• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Windows 9, 128-bit ?

There is a place for having such thing as a 128bit CPU. Mapping space! The universe?
 
drescherjm said:
Have no idea what /s means..
Click to expand...
It is common in online conversation among computer specialists to use a pseudo-HTML element: </sarcasm>. The tag is often written only after the sarcasm so as to momentarily trick the reader before admitting the joke. Similarly, and common in social-news-based sites, is a single /s placed at the end of a comment to indicate a sarcastic tone for the preceding text. source
Click to expand...

It's a thing.
 
JaiWebb said:
It's a thing.
Click to expand...

Thanks. Never recall seeing that in the 25 years that I have had Internet access. Although it could have just as easily flown by me. My sarcasm detector usually fails me when it is in any written form..
 
Last edited:
AndreRio said:
There is a place for having such thing as a 128bit CPU. Mapping space! The universe?
Click to expand...

I would believe a single CPU of today would be way too slow to use 18 exabytes of memory in a productive way.

I mean if my math is correct it will take 1554 days to access each address of the memory 1 time assuming you can read at 8 GB/s.
 
drescherjm said:
I mean if my math is correct it will take 1554 days to access each address of the memory 1 time assuming you can read at 8 GB/s.
Click to expand...
Point taken, however in the short term, memory will have bandwidths an order of magnitude higher in a single package; at a system level we might see 2-3 orders of magnitude more total bandwidth.
 
devman said:
I think it is likely we will see 128-bit registers
Click to expand...

Have you heard of this completely new technology, called SSE ? It's just 15 years old right now, so you may have not noticed it yet. Also there is AVX2 (Haswell and newer) with 256-bit reg/instructions and somewhere in 2015/2016 we will get AVX-512 with 512-bit reg/instructions.
 
jwcalla said:
Right now the idea is a solution looking for a problem of course. The move to 64-bit was very much necessary and well-timed and we just don't have that scenario for 128-bit yet. There are still a lot of inefficiencies to squeeze out of what we currently have.
Click to expand...
I agree.

Outside of security improvements and higher memory addressing and larger hard drive space, 64-bit was necessary. Unfortunately, many developers are still compiling 32-bit programs. Why? Too many people running legacy operating systems like Windows XP on older hardware, and many developers haven't seen a need to move their programs to 64-bit.

For example, I can tell you firsthand that Firefox is definitely in need of a 64-bit version. I've seen Firefox 29 crash once it hits 3GB of RAM usage in Task Manager. Waterfox, 64-bit Firefox, does not do this and can go above 4GB without crashing.

128-bit?

There is currently no need for it right now. 64-bit is about as good as it can get. We can already have more than 4GB of RAM installed and access 3TB-plus of hard drive space for consumer PCs. It's going to be a hell of a long time before we see 1TB or higher memory DIMMs in consumer PCs. Even longer with exabyte capacity RAM. 16GB DDR4 RAM is going to be mass produced in under five years with the eminent release of the Haswell-E processors and X99 boards.

Yet, we're discussing 128-bit operating systems when Microsoft limits maximum RAM to 128GB for non-Pro Windows 8 and 512GB for Windows 8.1 Pro and Enterprise. You'd need Windows Server 2012 Standard/Datacenter/Storage Server Standard/Hyper-V to access the maximum allowed physical RAM limit of 4TB.

To use all of that RAM, you'll need something like this:
http://www8.hp.com/us/en/products/proliant-servers/product-detail.html?oid=4231377#!tab=specs

4TB of RAM maximum and 128 DIMM slots. All for the "low" price of $33,585, or around the price of somewhat loaded 2014 Jeep Grand Cherokee (not the one that can go over $40,000, mind you).

There aren't even boards that support higher than 128GB for consumers (very few Socket 2011, X79 boards), and there are a few server boards that max out at 768GB available on Newegg. 32GB memory DIMMs aren't even cheap either.

I can probably see database servers and the like probably needing 128-bit operating systems, but I just don't see us-- consumer, enterprise, or datacenter-- requiring one for a very, very long time. 64-bit already has the ability to address and access up to 16EB (exabyte) of memory.

I'd say we're probably a few decades or more away from needing anything higher than 64-bit. Developers still haven't moved programs to 64-bit yet and ditched 32-bit altogether. So long as legacy hardware and operating systems are still in use, we will still be stuck with a large number of 32-bit programs and applications. By the time we will likely see a 128-bit operating system is when we have something like high-capacity holographic memory or quantum computing or the next non-silicon based processors and memory. Maybe something as exotic as RRAM that has already shown the ability to cram 1TB on a single chip, or higher capacity hybrid memory cube (HBM) beyond 4GB.
 
