When will we have 128 bit operating systems?

1000 years from now

probably when we'll quit using bits or RAM, well never with our actual computer architectures

BTW, what's the operating system in the context of the question?

Do you have lots of numbers wider than 64 bits to process?
Do you have more than 18,446,744,073,709,552,000 bytes of memory?

Other than moar RAM, I don't think increasing bits has helped with computational performance. 128-bit is useless.

>I don't think increasing bits has helped with computational performance.
There's a reason we're not using quad core versions of the 6502 clocked at 5GHz.
We SHOULD be using 64 core versions of the 6502, extended out to 64 bits, clocked at 5GHz, however.

Who cares about bits when every storage and program is atleast 1 gb?

>There's a reason we're not using quad core versions of the 6502 clocked at 5GHz.
Because 2^8 address space is not a lot, no?

2^16 for plain 6502, 2^20 for some variants, 2^24 for the 65816.

Why can't machine learning remove all the bloat in these programs??

Hopefully we will get a more elegant form of addressing before that is needed. 64 bit pointers are mostly waste of bits already.

Read the risc-v 128 bit spec. Soon (tm).

I see. Yes, I was wrong, but it's still not a lot.

Much sooner than you think. RISC-V, for example, already implies 128-bit registers and 128-bit addresation.
Hint: one might want to directly address not only some place in RAM, but also some place on a flash storage, directly connected to the system bus.

Assuming the upcoming 128-bit cpus are just 2 64-bit CPUs glued together, probably not for a long while.

When we start converting entire galaxies into RAM.

You severely underestimate how much memory 64 bit is able to address.

I imagine a breakthrough in distributed computing could warrant it, let's say 5 years from now, Average of ~20GB of RAM per phone, averaged over 1 billion phones 20 billion billion bytes of RAM, or 2e19 pooled RAM available for compute tasks.
20,000,000,000,000,000,000 bytes, slightly more than 18,446,744,073,709,552,000 bytes.
Massively parallel distributed compute could change everything, but no-one knows how.
Fuck, RAM over IPV6 could be more effective.

Tell this to Google. They were the main force behind all that 128-bit hustle in RISC-V architecture development.

>Google
Something tells me it's not "Google", it's one of their AI's, maybe nenet, getting pissy about having to bank swap exabytes of servers just to run compute tasks.