Some of you may be interested in this paper that caused quite a bit of discussion when it was published. (I'm linking to the IEEE version but there's also a publication at ISCA 2011.)
Dark Silicon and the End of Multi-Core Scaling, Esmaeilzadeh et al. ISCA 2011.
What is the significance of the "not floating point math" caveat?
@ak47: Floating point math takes even more energy to load memory and everything else for.
Edit: Just noticed that there are some more details about this on a later slide.
Some of you may be interested in this paper that caused quite a bit of discussion when it was published. (I'm linking to the IEEE version but there's also a publication at ISCA 2011.)
Dark Silicon and the End of Multi-Core Scaling, Esmaeilzadeh et al. ISCA 2011.
What is the significance of the "not floating point math" caveat?
@ak47: Floating point math takes even more energy to load memory and everything else for.
Edit: Just noticed that there are some more details about this on a later slide.