The Xbox 360 CPU is very similar already to a 3 core Atom - in order, simultaneous multithreading, etc. It can run 6 threads already, so by now, devs are *very* well experienced with multithreading. The number of cores shouldnt be an issue.
With 8 cores, a clock speed bump, a modern process and instruction set (like AVX) - that could be a pretty huge leap, even if at face value it doesnt sound as powerful as ivy bridge. Its going to be capable of exactly what they need it to do, and nothing more. That kind of architecture worked out very well for them on the 360, and either way, the GPU is the far more important piece of the puzzle.
I agree. Lets throw down this scenario:
MS goes with Intel / Atom processors for a 2013 console against Sony's PS4. The PS4, which is rumoured to have 4 AMD cores (once thought to be Steamroller but now maybe Jaguar) will have AVX (2 128bit ala BD?) on 28nm. If MS went with Atom they could (a) hit 22nm and (b) Tri-gate on Intel's mature 22nm technology. This chip will be smaller-per-transistor than what AMD can offer but, likely, cost as much. The big advantage comes from TDP and frequency as well as Intel's progressive roadmap. When you look and (c) Out of Order Atom (OOOe), (d) HyperThreading (HT), and (e) AVX these Atom cores should be more robust than the current PPEs in Xenon/Cell and with AVX and 8 cores offer a large bump in peak theoretical performance. One advantage of using simpler cores is more budget can be shifted to memory heirarchies (cache) which with many-core setups is a performance killer.
While Atom cores may be a big step down, core to core, from a IVB, SB, and even Core Duo, what you lose in IPC you gain in TDP, size, and number and leaving more budget for cache. Having 8 SIMD units so developers can vectorize their code offers the potential for long term performance gains. Yes, a 4 core IVB is going to tear an 8 core current gen Atom to shreads but if each new Atom has the same AVX resources as a current IVB core there are situations where this theoretical Atom could offer big gains (a number of developers have noted this happens on consoles already with VMX/SPEs where games with a lot of code that can be vectorized and hit peak performance hang well... of course normal code sucks as they are far below 50% IPC in most code).
From Intel's perspective getting AVX in the consoles is a win and they can cut a deal on Atom chips as it doesn't harm higher end margins. In fact it may give Atom some strong inroads that could help Intel's Atom ecosystem. If Sony is going Jaguar and MS is looking at Jaguar cores or some Power variant it would be worth Intel's time to toss out Atom availability. It may suck up some volume that could be diverted to higher margin products but it also gives a lower end model some much needed wind in its proverbial sails without harming the higher margin product line.
From a gamer perspective Intel's higher end stuff is off limits--why would Intel take huge margin cuts? This makes no sense. In terms of Jaguar and Steamroller an OOOe Atom with AVX on 22nm Trigate could be competitive. On the IBM front Power7 is large/power hungry and 478, A2, etc may be lateral moves from Atom/Jaguar at best and ARM would be a big step back in IPC and tools like compilers.
It probably wouldn't hurt MS to know they have a firmer road map from 22nm to 16nm and probably 12nm whereas with TSMC 20nm may, or may not, roll out in 2014 and FINFET has shifted down to 16nm (if it ever comes) and that won't be until, what, 2017/2018 for first volume production at best? Sure, Intel would charge more per mm^2 but you can pack more into each mm^2 and the TDP/Performance advantages when TDP, and not so much chip size, are the bigger barriers makes it sound like an appealing option IF MS could get Intel on favorable terms.
