(07-11-2014, 02:25 PM)KHg8m3r Wrote: [ -> ]Its actually closer to option B and what you said in the other post, minus the AVX stuff.
Sounds good to me.
Though that doesnt quite explain the big jump of Sandy to Ivy since the major changes were just the usual branch prediction improvements and a couple (arguably insignificant) ISA extensions.
And of course, moving to 22nm FinFET bulk, but that doesnt exactly inherently increase performance...
Quite confused...
It turned out that the original testing of Ivy vs Sandy had been done badly, and that they actually perform very similarly. The difference can be explained away as due to what is known to be different.
(07-12-2014, 12:29 AM)AnyOldName3 Wrote: [ -> ]It turned out that the original testing of Ivy vs Sandy had been done badly, and that they actually perform very similarly. The difference can be explained away as due to what is known to be different.
And now everything clicks with me just as it should. Thanks
Interesting replies guys, thanks. Say what you want about Intel's business practises but their R&D is still going strong.