Why we don’t have 128 bit OS or CPUs?
In: 927
Modern x86 CPUs already have some memory and operations which work on a 128 bit base (SSE and AVX extensions). It just does not make much sense or bring big advantages to use them as a general base.
64 bit addresses are enough for current memory and 128 bit numbers are not used so often (and for the cases where you would use them you have these extensions).
And 128 bit CPUs would most likely be more complex and need larger chip sizes (and therefore would be more expensive and need more energy). And doing that for little to no advantage makes no much sense.
No real need for them yet. Sixty four bit numbers can hold really huge numbers. And there aren’t many tasks that need bigger numbers. So while eight bit, sixteen bit and thirty two bit CPUs were limited and needed multiple instructions to do things, the current CPUs are capable for most things you want to do. And we have specialized GPU hardware for the things that a general purpose CPU isn’t good at.
A better question would be, why *would* we have 128 CPUs?
32 bit architecture gives us 2^32 values, a bit over 4 billion, which more or less sets the limit for memory addresses, capping your RAM at 4 GB. This was fine for a while but just wasn’t enough eventually.
64 bit is way more common now, which lets us use native values of up to 2^64, or 18 quintillion. RAM is capped by factors other than the number of memory addresses. 64-bit is great for things like IP addresses; 18 quintillion is really plenty for the foreseeable future.
128 bit would let us use native values of up to 2^128, or 10^39. We have no use for numbers this big. You could assign a unique ID to every star in the visible universe and only use 0.000000000000001% of the values available to you.
You’re much better off designing a 64-bit processor that’s half the size or draws half the power than designing a 128-bit processor.
I’m 5 when it comes to tech, so will explain as I know it: basically it’s possible and exists, there just isn’t a practical reason to go beyond 64 bit for most computing needs—the ability to store and use the amount of unique values a 128-bit system provides isn’t necessary for home use, gaming use, or most office use. And since 128 bit would require more power and resources, it’s not practical to just throw-in if it isn’t needed
there will never be a 128 bit os because there is no need.
you may find it shocking, but we currently dont even have true 64 bit cpus, only about 40 some.
the bit number is how many addresses a computer has to use, which basically means “what is the max amount of ram.” with 32 bit, that is 2^(32) bytes or about 4GB. So we had to upgrade, but 64 bit could hold 2^64 bytes, or 18 Exabytes (1 exabyte is 1000 petabytes (1 peta byte is 1000 terabytes (1 terabyte is 1000 gigabytes))), according to estimates made by https://what-if.xkcd.com/63/ this is how much data is in all google servers combined. but in ram. We have already pretty much hit the max of what 1 cpu can benefit from. even servers cant really use more than 1TB of ram, and the way we make computers faster now is by clustering (taking smaller computers and making them work together) so even if you had a 18 exabyte dataset, no machine woukd ever need to load it all in ram.
additionaly, each bit you add is a physical wire you need to add, so since you dont need more than 1tb of ram, and 1tb is 2^40, there is not even a need to make a truely 64 bit cpu.
since the os expects a 64 bit cpu, the cpu just fakes it and pads its 40 bit addresses with 24 0 bits
Because we don’t need it yet.
Imagine you want to give a number to all flats in an apartment building. If there are less than 10 flats, you can number them 0-9, which is 1 digit.
If your apartment building is large, you may need 2 digits. You’ll get 00-99.
If it’s a skyscraper, you’ll probably need 3 digits: 000-999.
But, assume you need to give a number to all houses in the US. How many digits will you need? We have about 350 million people, we definitely have less than a billion houses. You can use 9 digits, so 0-999,999,999. As you can see, you don’t need more than 9 digits.
It’s similar with the 64 and 128 bit namespaces. 64 bits could theoretically manage 18 quintillion bytes of RAM, which is about 18,000 Terabytes if my math is correct.
Wikipedia article > https://en.m.wikipedia.org/wiki/128-bit_computing
I actually have a related question. Why did all of the jumps happen by doubling bits? Like, why did we go from 32 to 64, with no stops along the way?
There is a fixed cost to more bits that you have to pay even if they are useless. It’s a very, very rare circumstance where there will be any value, & easier ways to fit those needs.
It’s like putting a second car on the top of your car, it’s a lot more money, more weight, more gas & more things to go wrong. The only advantage is if you flip your car over you can keep on driving, but there are better and cheaper solutions to that problem.
Because the need to address over 18 exabytes (18 million terabytes) of memory doesn’t come up very frequently.
Similar reason why we had 2 digit dates until Y2K approached. 4 digits took up twice the space and we were decades away from it mattering.
128-bit CPUs would need twice the size for many things but we’re decades (maybe?) away from hitting the current 64-bit limits. So it’d just be a waste.
Latest Answers