If Nvida's GPU could be used as a stand-alone processor complete with booting its OS, sans any other CPU in the computer, then that Nvidia GPU would no longer be referred to as a GPU as it would then be the CPU for the computer in question.
I thought Nvidia did port Linux to their Fermi arch? :awe: It just didn't run well... :hmm:
I think APU is the better term, rather than gCPU or just CPU. For me, APU refers to a heterogeneous-core cluster. Current APUs aren't
really good examples of APUs, because really they're like CPUs+GPUs. I think (hope) future APUs will have more varied configurations than the ones we see today.
For example, imagine an APU that integrates two strong x86 cores, a strong vector cluster, one or two weaker (but very low power) ARM cores, and a core that is able to natively run .Net or java bytecode. I could imagine a few uses of that...
Or imagine an APU that is designed purely for office work, and so the ratio of integer-to-FPU is even higher than BD's, with a built-in GPU that can (in a pinch) process floating workloads.
Or imagine an APU that is designed for folding :biggrin:
AMD's vision (heh) as far as I could tell was that the era of just pounding away at general-purpose computing power was over. We have enough of that, we can use the rest of the transistor budget for dedicated circuitry that can do things much, much faster. Honestly, I think Quiksync speaks to this more than anything I've seen from AMD yet, though I'm hopeful.
Edit:
And as far as AMD and marketing are concerned, they've done much worse than trying to differentiate their offerings from Intel. "Unlocked and loaded" (I realize they haven't started that one... yet) or "Ready, Willing, and Stable?" Now THAT was low...
That would've been like if last year Honda placed an advertisement in a magazine of one of their cars just stopped at a stop sign. Maybe I should go into marketing...