If they only had a lot of flexible programmable vector units sitting somewhere close by... Mmmh
. Once you balance most of the vector code between GPU compute and your AVX units, you are still left with a lot easier to tackle general purpose cores than the PPE/PPX cores ever were.
I remember your posts about the Mario GPU design or the custom eDRAM with shared CPU and GPU access, there is plenty to talk about its architecture and it would not be the first architecture with lots of manpower spent in ingenious ways that reaches a target that is not what the market or the company itself wanted to back (I still think we saw a fraction of what the IA64 architecture could provide and we never had it compete on the same manufacturing node as its internal Core competition not the same level of investment once the ex EV7/8 team was handed the reigns).
Nintendo, who I am more and more convinced has nowhere near the grip on their supply chain than company like Apple or even Sony and Microsoft have (considering how Wii U was barely making a profit at launch), targeted something (and in a way that could have been a lot more third party friendly) which the third party community did not really want and that limited the reach of its USP's too. It is not just having a machine which could not run circles around the 8 year older competition at the kind of launch price it launched at, but it is difficult to argue that better specs would have allowed the Wii U OS designers to deliver a much faster Susi and user flow through the console, keep MiiVerse always loaded or find a way to avoid long time consuming game-to-MiiVerse-to-game trips.
PS4 and Xbox One were also not coming out just a year before the next generation of consoles was supposed to arrive... which makes them a lot better positioned for developers to target and optimise code for. Overall they are a far larger jump from PS3 and Xbox 360 than Wii U is.