I think Sony or Microsoft can put a middle range AMD "Sea Islands" GPU to be released next year... Radeon HD 8850 (Oland) can be close to what you saw in these demos and the price is not that high like HD 7970 or GTX 680... so a PS4 below $600 is possible.

There is the Radeon HD 8770 (Orust) to be released too... even more cheaper... less than $150 final retail price (for Sony or Microsoft will be way less).

For me a PS4 with a good CPU plus HD 8770 and 8GB of RAM will be priced the same than Wii U (<$400).

PS. The Wii U have a dated hardware... so if it didn't have the Pad the retail price could ended less than $250.

I don't really know much about this tech stuff, surely if the PS4/720 were to release in 2014 rather than 2013, the cost of this GPU would drop down slightly. I mean I do want the new consoles to come soon, but if its at the cost of performance it might be better to wait another year.

I'm almost 100% positive that both companies will do exactly that, which is why I'm just as sure that we won't see consoles at this level of hardware or anything near it. With the success of the Wii and the Kinect, I don't see either Sony or Microsoft having the appetite for the raw power paradigm. I expect Kinect 2.0 and Move 2.0 to be built in, and I expect Sony to try that weird glasses nonsense that was rumored recently as well.

Nor do I expect the next round of AMD GPUs to be a huge jump over the current ones. Not when AMD has been getting it's shit rocked financially and has gone through several rounds of layoffs.

By the HD 8000 series specs leaked you can see a 30-40% boost over the HD 7000 series... not like the previous jumps (50-70%) but this jump is good... really good.

So you can have a HD 8000 mid-rage with the same performance than a HD 7000 high-end.

^ I'll believe that when A) we see actual benchmarks that prove as much--both for those cards and the cheaper 87xx cards--and B) when it's confirmed that the next gen consoles will get 8000 chips that match 87xx cards in performance. Making that kind of prediction requires too many assumptions for me to be comfortable with it; I need more concrete proof from AMD itself.