Just look at how they used SPECfp, the most exploitable benchmark, in almost all of their comparisons in the video.
I am kind of suspicious of this, especially if Sun was using
its in-house Solaris compiler. Sun's 171.art crack is well
known in the industry and it inflates its SPARC SPECfp2k
scores by 15% to 20%. It is a machine independent code
transformation so it will apply equally to Opteron although
its integrated memory controller means the benefit will be
somewhat less.
Oh well, it will be easy enough to tell when Sun's offical
submission(s) are posted on the SPEC web site.
Market Data 
Markets 
AI
