In terms of Vega, 'go talk to people who sound like they know more about this than you do' is not an argument. That's an admission of ignorance.
Not really - maybe you should go onto those threads,since there are people who work in the area,etc who have some very interesting discussions over there and maybe you can read the discussion instead of calling people ignorant and that is again you saying your knowledge trumps everybody else via a different way,like calling anyone who does not see your way as naive.
You can't can't even back up any of you assertions either so show ME evidence that AMD won't redirect more of its R and D resources towards GPU development after Zen launches. I ask for your evidence then.
Only in your world does a brand new CPU design which needs more validation for errors and performance need as much money as iterative improvements - yet the whole engineering world contradicts you.
Hardware enthusiasts complain that Intel is not pushing large improvements at every generation since they don't get that iterative improvements are a safer bet - Intel has only had five major ground up new uarchs in nearly 20 years and two of those do not even target what we are interested in. Pentium PRO,P4,Core,Atom and Itanium. The rest have been gradual improvements over time.
Yet,P4 was a failure,Itanium was a failure and Atom is close to being a financial failure.
That explains why Intel is sticking with the "if its not broke don't fix it mentality"whilst enthusiasts start moaning at them.
Risk=cost.
AMD has had a few ground new arches since 2000,the original Athlon,Brazos and Bulldozer. The Athlon 64 has its lineage in the original Athlon. The Phenom had its lineage in the Athlon 64. Bulldozer was a failure and nearly destroyed them.
Risk=cost.
New uarch=more risk.
Iterative improvements=less risk=less cost.
This is why both AMD and Nvidia did the same with Pascal and Polaris.
The last ground up fully new GPU designs from AMD and Nvidia are GCN and Fermi. Everything since has been derivatives(although Maxwell is somewhat of a halfway house with some major changes especially with Tile-based Rasterization).
Edit!!
Do you even realise the Athlon 64 was a derivative of the first Athlon??
The follow up K9 was cancelled and the Phenom series still has its roots in the Athlon 64 and Athlon. Cost was the reason.
So if developing new uarchs required as much r and d spend as iterative developments we should see AMD and Intel pushing them much more often. History does not agree with you.
Plus new uarchs need the software to catch-up with them too - so ultimately that is the other side of the equation. It makes very little sense to keep pushing new uarchs all the time if you have a solid one already.
No,I know why people want to push that AMD HAS to spend as much money on Zen+ as with developing Zen and so on,then they can spin the line AMD won't spend more developing new uarchs on the graphics side and say AMD is doomed for eternity with graphics.
You know what we can agree to disagree.