Sure and it will include one fluffy kitten to remove all virus, malware, and porn.
 
faugusztin said:
Have you heard of this completely new technology, called SSE ? It's just 15 years old right now, so you may have not noticed it yet. Also there is AVX2 (Haswell and newer) with 256-bit reg/instructions and somewhere in 2015/2016 we will get AVX-512 with 512-bit reg/instructions.
Click to expand...

Classy.

I was NOT referring to vector storage smaller data types. SSE also existed back in 32-bit CPUs as well. I see no reason why SSE or AVX would prevent adoption of 128-bit registers beyond vector registers.
 
devman said:
Classy.

I was NOT referring to vector storage smaller data types. SSE also existed back in 32-bit CPUs as well. I see no reason why SSE or AVX would prevent adoption of 128-bit registers beyond vector registers.
Click to expand...

Agree on SSE and most SIMD ISAs not being 128b arithmetic in the conventional sense.

However - quad precision floats have been supported in some hardware as early as System/360.
 
drescherjm said:
I think it will be quite a bit longer than that to put the equivalent of 268,435,456 64 GB dimms into the typical 1 to 4 slots a PC has.
Click to expand...

Guys,

I think as a group we are making the mistake of thinking that our needs are typical, or comprehensive. A consumer PC, probably not. But in 5-10 years, a DIMM (or whatever that looks like then :) ) could be 256 GB or even 1-4 TB. And "we" aren't the entire market. What about servers? Right now Intel has separate Xeon CPUs for server and embedded markets, at much higher prices than consumer CPUs. What's to stop Intel from introducing 128-bit Xeon CPUs in 2-4 years? If that's even remotely possible, then Microsoft had damn well be working on a 128-bit Windows. That Windows could be very high-priced, with only server-based versions.
 
x509 said:
Guys,

I think as a group we are making the mistake of thinking that our needs are typical, or comprehensive. A consumer PC, probably not. But in 5-10 years, a DIMM (or whatever that looks like then :) ) could be 256 GB or even 1-4 TB. And "we" aren't the entire market. What about servers? Right now Intel has separate Xeon CPUs for server and embedded markets, at much higher prices than consumer CPUs. What's to stop Intel from introducing 128-bit Xeon CPUs in 2-4 years? If that's even remotely possible, then Microsoft had damn well be working on a 128-bit Windows. That Windows could be very high-priced, with only server-based versions.
Click to expand...

I don't think so. See: previous posts in this thread. I think you're applying trends in TV and camera consumer markets to the silicon industry inappropriately.
 
x509 said:
What's to stop Intel from introducing 128-bit Xeon CPUs in 2-4 years? If that's even remotely possible, then Microsoft had damn well be working on a 128-bit Windows. That Windows could be very high-priced, with only server-based versions.
Click to expand...

The article was about making a consumer OS (Windows 9) for a server processor that would likely never ever exist. I mean a 128 bit version of Itanium. Remember IA-64 = Itanium.

But in 5-10 years, a DIMM (or whatever that looks like then ) could be 256 GB or even 1-4 TB
Click to expand...

But not exabytes.

1 exabyte = 1048576 terabytes

64 bit addressing can access exabytes (well if it truly was 64 bit not 48 as it is today).

http://en.wikipedia.org/wiki/Exabyte
 
Last edited:
I think as a group, the people who think we're going to see a 128bit addressing in a CPU in the next 30 years, are making the mistake of not comprehending orders of magnitude. Lets assume in 5 years we somehow get a 4TB DIMM for server use. you still need four MILLION of those DIMMs in a single machine before you hit the memory cap. And a 4TB DIMM in 5 years is an absurd assumption. I'm not sure if you took all the RAM in existence in the world today if you'd be able to hit the 64bit limit. I think we're getting close, you'd need ~4 billion computers with 4GB RAM each, we're in the ballpark, if we are over 64bits of total RAM I highly doubt we'd need more than 65bits. That's for ALL THE RAM IN THE WORLD.

What's to stop intel from introducing a x128 chip in the next 2-4 years? The fact that they have engineers who DO understand orders of magnitude.
 
