Executive summary: The M14K and M14Kc have the right performance and cost for today’s MCU markets, and the new microMIPS instruction set cuts code size by 35%.  Compared to competing ARM cores, the M14K and M14Kc offer significant advantages in cost, power, and performance.  However, MIPS must overcome ARM’s big head start in 32-bit MCUs.

Review: The new M14K and M14Kc cores are strong entries into the 32-bit MCU market.  MIPS is positioning these cores against the ARM Coretex-M3 and ARM926EJ-S, respectively.  Compared to these ARM architectures, the MIPS cores offer important advantages in performance, cost, and power efficiency.   The new MIPS cores also close the gap with ARM on code density thanks to the new microMIPS ISA.   This ISA uses a mix of 16- and 32-bit instructions.  Compared to the 32-bit MIPS32 ISA used in previous cores, microMIPS offers a 35% reduction in code size while achieving nearly identical performance. 

The following tables show the M14K specs and the Cortex-M3 specs.  All figures are for a 90nm process, and do not include area or power for cache.  Comparing the two cores is a bit tricky, but the MIPS41K is the unambiguous winner on performance and die area.  (Die area translates directly to cost.)  The performance of the area-optimized M14K is particularly noteworthy.  This 0.21 mm² core cranks out 290 DMIPS, about four times faster than a similarly-sized Cortex-M3.

The comparison gets murkier when it comes to power.  The M14K turns in a respectable showing, and it is a better choice if you need the 190 MHz of performance.  However, ARM is the better choice if you can make do with 50 MHz of performance.

Now let’s look at the M14Kc specs and the ARM926EJ-S specs.  Again, all figures are for bare cores (i.e., no cache) in a 90nm process.  Here the tables are turned: MIPS is clearly ahead in area and power, but the winner of the performance comparison is unclear.  ARM has the lead in maximum performance, but MIPS can boast better performance for a given die area.  (Note that the ARM performance numbers are MIPS, not DMIPS, so the comparison is far from perfect.)

Finally, it’s worth taking a look at code size.  The microMIPS ISA is highly similar to the Thumb-2 ISA used in the ARM Cortex family.  It is reasonable to assume microMIPS and Thumb-2 will achieve similar levels of code density.

While the new MIPS cores are highly competitive, they are getting something of a late start.  You can already buy MCUs based on the ARM926EJ-S and the Cortex-M3 from a number of vendors.  Depending on whose numbers you believe, these chips have helped ARM capture as much as 25% of the MCU market.  In contrast, M14K and M14Kc parts won’t be out until 2011.

Having said that, MIPS has made inroads into the MCU market with its existing cores, the M4K and M4Kc.  Most notably, the M4K powers the Microchip PIC32.  Microchip says it picked the M4K because it offered the best combination of power, performance, interrupt capabilities, code density, and potential future customizations. (See Inside DSP for details.)  These attributes have only improved with the M14K and M14Kc, so I am optimistic that these cores will attract additional design wins.