"Based on K8L architecture, Altair has upgraded Instruction fetch to 32B, enhanced Branch Prediction and Out-of-order load execution, with Double Precision FLOPS/cycle, 128Bit SSE, and Load per Cycle. AMD claim that the new architect would get a great improvement with more than 40%!"link