-Dragon- said:
I think as a group, the people who think we're going to see a 128bit addressing in a CPU in the next 30 years, are making the mistake of not comprehending orders of magnitude. Lets assume in 5 years we somehow get a 4TB DIMM for server use. you still need four MILLION of those DIMMs in a single machine before you hit the memory cap. And a 4TB DIMM in 5 years is an absurd assumption. I'm not sure if you took all the RAM in existence in the world today if you'd be able to hit the 64bit limit. I think we're getting close, you'd need ~4 billion computers with 4GB RAM each, we're in the ballpark, if we are over 64bits of total RAM I highly doubt we'd need more than 65bits. That's for ALL THE RAM IN THE WORLD.

What's to stop intel from introducing a x128 chip in the next 2-4 years? The fact that they have engineers who DO understand orders of magnitude.
Click to expand...


/end thread (IMO)
 
look on the bright side if 128bit thing does become a thing and they get that whole ram situation figured out you can say goodbye to HDDs or SSDs cause you'll have ram by the exabytes to store everything. i dont know if my logical holds water either so...... :/ i can get pretty technical with computing but when it comes to the technical of the technical i lack the IQ points :) LOL
 
Gentlemen, we simply MUST have more bits! We need not know why or how, but more bits must be had! 128, 129, whatever it takes!
 
techtechie said:
look on the bright side if 128bit thing does become a thing and they get that whole ram situation figured out you can say goodbye to HDDs or SSDs cause you'll have ram by the exabytes to store everything. i dont know if my logical holds water either so...... :/ i can get pretty technical with computing but when it comes to the technical of the technical i lack the IQ points :) LOL
Click to expand...

Even if they make RAM that can retain data after powering down, it still wouldn't need 128 bit for massive amounts of storage. As said above, you would need 1048576 terabytes to equal one exabyte. You would need like 5 million 16k videos to fill that space up (random numbers pulled out of nowhere).
 
Unknown-One said:
Windows 8.1 Update 1 can be installed in as little as 4GB, actually. They've added a new "WIM Boot" feature that seriously reduces the install footprint. Details here: http://technet.microsoft.com/en-us/library/dn594399.aspx
Click to expand...

That's actually pretty cool! It essentially provides NOR Flash Execute In Place functionality on NAND flash. That's a serious piece of work right there, but has the downside of not working with regular disk drives. :D

About goddamned time too, since the punitive size of the 64-bit install was keeping Windows off of most tablets.
 
JoseJones said:
Will Windows 9 be 128-bit or not? Some websites discussing Windows 9 are also talking about the possibility of 128-bit. I certainly do remember the 32/64-bit fiasco. Have we already maxed out 64-bit? Will going to 128-bit be another fiasco? Are we ready for 128-bit?

Microsoft Working on 128-bit Windows

What are the pros and cons of a 128-bit Operating System?
Click to expand...

Yes.
I know I maxed out 64 last week some time. Not sure what anyone else is talking about.
I am ready for 128-bit.
I can't think of any cons.
 
Tsumi said:
Even if they make RAM that can retain data after powering down, it still wouldn't need 128 bit for massive amounts of storage. As said above, you would need 1048576 terabytes to equal one exabyte. You would need like 5 million 16k videos to fill that space up (random numbers pulled out of nowhere).
Click to expand...

You could fit 52.43 million 20GB blu-ray rips into an exabyte drive.

I'll take 2, please! :p
 
thedocta45 said:
Yes.
I know I maxed out 64 last week some time. Not sure what anyone else is talking about.
I am ready for 128-bit.
I can't think of any cons.
Click to expand...

Making the rash assumption that this isn't a troll post...

Maxed out 64 what?
 
Why does everyone seem to think 64bit and 128bit are only good for addressing additional physical RAM? :confused:

There are features that can take advantage of the entire 64bit (and 128bit) address space, even if you only have a couple of gigabtes of RAM installed. Address Space Layout Randomization (ASLR) uses the entire address space, and randomly assigns chunks of it to physical resources. This makes it impossible to attempt exploits that target specific memory addresses.

ASLR can be defeated by scanning the entire address space with some intelligent pattern-matching, which is easily-done on 32bit systems (only 4GB of addresses to scan). This becomes MUCH more difficult on 64bit systems (16 Exabytes of addresses to scan, even if there's only 4GB of physical RAM installed). As you can imagine, 128bit would make ASLR even more effective (and no, the OS doesn't need to be 128bit native to handle 128bit addressing)

There are also benefits that have nothing to do with address space... Some operations (like high-grade encryption / decryption) are legitimately faster to compute with 128bit registers.
 
thedocta45 said:
Yes.
I know I maxed out 64 last week some time. Not sure what anyone else is talking about.
I am ready for 128-bit.
I can't think of any cons.
Click to expand...
No, you likely didn't max out 64-bits on a desktop system.

