Ofcourse they would need to redesign the Apu but didn't MS changed the Xenos and Xenon and made it a single chip. Whats stopping Sony to invest some Money into a redesign later if it saves them alot more money. Whats AMD stopping from doing it ?

Whats the reasoning behind not taking a chance to make money and get more marketshare if a redesign won't influence the older models. Its not supposed to increase performance of a PS4 just make it cheaper its basically still a fixed unchanged hardwaredesign on the outside.

Maybe AMD layed plans out already for that to happen eventually and is part of the deal made with Sony ? Also if AMD wants Apus to go anywhere they need stacking, the GPU parts in an Apu for a PC can't grow due to DDR3 limiting bigger APUs. 

its all part of the plan from the beginning imo.  

as Kasz216 said I am talking about a slim redesign maybe in 5 years time

@Kasz

What makes you say that its unlikely ? Costreduction potential was probably a priority, this time around. PS3 is a disaster in that regard.

Sony wouldn't have used GDDR5 at all if they don't see a way to drastically cut costs later. It draws alot of power and through stacked Ram and a single chip PS4 will be tiny by the end of the gen. Sony are going to switch to stacked as soon as it becomes cheap enough. There is no reason why they wouldn't do it if they can. Saying thats unlikely is hard to understand looking at things. So I assume you know something I don't ? Please tell me what am I missing ? Or is that assumption  based on a feeling ?

Latency and bandwith. So why used I fast ? Because through bandwith you can transport a higher volume of data in a shorter timeframe which is why GDDR5 is considered faster aka More Data available per frame.

Latency is reaction time. Lets take a look at cars.

DDR3 is a lightweight tuned Volkswagen Golf with Schumacher on the wheel its faster from 0-5mph because the driver steps earlier on the pedal. GDDR5  is a Bugatti Veyron with a 500 pound guy behind the wheel. Schumacher will win on very curvy roads. Like a road through the Alps. But the Veyron with the 500 pound driver is still faster in 99% of the (gaming related) cases.

So yes the Veyron wins in  the (gaming related) cases even with a shitty driver. 

Also faster is used because it actually is faster doing a thing. And I mean the technology runs faster. One thing does something 1 Billion times per second and the other does it 5 Billion times.

So yes saying GDDR5 is faster than DDR3 is a perfectly correct way to describe the situation. Sure its sometimes slower but mostly not.

The Latency issue is not a big issue anyway I assume you are hinting at DDR3vsGDDR5 debate coming from nextgen talk ?

Caching in the CPU solves most problems. Throwing way more bandwith (HP) at one problem than necessary also. And yes the Latency issues are big compared MHZ to MHZ and Bit per Bit but are really not that big of a deal with 5 times the clockrate and 2.5 times  more bandwith. Even if the per MHZ latency is 6 times higher its not as bad if you have 5 times more MHZ.

But when I think about it MS reason for their Ram choice makes sense. Kinect might suffer big time on a PS4 the latency would really be noticeable here. As the Kinect needs alot smaller packages fast the GDDR5 larency would add up each time the software needs another package from the Ram. The lag would increase alot. This might be the reason for Esram+DDR3 in a console.  

It'd be an actual chip redesign, it's not just slapping RAM on top of the GPU and calling it good ^_^; I think they'll come up with better cooling solutions down the line and maybe do the stacking on the side to reduce overall size but I don't think they'll tie the RAM to the chip directly.

I don't see the point in changing the analogy that way.  It may accomplish a lot of tasks quicker, but it's still slower.  That's specifically what latency is measuring.  The speed.

Latency is speed,   Bandwith is amount of data able to be carried.  You can change the analogy how you want... but that's what it is in reality.

Even if GDDR5 is faster at accomplishing the particular goal.  It's slower.  It accomplishes things faster by "carrying" things quicker.

So to call it faster isn't really correct.  If you want to say it accomplishes specific tasks quicker... sure.  However it's in fact slower.

It's like saying Brian Shaw is faster then Usain Bolt because he can clear out a garage faster.

He's faster at doing that, but he's not actually faster.  Meaning that such a qualifier should always be put in conversation.   Brian Shaw is insanely fast "At cleaning out garages".  etc.

Call me a literalist i guess.

As for the PS4.  I've got to think the cooling solutions for this will be pretty large... sort of defeating the point of a "slim".  Afterall the main reason people don't use memory stacking right now is because of the heat generated between the layers. 

If you are going to wait that long you might as well just wait a bit longer and get a couple GTX 780s (said to be bassed on the same GK110 core as Titan). I mean the Titans are limited supply so it's not like they are going to get any cheaper even if you can even still get 2 at the end of the year. 

As for stacked Ram. A 14nm PS4 Apu is small and cool.

This is an error most people make. Actually, the situation is exactly opposite of what you think. A 14nm APU might dissipate only 60% of the heat of a 28nm APU (assuming leakage problems can be solved which greatly increase as the process scale decreases) . However the heat density is now much higher, because the die size of the 14nm APU is considerably smaller than the 28nm APU. Cooling becomes more critical the smaller the die surface is, because you have "less surface area to dissipate heat".