Oh and I don't believe Raja Koduri will achieve as much as many hope with intel. A few reasons:
1. He appears to be a man that can see an endpoint, but not find appropriate intermediate steps on the way. Rome wasn't built in a day. Vega reaches for the moon in some respects, yet the foundations aren't there to make that reach anything but unstable.
2. Leaving AMD for a sabbatical right when the GPU team was deep in brown stuff won't endear him to anyone below, from the next in the management chain down to the lowest worker bee in the office. They'll all remember "when the going gets tough, Raja ___ks off", you don't want to jump in the trenches under that kind of "leadership". Some posts in AMD subreddit already allude to this.
3. With Intel being chipzilla, the expectations will naturally be huge, indeed, we can say now they'll be wholly unrealistic. There was much hype around what Larabee could result in, yet absolutely nothing came out of it for the consumer. Starting at such a (relatively) low base, with a long hard road ahead of it, the graphics department will have to very carefully dampen down all hype so that the executive board don't lose faith when (not if) products don't meet public expectations. Raja does not appear to be someone who can manage those expectations appropriately - this is possibly a result of point (1) - even he expects more than what actually is delivered.
I predict:
- Surprisingly few will leave AMD to go with Raja to Intel.
- Intel will build 2 or 3 discrete GPUs under this initiative - not sure if all will be released to the public.
- All will be utterly underwhelming compared to what media/public expectations have been allowed to build to (which is not entirely Intel's fault, but they will retrospectively wish they had categorically said performance won't be as good as X, Y or Z).
- The Intel board will (once again) shelve plans to get into high-end graphics
- This will all be done before ~ 2021.