =A0So my dislike of low-end PIC "architecture" is
> based on experience.
>
> - Britt

I have worked on all : 10, 12, 16 and 18 series PIC micro-controllers
and the product line includes: banking solutions, industrial
automation, home appliances, automotive etc.

BTW,  your experience just with 12 series, says it all.

-dK