three things.
1. Nvidia needs high end core to attach their GPUs to. AMD and Intel moving GPUs to the same die as the CPU is going to freeze them out of much of the market over time.
That doesn't mean that the offering is going to get much of an uptake.
NOTE also that Nvidia is also persuing building A15 designs for cellphone/tablet market. Where these 64 versions are heading is unknown.
If Windows/ARM doesn't take off and Intel's and AMD offerings get much better, then this is a dead-ender.
[ NOTE: Windows NT had multiple ports that didn't go anywhere either. Like Apple sticking it to Intel , similar issue with Microsoft pushing an ARM port. If Intel just becomes a better dance partner they other will get dropped. ]
2. The statement from the ARM CEO came
after the Nvidia announcement. ARM knew full well there were folks running around talking up forks. They aren't going there yet. Nor do I think they are going to feel obligated to buy into the designs banged out by Nvidia. Once the ARM goes 64 bit internal, add in superscaling , and pipelining to keep up with AMD and Intel performance then it isn't so clear there will be a large gap between ARM designs and x86 ones. ARM power efficiencies come in part because of the stuff they leave out by making a power vs. performance tradeoff. Once those are partially reverse it isn't so clear the "big win" is going to be there. Especially, since Intel can just "out process" you. Your fab is at 32 when they are at 28.
If ARM can get a decent hook into the "high energy efficient" server market with just a variant of the A15 they will probably persue that course. If the A15 can't crack the market they they may go down the 64-bit path. I don't think they really want to go that way if they don't have to. Until 3-4GB of RAM gets common in mobile devices ( very far from that at the moment) there isn't lots of pressure to go 64 bits.
3. One reason why Apple has $70B in the bank is that they outsource lots of stuff to people who have lots of customers to spread R&D costs over. Apple buys cheaper and sells with a higher margin. That gap is what built that $70B.
Tweaking the ARM baseline is one thing. Venturing off the baseline will substantially increase R&D costs. Nvidia is going to need a much broader customer base than the number of Macs Apple sells for their 64-bit fork to pay off. Apple somewhat redundantly doing the same work just for the Mac market is a much bigger gamble. Apple isn't a big gambler.
They spin stuff as being big gambles but most thingst they do these days are not.