AVR32 32-bit MCU - Benchmarks

AVR32 AP Outperforms Major 32-bit Architectures

The AVR32 AP core consistently outperforms major high-end 32-bit standard MCU cores in every EEMBC® benchmark for performance and code density.

In EEMBC benchmarks, AVR32 code consistently requires 5% to 20% less code than the ARM® Thumb® instruction set to execute the same functionality. More significantly, for high-performance applications, when optimized for speed of execution, AVR32 code is 30% to 50% more compact. AVR32 performance-optimized code density is also consistently better than that of the ARM ISA in every EEMBC benchmark.

EEMBC code size benchmarks

Figure shows the advantage of the AVR32 in regards to code size with a lower value indicating better performance.

EEMBC code size benchmarks are based on publicly available numbers from EEMBC, based on ARM's simulated numbers for the ARMv5 Thumb ISA. AVR32 code was compiled with the IAR AVR32 2.09 compiler while ARM-code was compiled with ARM's ADS 1.2 compiler.

The AVR32 AP consistently outperforms both ARM9™, ARM11™ and NEC V850E cores in EEMBC's TeleMark™, OAMark™, AutoMark™, ConsumerMark™ and NetMark™ benchmarks. Its performance exceeds that of the ARM11 by over 35% on the ConsumerMark benchmark.

EEMBC performance benchmarks

Figure shows the architectural performance when comparing CPU cores with a higher value indicating a better performance

The EEMBC benchmark numbers are normalized to the same clock frequency. The ARM926EJ (based on the Freescale i.MX21 values using GCC 3.3.2), ARM1136JF (based on the Freescale i.MX31 values using GCC 3.4.3) and NEC V850E (using Green Hills multi 2000 4.0.1 compiler) benchmark data can be found at the EEMBC website.