LordOfChaos
Member
If that's the case something went terribly wrong with Sammy's die shrink. A smaller chip with the same taping/capabilities should never run hotter than the identical chip on a less efficient last gen (larger/less efficient) process. Sammy certainly charged more per chip based on what would be almost guaranteed to be a cooler, more efficient design
I don't follow. The only "last gen" process to compare it to is the A8, not the A9, and then there's different clock speeds and memory controllers and everything to throw into the mix that make them not directly comparable process to process. There's no A8 on 14/16nm, there's no A9 on 22nm. Not sure what you're comparing.
If you mean Samsung 14nm vs TSMC 16nm, Samsungs and TSMCs processes are on the same generation, and advertised sizes aren't really comparable between fabs as everyone advertises and counts different ways. Look at the difference between everyone elses 14nm vs Intels, plus TSMC 16nm vs Sammy 14. TSMC 16nm is pretty comparable to Samsung 14nm, Samsung is definitely not a generation ahead. And as some peoples results show, TSMCs advertised 16nm is actually better than Sammy 14.
It'd be like a phat PS3 running cooler than the die shrunk slim models that replaced them. It all adds up to tells me somebody made a fairly significant design or layout/configuration mistake. But I'm not smart enough about the current fabrication procedures to know what they actually did specifically
It would be more like comparing the PS3 to PS4, as the chips aren't the same between process shrinks. There's no straight shrink of A8 to compare to here.
Anywho, I hope Ars or Anandtech or someone grabs models of each and runs throttling tests.