The BMAR project aims to provide a detailed reference for the performance and other characteristics of Arduino and Arduino compatible microcontroller boards. Using a range of benchmarks, the performance of each board is measured and stored in a handy-dandy spreadsheet. Fancy finding the Arduino with the lowest power draw? We've got you covered. Need to do lots of floating-point arithmetic locally? There's an Arduino for that, too. Need all the pins? Just take a look at our results.

There are three main benchmark sketches used to generate BMAR results.

The classic floating-point benchmark, here largely unmodified beyond a switch to single-precision in order to not spoil the fun for 8-bit microcontrollers incapable of handling double- or higher-precision arithmetic. This provides a glimpse of a given microcontroller's ability to handle floating-point maths.

Where Whetstone measures floating-point chops, Dhrystone looks at integer arithmetic. This version of Dhrystone has been modified in order to allow it to run on the ATmega328P microcontroller found in the Arduino Uno and compatibles, where it uses over 80 per cent of the dynamic memory (SRAM).

A custom benchmark written specifically for BMAR, IOBench times a number of digitalWrite and digitalRead operations on a board's pins and provides results in kilohertz (KHz) - i.e. how many operations the board can carry out per second on a single pin. If you have access to a frequency counter, IOBench also has a third test mode: at the end of the 'soft' benchmark, IOBench goes into a 'hard' mode in which it physically toggles Pin 7 on and off using digitalWrite in an infinite loop. Hook your frequency counter up here to get the "Digital Write (Hard)" result.

As well as the performance benchmarks, BMAR looks into other aspects of a microcontroller's characteristics. These are:

BMAR tests the power draw of the Arduino boards via the USB port in three modes: Load, Idle, and Sleep. To test for Load draw, the Whetstone benchmark is run and the power draw recorded via an ammeter; Idle results are obtained while the microcontroller is in a delay() cycle; and Sleep results recorded when the microcontroller is in its interruptable low-power sleep mode. Note, however, that this is not always the absolute lowest power possible: for the purposes of BMAR, only sleep modes accessible through the Arduino IDE in a simple sketch are used. For some boards, no sleep modes are available without various workarounds and hacks - these are immediately obvious thanks to the fact that the idle power draw will match the sleep power draw in the results!

Having a load of flash memory available for sketches isn't much use if your sketches are huge. To obtain these results, the BareMinimum.ino and Blink.ino sketches included with the Arduino IDE are compiled and the size of the resulting binaries recorded.

The width, breadth, and height of each board is measured with calipers. Note that the results may differ from the official measurements found on the manufacturer's documentation, as the measurements for BMAR take into account any components sitting proud of the overall PCB.

If you've got a board you'd like to see included in BMAR, feel free to run the tests and send us a pull request to get it added! Ideally, though, avoid boards which are clones of existing entries: the Arduino Duemilanove, for example, is functionally identical to the Arduino Uno (as they both use the same microcontroller), so there's little reason to include both.