AMD has publicly demonstrated a Barcelona system with a pair of 200W R600 cards running in Crossfire mode which is capable of hitting one teraflop. This achievement represents a ten-fold performance increase over today’s high-performance server platforms, which deliver approximately 100 billion calculations per second. Researchers at Stanford are already writing supercomputing applications for R600 and that R600 uses 320 multiply-accumulate (MAC) units which could imply 40 vec4 per GPU and ~800MHz clock. The reasons for R600 delay cited by Henri Richard: “R600 is doing very well and there’s a reason we’re going to launch it when we’re going to launch it…I’ll take the blame.