Is a 2 MegaByte Flash part, still a Microcontroller ?

jmg

I see Atmel claim to have a 2 MByte (!) M4 sampling

The Atmel SAM4SD32 is sampling now with general availability in September 2012. The device is available in QFP100 packaging priced at $7.94 in 1000-piece quantities.

This comes even down to a TQFP64 package, with 160KB of SRAM.
That is certainly not your Grandfather's 8748, but it is still one chip, and costs less than the 8748 did (I am told by historians)

Does this make the Ardunio Due obsolete, before it even ships ?

rod1963

Oh you mean that Atmel, the company that's notorious for announcing chips that aren't available for years. Remember the XMega? I'd never develop a board based on their micro's unless it's already available from say Digikey.

The Cortex M4 series seems to be real hot at the moment. Every IC house and their pet monkey has a variant out there. NXP, ST, TI, FreeScale, Atmel, etc. One company has already beat Arduino people to the punch - GHI electronics who has put 2 STM32F4 boards for under $30.00. They also have competition from the Maple people. Then there are Expresso and Mbed ARM alternatives.

IMO the Arduino Due is irrelevant of the other offerings out there, not to mention they kept their customer base in the dark, getting info from them was like milking a boar and waiting too long - it should now be called the "OverDue". And they are just now getting out eval boards - which translates into not being released until end of year at earliest. Banzai really bit off more than he could chew not to mention lacking basic communication skills.

They also face competition from the PIC32 people as well. You have the Olimex DuinoMite(a Maximite clone that does low end video) and Digilent boards like the Uno32.

LoopyByteloose

I suppose it is, but it seems rather silly. Micro-controllers generally use memory for buffers and pass searches and sorting on to faster, larger micro-computers. A lot of the AVRs already have more memory that one would ever need, unless you want slow independent processing.

rod1963

Just look at how Prop users are always trying to find ways of getting more memory for it. Not everybody is happy with 512 longs. There are apps that need more.

And slow independent processing. What exactly do you mean by that? The M4 series will smoke a Prop in terms of sheer processing power.

LoopyByteloose

How would I know? I'd have to order 1000 to get that great price, sort through all the documentation, likely be frustrated by a few leading edge glitches, and maybe have to write an IDE for it.

All beyond me. The point was that micro-controllers versus micro-computers don't need a vast OS and the depth of storage for compiler libraries and so forth. I am happy with the chips I work with and an not likely to ever be an expert.

Can you really go in one jump from 512 32bit longs to 2Mbytes. What the heck is M4?

Smoke? Have had plenty of that already, and mirrors too.

Leon

It's the ARM Cortex-M4 which has DSP functions and floating-point:

http://www.arm.com/products/processors/cortex-m/cortex-m4-processor.php

ST has this STM32F4 Discovery board which uses it for under $16:

http://www.st.com/internet/evalboard/product/252419.jsp

Mark_T

I suppose the question is does one consider a DSP to be a microcontroller (albeit a powerful one), or a different class of chip. Perhaps "DSP/microcontroller" ought to be the description if it serves both functions. Maybe "DSP/microcontroller/flash-memory" is more revealing.

Ultimately "microcontroller" is a very wide category and one can always dispute where the boundary between a MCU and MPU is. Perhaps in reality its what the manufacturer's datasheet says it is!

Leon

The M4 is still a micro-controller. It simply has additional DSP instructions and hardware. It's similar to the dsPIC digital signal controllers made by Microchip, but offers much more performance. Microchip makes the PIC24 range of controllers which are basically dsPICs without the DSP functionality.

rod1963

A chip like the STMF4 IMO blurs the boundary between microprocessor and micro-controller. It's processing power surpasses that of mid 90's workstations, yet it can work as a single chip embedded solution. It also has two other components that were once IC's unto themselves - floating point and DSP units.

Integration is blurring a lot of lines.

Leon

It's mostly going to be used in embedded applications, which makes it a micro-controller. In the old days, we used micro-processors like the Z80 as embedded micro-controllers, as well as CPUs in computers.

Invent-O-Doc

I agree with Leon on this one.

Mike Green

We did use Z80s as embedded controllers, but these were not the original Z80 chips. They included things like clock generators, interrupt controllers, small amounts of memory, and some simple peripherals (like parallel I/O), sometimes as matched support chips. We also used versions of the 8080 (like the 80186 and 80188) as embedded controllers with the same sort of mix of simple peripherals. The original microprocessor chips required too much in the way of external logic to be effective as embedded controllers.

LoopyByteloose

Small is beautiful. Dr. Dobb's Journal provided the aesthetic with 'running light on overbyte'.

Floating point may be an imperative for scientific and financial display and data acquistion, but I am more and more wary of the need for it in most embedded setting.

Besides, the educational setting needs to get students started with limited devices before they hand them seemingly limitless devices. I can't seem to figure out what I need 32bits for or more than 8 relays - certainly not to conserve energy. The fundamental concepts and the beauty of economy get lost in vast speed and huge capacity. The bugs tend to have more and more places to hide.

Of course, the military application always want to have an edge by domination via brute force of greater capacity. I am just a lowly consumer.

Leon is some kind of visionary. We will just have to wait and see if he is right.

Leon

I used the Intel 8255 for parallel I/O and 8251 for serial I/O, with the Z80. I never used any of the integrated Z80s, as the 68000 had come out, and there were several smaller devices like the Z8.

kwinn

In the early days the distinction was very simple. If a chip had an address and data bus for addressing external memory and peripherals it was considered a microprocessor. If it had memory and peripherals on chip and used the pins as I/O ports it was a microcontroller. Nowadays the issue is more complicated. I have to agree with Leon to a point, it is more of an issue of how the chip is used than what hardware it includes.

The distinction between a microprocessor and a microcontroller is fading. There is no doubt that a chip like the propeller is a microcontroller, although it could be used as the cpu for a microprocessor. At the other end of the spectrum the Intel and AMD chips are definitely microprocessor chips even though they are sometimes embedded in equipment.

The floating point issue is somewhat irrelevant as far as the controller/processor issue is concerned. Early instrumentation with embedded computing varied widely in design and included designs built around the 74181 ALU chips, calculator chips, micros such as 4004, 8008, 8080, Z80, 65xx, 68xx, and 1802. Many of these instruments performed floating point operations in software, but later used cpu's that included floating point hardware. Having floating point available through hardware or software made no difference as far as the chip being a microcontroller or microprocessor.

IMHO with the available chips spanning the entire range from address/data pins with no on chip memory/peripherals to chips with memory/peripherals on board and no external address/data bus the distinction has become irrelevant.