Just an odd thought, but wouldn't it be possible to just build in a die shrunk version of an old CPU along with the other cores, and just use that to run any weird legacy code? I mean a Ryzen 7000 series chip has 6.5 billion transistors on the chiplet die; in comparison an Athlon XP Thoroughbred chip had 38 million transistors, so including a die shrunk version would only cost you like half a percent of the die space to get perfect backward compatibility.
What comes next may not really have a designator. Look at what is happening with AI accelerators GPUs and with CUDA and Intel's one API. In those cases you access compute resources on the other side of an API. The actual hardware doesn't need an access layer of an immutable ISA. Basically I'm arguing that most of the compute will be done on weird bespoke accelerators. In that future what does traditional ISA wars even mean? Microarchitecture and I guess systems macroarchitecture will matter all.
x420
hell yeah 420 HH my dude
>HH
paizuri
snapdragon
x87
https://en.wikipedia.org/wiki/X87
xfree86
There might be a general push to just use ARM with everything, but the idea of all of the hardwire being soldered in is not a personal favorite
x86_64
XXX
RISC-V
will take 10 years according to jim keller
Nah, just 2 more weeks
Oh good
Intel will fund a submarine rescue to bring back Itanic
there is no vehicle capable of reaching those depths
Itanium
wow i bet his crotch is even sweatier!
Nothing, we're stuck with x86 forever.
I think they're planning on x86_64, with all of the legacy shit removed.
However ARM is where its at.
homie the x86_64 is what we use today. What you mean is x86s
RISC-69
I HAVE FOUR WORDS
>64-Bit Only: x86S processors exclusively operate in 64-bit mode, eliminating legacy 16-bit and 32-bit support.
>Virtualization: Legacy features are emulated through virtualization technologies rather than direct hardware support.
>Benefits: 64-bit-only CPUs offer improved efficiency, faster boot times, and potentially reduced security vulnerabilities.
>Industry Impact: While not imminent, x86S represents a forward-looking approach that may simplify processor architectures and enhance performance.
I wonder how well games will perform under that virtualization.
Doesn't matter good goy, who cares if your favorite game can't be played anymore? Time to play some Fortnite and EA Sports instead!
Just an odd thought, but wouldn't it be possible to just build in a die shrunk version of an old CPU along with the other cores, and just use that to run any weird legacy code? I mean a Ryzen 7000 series chip has 6.5 billion transistors on the chiplet die; in comparison an Athlon XP Thoroughbred chip had 38 million transistors, so including a die shrunk version would only cost you like half a percent of the die space to get perfect backward compatibility.
They can do that with just microcode. AMD has been emulating 16-bit instructions in microcode since the first Opteron.
x88/14
Nothing
x88
y86
RISC-V
AMD128
What comes next may not really have a designator. Look at what is happening with AI accelerators GPUs and with CUDA and Intel's one API. In those cases you access compute resources on the other side of an API. The actual hardware doesn't need an access layer of an immutable ISA. Basically I'm arguing that most of the compute will be done on weird bespoke accelerators. In that future what does traditional ISA wars even mean? Microarchitecture and I guess systems macroarchitecture will matter all.
Sex86
128 bit
x86 works
just keeping it
x86 pro max
POWER11, or z/Architecture microcomputers. The future belongs to IBM.
Arm86_64