I haven't seem slides on it, so it's an assumption, but I think it's a solid bet that discrete GPU's are on a finite timetable. maybe AMD is just missing it completely. Nvidia seems to be evolving GPU computing, AMD seems to be refining gaming chips. Just seems to be a canyon in MO's between them. i guess I assumed a plan on both sides. If the above are the 'plans' for both, AMD seems to be way behind the curve if it is still years from on die GPU.

Hard to know what the f___ AMD are planning IMO. The HD 7000 series arch (GCN) was laid down in 2009 as far as I'm aware -which was while Dirk was still captain. I sincerely doubt that a change in CEO and a BoD can influence a design is intended to keep AMD going in the graphics business for the next 2-3 iterations. I'd assume, like VLIW5 before it, AMD are tied to GCN for at least the HD8000 series if not the HD9000. Like many other facets of AMD's business, I don't think their change in headman and business direction will be readily apparent until the present tech has run it's course.

AMD can't very well back out of the high end (read: workstation/enthusiast) card market in any case- not without cutting the collective throats of dev's that have committed to AMD's OpenCL SDK.

The only points of interest as far as I can discern will be what kind of effort AMD are willing to put into future drivers, gaming development and open source software. Nvidia just posted record revenue and profit numbers from their workstation/hpc graphics- so either that market is now growing or they have taken marketshare away from FirePro's (using Cayman GPU's) while still utilizing an old (Fermi) architecture - I suspect that it is the latter.
Nvidia have been pilloried for their memory controller issues in recent times- a large part of which was due to moving away from the 256-bit memory bus (note that the GTX460/560 don't suffer from memory speed issues) and using 72 bit ECC memory modules (as opposed to 64-bit) in addition to being late to the GDDR5 party. AMD still have to demonstrate both these attributes (or to convince enterprise users that EDC will suffice in lieu of ECC) to deliver in the high ASP/high margin workstation market in addition to providing a suitable software ecosystem and overcoming the inertia inherent in the business sector.
The performance/enthusiast desktop graphics market isn't that big, so I think the prime mover in it's ongoing existance is sharing architecture with enterprise GPGPU -true in Nvidia's case since they have already made a case for Kepler and Maxwell.

TL: DR
I Don't think either company will be exiting the high-end gaming business unless they are willing to surrender the highly lucrative workstation market.
It's a certainty that Nvidia won't walk away from either -not with the profile that Quadro and TWIMTBP enjoy- and the fact that GPGPU factors highly in Nvidia's future plans.
AMD...well, who knows what's going through their minds- aside from some fist pumping/chest beating from Rory and PR drivel from John Fruehe, I haven't actually heard anything from AMD. And what is anticipated are still products of the old regime.

AMD's main problem is they are fighting a war on multiple fronts - against Nvidia in graphics (who have superior cash reserves and R&D), against Intel (who have cash, R&D, and fabs that are purpose built for their product), and a future problem with ARM x64 in the ultra-low voltage sector. AMD have limited resources- something has to give.

Probably depends on if AMD plan on moving into the ultra-portable market. If they plan on getting into smartphones and the like then they are starting from further back than the companys that already have an ARM licence and have to fight smaller more cost efficient chips with SoC's that are burdoned with x86 overhead (i.e. overly complex for what is required). If they plan on desktops/laptops/tablets/enterprise then APU incorporating GCN graphics would serve them just fine and Trinity should stand them in good stead...but an APU isn't a SoC. Moreover, Rory and the Board seem to be putting AMD on the Jenny Craig diet- cutting staff/payroll and divesting yourself of code writers and engineers (amongst the PR monkeys)doesn't -at least to me, sound like they are gearing up to fight on every front. To me it seems like Rory and Co are getting ready to narrow down AMD's portfolio

Well I suppose I can see how this could happen. Thinking I'm in need of another pump, and I'd like to get some 45 and 90 degree fittings. The little things add up to. I'm trying to stick to a budget here! Story of a pc enthusiast gone wrong. lol.

well there you have it...i guess.I am getting the impression The interviews from both sides regarding the time-line of the demise of discrete GPU must not be in play, or that would not even be an option for AMD. My original inquiry was for the time between 5 TF on die in now inside 4-5 years, and now. How is each GPU manufacturer going to deal with the the transition regarding the integration of GP²U. given that, it seems that Nvidia is far ahead of AMD.

I just read your edit:
SoC was the stated goal, I was more focused and wondering (apparently with a faulty premise) more immediately about the demise of the discrete GPU and being on die. It sound like system ram wont be an on die possibility until around the time of the 8nm shrink anyway. like you pointed out, with the gutting of AMD staff, any previous stated pursuits are now moot anyway, in short who knows.

I'm just going by what's been published and extrapolating based on what could eventuate. Dirk Meyer's axing and AMD's seeming change in emphasis tend to show that focus and direction aren't set in stone. I'd wonder if Nvidia's future would remain as it's set out if Jen-Hsun stood down.

AMD certainly have the expertise for forging ahead with iGPU. Llano certainly makes a strong case, as will Trinity I expect. But to my mind these are products destined for entry/mainstream level consumer products -laptop/desktop, probably moving down to tablet.

There seems to be a lot of talk about discrete graphics going the way of the passenger pigeon and the dodo, but I would think that while you still have a graphics AIB enterprise market then AMD and Nvidia will both ensure that the gaming market requires high-end discrete graphics.
I'm not certain that a large percentage of PC gamers would for instance be happy playing BF3 at the resolution or IQ that Llano would offer. Even if you could somehow shoehorn HD 6870 or GTX 560 level graphics power into an APU package ( die area, heat dissipation, TDP), that still leaves a considerable amount of gaming out in the cold - 3D, Eyefinity/Surround, conventional AA, HB and SSAO, particle effects etc.- and that's without taking into account that AMD, and Nvidia especially, would likely keep pushing graphical effects to keep the uptake of new hardware high (Crysis 2 anyone)...I'm kind of thinking that Metro:Last Light and S.T.A.L.K.E.R.2 won't be too friendly with entry level graphics. Add DirectX 11.1 (Win8) into the mix and you have a graphics landscape that moves just fast enough to outpace whatever the low end of market produces.

Even if AMD pulls the plug on Gaming Evolved (and that wouldn't surprise me), I think Nvidia still pushes gaming harder and faster -they pretty much have to- other than graphics they don't have a lot else to offer at the present time. Given that there seem to be some rather graphically intensive game engines in development- Unreal Engine 3 is now supported by DX11 (Nvidia sponsored "Samaritan" demo to it's name) and Unreal Engine 4 is slated for 2014 for example, I wouldn't see Nvidia pushing the development if they weren't going to gain something out of it...especially if AMD end up with their GPU's in the next-gen consoles.

Yeah, and that seems to be where my understanding (or buying into) what amount of power they planned to have on die (and a way to upgrade it for that matter) in the next few years falls of the rails.