The software implementation should be able to decrypt a 1080p30 stream given a suitably fast dual-core processor and about 1.6GB of RAM. The poor performance is due to the nature of the algorithm. HDCP was designed to be cheap and fast for hardware manufacturers, but operations that are quick and easy in hardware are often slow and inefficient in software. In spite of this, the developers believe they have opportunities for further optimization and improvement, making real-time decryption on more modest hardware feasible.
Just wait until someone ports this to FPGA or custom ASIC, bearing in mind that Altera dev boards start at $200. Instant HDCP cracking. What's at stake here is pixel-for-pixel rips of ANY content that plays on your TV.