Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

can anyone explain to me, why we need an 128-Bit OS anytime soon? I thought the problem with 32bit was mostly about memory address space and 64bit solved that issue for good.


Just going off the deep end, what if you ran an entire server farm as a single OS instance? It's certainly an interesting model for a distributed system, but I don't know how practical it is.

Even then, I don't see anyone getting 3.16 * 10^29 gigs of ram, so maybe it's just more marketing?


64 bits can only address about 4GB on about 4 billion computers.

If you were to hypothesize a globally distributed object oriented system where objects can migrate / replicate from one machine to another, a 64-bit virtual memory space would be insufficient.


Because processor technologies aren't designed the weekend before they're shipped. AMD's x86-64 technology (aka 64-bit) was publicly announced in 1999 (see: http://en.wikipedia.org/wiki/64-bit). It likely was in planning stages years earlier.

When was the first time you needed x64? (i.e. when did you run into the 4GB ram wall?) This year, maybe. Last year at earliest. That's more than a 10 year gap from when a technology was conceived and when it was needed.


Well, according to Wikipedia, 32 bit architectures started to come around in the 60s. We reached the 32-bit memory wall in some applications at earliest 1990. Thats 30 years to go between the first 32-bit archs and the "topping out of them." Then, you need to remember that this is an exponential change in address space, not a linear one. 2^64 is MUCH MUCH bigger than 2^32, you can't just think of it as "twice as much." In the 60s, they thought 4 billion bytes would be much more than they would ever need, but 32 bit's headroom was just 1000 times more than the biggest they used in that day (assuming 4MB, even though IBM's System/360 could access 8MB, which in that case the headroom would only be 500 times as much memory). 64 bit tops out at about 17.2 billion GB, and if we estimate that today's common RAM size is 2GB, the headroom we have for 64 bit is about 8.6 billion times what most computers have commonly today. That means that the headroom for 64 bit is 8.6 MILLION times the headroom that was in 32-bit. For some reason I think that it will take us much longer to fill up 64-bit than it did for us to use up 32-bit.


this is an exponential change in address space, not a linear one

To be fair, Moore's Law describes an exponential change in usable address space that we're still living through.

The numbers are kind of spooky. The stupidest statement of Moore's Law is "everything doubles in size, at constant cost, every 18 months." If we take that statement literally, and also assume that the reaching of the "32-bit memory wall" became a widespread phenomenon around, say, 2009 [1], a backwards extrapolation suggests that the RAM industry should have begun in 1961 with a single byte. This is frighteningly close to agreeing with actual history. [2]

The implication is that another 32 bits only buys us another 48 years. After that we'll need a 128-bit address space in order to properly index our personal catalogues of everything ever. [3]

Or, perhaps, at some point we will just get tired of making RAM.

---

[1] This is when I first noticed it. Yeah, people with millions of dollars hit this wall in 1990. Must have been nice to be them.

You will note a certain degree of slight-of-hand in my numerology here. ;)

[2] Intel's first memory product was a 64-bit unit in 1969. In the fantasy modpunk universe where Moore's Law was operating during the Sixties, that would imply a starting point of a single bit in 1960.

If we reject my narcissistic assertion that nobody important cared about the 32-bit memory wall until this year, and instead accept oomkiller's assertion that the 32-bit memory wall first appeared on the radar around 1990, that would push back the birthday of the first bit to around 1941. I give you Wikipedia:

The [Zuse] Z3 (1941) was the first working machine featuring binary arithmetic, including floating point arithmetic and a measure of programmability. In 1998 the Z3 was proved to be Turing complete, therefore being the world's first operational computer.

Cue the Twilight Zone theme.

[3] We probably won't need that 256-bit address space in 100 years, since there are estimated to be only 2^266 atoms in the observable universe.


Exponential growth is unsustainable. In a universe where the infinite does not exist (either big or small), we are strictly bounded by things. At some point Moore's law will show it's true colors (a logistic curve) and taper off.


Huh? Plenty of things needed more than 4 GB RAM in the 1990s. Scientific and database applications, to name a couple examples. That's why IBM, Sun, HP, SGI, DEC, etc. were all shipping 64-bit machines then. There were also a lot of contortions implemented to make >4 GB happen on 32-bit Intel machines too.

What AMD did (and Intel subsequently copied) with x86-64 was to significantly drop the price of 64-bit capable hardware, to the point where it shipped on personal workstations which didn't really need it then.


Very true: One thing you've missed is that current x86-64 is not yet providing a true 64 bit address space. Only 44 or 48 bits are provided in shipping hardware. This will no doubt be expanded in later revisions of the hardware, but it seems that 48 bits is enough for even the largest users today.

http://en.wikipedia.org/wiki/X86-64#Virtual_address_space_de...


Keep in mind that the x86-64/amd64 architecture also added a lot of other architectural improvements (like additional registers) and the processors supported running unmodified native x86 code much, much faster than the 32-bit only processors of the time. 5 years ago, if you wanted the absolute fastest x86 machine possible your best bet was to get a dual Opteron (amd64) system and just put a 32-bit OS on it. Running actual 64-bit apps also showed a significant speed improvement, regardless of ram usage.

This performance improvement, more so than the ability to address more than 4gb of memory, is the primary reason for 64-bit processing becoming mainstream, in my opinion.


From what I understand 128-bit is for HDD allocation purposes only. Memory allocation / everything else will stay 64 bit. Very large disk storage units is looking like it will challenge the 64-bit architecture in the coming years - for the heaviest users.




Consider applying for YC's Winter 2027 batch! Applications are open till November 2.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: