Which means several years of development ahead to have working silicon, and that would mean AMD64 v1, which Windows and many libraries/application in Linux doesn’t support anymore.
In Debian Unstable, for example, ld-linux-x86-64.so.2 reports that it only supports v2, v3 and v4. v3 architecture , so CPUs from Buldozer/Nehalen generation or later. That version of the architecture will still be protected for a few more years.
Since both Intel and AMD are competitors on both CPU and GPU markets, Nvidia’s only option is Zhaoxin, a joint venture between Via Technologies (who has a license for box X86 and AMD64) and Shanghai municipality.
Failing that, they would have to go with ARM and emulation, which would come with a performance penalty, or separate CPU and GPU chips, which would make the devices bigger and less power efficient than competing models with APUs.
In conclusion, don’t hold your breath. This talk about Nvidia handheld PCs is just to appease their shareholders and create FUD on AMD and Intel ones.
Even if it was unrestricted, x86 is incredibly difficult to optimize well. Most of the people who know how to do it already work at Intel or AMD. Actually, they might all work at AMD.
They definitely don’t work at VIA, which is the forgotten third company that makes x86 chips. Forgotten for a reason.
Now, Nvidia has a big pile of cash and can solve the problem that way. More likely, though, they’ll use ARM like they have been.
The first batch of x86_64 patents elapse next year.
Which means several years of development ahead to have working silicon, and that would mean AMD64 v1, which Windows and many libraries/application in Linux doesn’t support anymore.
In Debian Unstable, for example, ld-linux-x86-64.so.2 reports that it only supports v2, v3 and v4. v3 architecture , so CPUs from Buldozer/Nehalen generation or later. That version of the architecture will still be protected for a few more years.
Since both Intel and AMD are competitors on both CPU and GPU markets, Nvidia’s only option is Zhaoxin, a joint venture between Via Technologies (who has a license for box X86 and AMD64) and Shanghai municipality.
Failing that, they would have to go with ARM and emulation, which would come with a performance penalty, or separate CPU and GPU chips, which would make the devices bigger and less power efficient than competing models with APUs.
In conclusion, don’t hold your breath. This talk about Nvidia handheld PCs is just to appease their shareholders and create FUD on AMD and Intel ones.
And the other extensions can be emulated. The result should be much faster than full x86_64 emulation on ARM.
Even if it was unrestricted, x86 is incredibly difficult to optimize well. Most of the people who know how to do it already work at Intel or AMD. Actually, they might all work at AMD.
They definitely don’t work at VIA, which is the forgotten third company that makes x86 chips. Forgotten for a reason.
Now, Nvidia has a big pile of cash and can solve the problem that way. More likely, though, they’ll use ARM like they have been.