Arduino Due

Created Sat, 17 Sep 2011 02:46:40 +0000 by Darth Maker

Darth Maker

Sat, 17 Sep 2011 02:46:40 +0000

[url]http://arduino.cc/blog/2011/09/17/breakfast-at-arduino/[/url]

This is going to be interesting. Oddly, though at first glance it would seem to be direct competition for the ChipKIT, I'd be willing to bet that it will be two or three times the cost of a Uno32, and probably more than the Mega32 also.

As far as chip capability, it looks like the SAM3U is more powerful than either of the PIC32s used in the ChipKITs, except that the Mega32 has more memory.

I'm hoping to get a look at one on Sunday at the Maker Faire.

WestfW

Sat, 17 Sep 2011 09:11:41 +0000

I think this is the first confirmation that "The Arduino Team" may be adopting multiple processor (family) support in the Arduino IDE. That's a good thing; it's a lonely job trying to keep a non-mainstream product (like ChipKit) compatible with a moving target like the Arduino libraries and IDE if that target doesn't even support multiple targets. (Said another way, it's easier to support the 3rd core than the 2nd core.)

avenue33

Mon, 19 Sep 2011 14:04:30 +0000

This news is highly interesting because the Arduino and alike boards are jumping from 8 to 32 bits.

Given that the Arduino IDE has to manage two different platforms, AVR-specific code needs to be rewritten, as SoftSerial.

So the benefit for chipKIT users is better support for the gray areas of limited incompatibility between AVR and chipKIT.

jumpin_jack

Mon, 19 Sep 2011 18:36:43 +0000

As far as chip capability, it looks like the SAM3U is more powerful than either of the PIC32s used in the ChipKITs, except that the Mega32 has more memory.

That is interesting. I'd like to compare the speed of the Thumb-2 ARM with the MIPS32 PIC32. I haven't seen CoreMark benchmarks for the SAM3U, but the TI Stellaris benchmarks seem to show that the PIC32 with the MIPS32 core is clock-for-clock faster than the Cortex-M3. The PIC32 at 80 MHz using the GCC 4.5.1 compiler scored 203.68 while the TI Stellaris Cortex-M3 scored at 80 MHz using the Keil ARMCC compiler scored 127.60. What do you guys think? http://www.coremark.org/benchmark/index.php

cobbj

Tue, 20 Sep 2011 16:00:30 +0000

jumpin_jack -

PIC32 scores higher in Coremarks per MHz than any Cortex-M3 product that has been tested. Atmel isn't a member of EEMBC so there isn't as much Coremark work done for their products. Dhrystone MIPS per MHz - the Atmel SAM3U parts run 1.25 DMIPS/MHz, while PIC32 scores 1.56 DMIPS/MHz. 96MHz SAM3U is still lower in DMIPS than PIC32 @ 80 MHz. There are some other peripheral differences between the two parts - ADC speed, # channels, # serial peripherals, etc. If ya'll would like, we can put together and post a comparison of the parts.

jumpin_jack

Wed, 21 Sep 2011 00:20:20 +0000

So you're saying that the 80 MHz PIC32 should out perform the 96 MHz Cortex-M3 at least just for the core, right? That's what I would guess judging by the Coremark scores. Sweet. :D

Darth Maker

Wed, 21 Sep 2011 01:27:36 +0000

cobbj, I would appreciate the comparison. It would be really great if you also included the ST32 parts used by LeafLabs.

I got to examine a Due at the Makerfaire on Sunday. Honestly, unless there was a major software reason to use it, I'd rather use an Uno32 or Maple. One of my main reasons is the form-factor. I really dislike Arduino Mega-sized boards. Anything that big should have motor drivers and three pin headers for its I/Os. But I'm a roboticist, so I would naturally feel that way. The other reason is 5V tolerant pins. From what I can tell, the SAM3U doesn't have any.

I still would love to get one of the community released prototypes in a couple months... And I already own 32-bit "xDuinos" from two companies, so I could do some RL comparisons.

cobbj

Wed, 21 Sep 2011 15:37:30 +0000

jumpin_jack:

"Perform" is a very strong word in the MCU world... according to DMIPS benchmarks and published numbers, yes, a PIC32 w/ MIPS M4K core @ 80 MHz outperforms a 96MHz SAM3U w/ Cortex-M3. Benchmarks try to mimic the real world, it is generally accepted that Coremark does that better than DMIPS, but they still aren't the real world.

In general, yes, your statement is correct - there's just caveats for everything...

Darth Maker:

I'll work on getting those comparisons done - they'll probably be posted by the user marc_mccomb.

hairymnstr

Thu, 22 Sep 2011 10:17:25 +0000

I've not looked in detail at the way the MIPS4K architecture is designed but I think although it is faster in a straight run, the Cortex-M3 has context switching optimised for faster interrupt handling and the like. To be honest there isn't enough difference to matter I don't think, if the PIC32 isn't fast or powerful enough you'll need something more like a Cortex-Ax (BeagleBoard etc.) rather than just a side-step into a Cortex-M3.

I think this announcement is excellent news for the open hardware world and we have Microchip and Digilent to thank for shaking up the market and forcing some additional innovation.

cobbj

Thu, 22 Sep 2011 20:44:53 +0000

Here is the quick peripheral / core comparison:

[attachment=0]Comparison.jpg[/attachment]

WestfW

Fri, 23 Sep 2011 01:51:02 +0000

I don't think the "4 PWM" number is correct for the Due. The chip involved has some special PWM peripheral with 4 outputs, but I think those are separate from the (more primitive) PWM possible based on the timer peripherals...

Darth Maker

Sun, 25 Sep 2011 02:57:29 +0000

Are you going to make another more inclusive list with more processors on it?