The article mentions getting SPEC CPU running but doesn't share performance results or scalability results (now the CPU can decode twice as many instructions, etc). Can someone who has been following the research in this area share some results?