Or they could try to buy VIA, which is outside of US jurisdiction, if they really wanted x86. Most likely scenario is they find some other way to get the CPUs they want.
I don't think x86 compatibility is the issue here. They can already run Linux on their homegrown Longsoon chips which is often the base OS for their supercomputers anyway. It is not outside the realm of possibility that they could just run a bunch of Teslas or Voltas in Longsoon nodes, though that would requires some hackery. Or they could run them on Opteron or POWER nodes. ORNL's Summit will be POWER9 + Volta.
Unlikely. Anything advanced enough inside AMD is likely considered restricted.
It also means that any other US based technology company is likely in the same spot as Intel. AKA, no they can't just switch to Power and Nvidia GPUs.
What was considered restricted before this particular ban went into effect? GPUs from AMD, for example? I mean you have to deal with the obvious headaches of GPGPU when using them, but GCN cards are easily fast enough to be useful in a supercomputer, especially if you get some of the 1/2 64-bit cards.
If what you say is true, then it means that the US gub'ment was asleep at the wheel for a long time wrt Phi. It took them this long to get around to classifying it in the same category as high-performance GPUs?
So why are we so unsettled about the proliferation of an Intel chip, to a country that already manufacturers so many other parts for the world-wide industry?
There are obvious answers to this, but to go back to what I said earlier in the thread . . . the US has been selling out sensitive tech to the Chicoms for decades. Anyone remember Loral or Global Crossing? I don't think that stuff ever stopped.