Hardware performance features available for embedded software to exploit are reviewed, including both recent advances as well as past trends. Hardware performance depends on clock speed and a large and still growing bag of tricks to discover and exploit concurrency in the problem being solved. The embedded software designer should be aware of the underlying hardware performance features in order to make the best design decisions. In particular, the recent advances in SIMD instruction sets, hardware multi-threading and single chip, multicore processors require explicit measures to exploit them in software.