The limits of memory capacity of 48-bit physical addressing is 256TB, and the virtual address space is 16 exabytes (2**64), both far higher than can be attached directly to any processor by several orders of magnitude*. The maximum process size on Windows x64 is 8TB, and that's an arbitrarily low limit.

Are you suggesting that you ran into limits of precision available to the integer core when working with 64-bit operands?

Please explain why you think you ran into some limitation of 64-bit that would be solved with 128-bit. Because as it's been explored for the past couple of pages, there is zero compelling argument to widen x86 past 64-bit right now. The imperative would be seen at the high end of the HPSC and mainframe segments long before it became a necessity on the desktop.

* as I explained earlier in the thread, it would take 20 more years of perfect Moore's Law improvements before 256TB memory sizes would even be feasible and somewhat cost effective for the highest end consumer products using a reasonable number of memory modules.
 
Last edited:
pxc said:
Please explain why you think you ran into some limitation of 64-bit that would be solved with 128-bit.
Click to expand...
Plenty of software takes advantage of SSE (and the 128bit registers it includes). Maybe that's what he's talking about.
 
Unknown-One said:
Plenty of software takes advantage of SSE (and the 128bit registers it includes). Maybe that's what he's talking about.
Click to expand...
Widening x86 to 128-bit wouldn't do anything to change that. If someone wants 256-bit wide SSE integer and AVX execution resources they can use AVX2. :p

The upcoming AVX3.x widens that again to 512-bit.

IMO the people calling for 128-bit wide x86 don't understand why it's unnecessary for the foreseeable future.
 
pxc said:
No, you likely didn't max out 64-bits on a desktop system.

The limits of memory capacity of 48-bit physical addressing is 256TB, and the virtual address space is 16 exabytes (2**64), both far higher than can be attached directly to any processor by several orders of magnitude*. The maximum process size on Windows x64 is 8TB, and that's an arbitrarily low limit.

Are you suggesting that you ran into limits of precision available to the integer core when working with 64-bit operands?

Please explain why you think you ran into some limitation of 64-bit that would be solved with 128-bit. Because as it's been explored for the past couple of pages, there is zero compelling argument to widen x86 past 64-bit right now. The imperative would be seen at the high end of the HPSC and mainframe segments long before it became a necessity on the desktop.

* as I explained earlier in the thread, it would take 20 more years of perfect Moore's Law improvements before 256TB memory sizes would even be feasible and somewhat cost effective for the highest end consumer products using a reasonable number of memory modules.
Click to expand...

I could tell you but then I would have to... well you know.
 
Unknown-One said:
There are also benefits that have nothing to do with address space... Some operations (like high-grade encryption / decryption) are legitimately faster to compute with 128bit registers.
Click to expand...

With the limited applicability of 128-bit computing, I think that individual use cases that pose significant computing bottlenecks will continued to be addressed with special instructions and also special registers (or using exiting vector registers like SSE).

AES-NI uses SSE registers to hold 128-bit round keys and 128-bit data blocks for its operations.
 
pxc said:
Widening x86 to 128-bit wouldn't do anything to change that. If someone wants 256-bit wide SSE integer and AVX execution resources they can use AVX2. :p

The upcoming AVX3.x widens that again to 512-bit.

IMO the people calling for 128-bit wide x86 don't understand why it's unnecessary for the foreseeable future.
Click to expand...
devman said:
With the limited applicability of 128-bit computing, I think that individual use cases that pose significant computing bottlenecks will continued to be addressed with special instructions and also special registers (or using exiting vector registers like SSE).

AES-NI uses SSE registers to hold 128-bit round keys and 128-bit data blocks for its operations.
Click to expand...
I never said anything about widening the x86 bus (I've been advocating the use of SSE for 128bit when it makes sense instead of going native 128-bit x86 since page-one of this this thread). I'm pretty sure we're all making the same point here :p

jimmyb said:
How much more effective?
Click to expand...
When moving from 32bit to 64bit? ASLR suddenly has 4294,967,296 times more addresses to randomize within.

Finding the specific bit you want to exploit in a randomized pool of 147,573,953,000,000,000,000 bits (148 quintillion bits, the 64bit address space) is a HELL of a lot harder than finding the same exploit in a randomized pool of only 34,359,738,368 bits (34 billion bits, the 32bit address space).

An exploit can brute-force the entire 32bit address space in a matter of minutes on a modern consumer CPU. Moving to 64bit means brute-force suddenly takes tens of thousands of years.
 
Last edited:
You must log in or register to reply here.
Back
Top