Microsoft to Launch 64-bit Windows on Monday

maotx writes "Several news outlets are reporting that Microsoft will officially roll out 64-bit versions of its Windows operating systems on Monday. As compared with existing 32-bit versions: 64-bit Windows will handle 16 terabytes of virtual memory, as compared to 4 GB for 32-bit Windows. System cache size jumps from 1 GB to 1 TB, and paging-file size increases from 16 TB to 512 TB."

Re:Are there any 32-bit-only OSes left worth menti

[taskforce]

Mmm... well sort of. AFAIK, Tiger isn't fully 64bit. The only thing that is 64 bit in Tiger is its support for 64 memory for POSIX based apps...(Command line apps; server daemons and such) Tigers kernel will reside in 32 bit address space so that it can still run on the G3 and G4. XPx64 is fully 64 and as such can only run on the x64 architecture and is by no mean a patch up job.

Re:Are there any 32-bit-only OSes left worth menti

[Anonymous Coward]

This isn't the first version of 64-bit Windows; it's the first x86 64-bit version.

64-bit Windows has been available for Itanium for several years now.

Re:Serious performance problems..

[Abreu]

Goatse alert!
Mod parent down

Not 64-bit, just x64 editions

[cyberjessy]

The submission is absolutely misleading.

Windows Server 2003 has supported 64-bits (Enterprise Edition and Datacenter edition) since its launch on IA64(Itanium). Before that, they also had 64-bit versions of Windows 2000 Server.

Windows XP Professional also had a 64-bit version since 2003, again running on the Itanium. However, XP on Itanium was discontinued as no one was using it outside MS testing labs.

Whats gonna be launched are x64 editions of XP and 2003 Server.

Re:Are there any 32-bit-only OSes left worth menti

[Anonymous Coward]

DEC had 64 bit ALPHA processors in 1992, Linux didn't gain 64 bit support until 1996. That's still early, but your statement is not true.

Re:Are there any 32-bit-only OSes left worth menti

[Panaflex]

Technically, Windows NT on Alpha (and Solaris I do believe) was merely running in 32bit "mode" (Still used 64bit pointers, however the top word was merely masked off.)

It wasn't until Windows 2000 for Alpha (the version that was literally cut right before shipment... some people managed to get a copy) that full 64bit apps were available!

-Pan

Re:32-bit Windows != 4 GB memory

[Ann Elk]

Yes, I believe that's why the summary states "64-bit Windows will handle 16 terabytes of virtual memory, as compared to 4 GB for 32-bit Windows."

quantum leaps

[Anonymous Coward]

A quantum is the smallest discrete amount.
So, the key word here is leap. It's a 32 bit leap.
Now this is a pretty big leap. Even at one G, it is much greater than say a 'giant leap'.

Think about it a minute (which is a huge number of Cesium state transitions). A small step might bypass one quantum unit. i.e. 5 bits to 7 bits. A giant step might bypass 2 - 6 quantum. The mother of all leaps, some 16 bits pales in comparisonto 32. NO! this is a mind boggling leap. More in the category of Rebus Kaneebus jumping to the center of the earth.

But don't forget. We owe it all to two state signalling. Thank you Morris the code cat.

Which brings to mind. If a chicken can learn to play the piano, why can so few humans learn Morris code?

Re:still a 32-bit file system?

[BlacBaron]

Don't know what you're problem is but I have Windows 2000, and using NTFS i've got some files that are over 8gb.

Re:It has been out in beta for a while

[NetNifty]

Ignore other post without paragraphs.

I'm not the original poster, but I'm running XP x64 RC2 and have had problems with Nero (asks for enterprise key just to run, then it works fine), printer drivers for Canon IP2000 (although driver problems are expected, and the built in BJC-8000 drivers work fine for printing, have to hook it up to 32-bit Windows machine to do head cleaning etc), ZoneAlarm doesn't install (although Tiny has a 64-bit Windows firewall available now), a few motherboard utils for my A8N-SLI Deluxe motherboard won't run (but 64 bit versions seem to be appearing), Doom 3 and some other software complains when installing - but editing the MSI file, or running in Windows XP compatiblity mode to get around this usually lets it install and run fine. Had a problem with GetRight crashing so switched to Free Download Manager (shared internet connection so really need the speed capping), haven't tried any BitTorrent apps (hacked together an app which passes torrents to my laptop) but presumably will have same problem as 32-bit SP2 - initialising socket caps.

Apart from my printer, all my hardware works fine (A8N-SLI Deluxe motherboard, NVIDIA 6600GT PCI-Express graphics card, 1GB Crucial PC4000 RAM, 200GB Maxtor Diamondmax 10 SATA HD, 120GB Maxtor Diamondmax 9 ATA133 HD, NFORCE4 onboard sound, NEC ND-3500 DVD burner, and some other generic 8x DVD reader), although it can be a big sluggish when copying large files from/to HD I think that's down to drivers rather than anything else.

Using Firefox 1.0.3 for browsing, Media Player Classic 6.4.8.2 for video, Winamp 5 for music and never had any problems with them, so don't know what poster above is talking about unless is using a very early build (used 1218 previously and only had same issues as I do now - only difference I noticed was upgraded Windows apps - IE got SP2'd with popup blocker, Solitair is 64-bit etc).

Re:Third party apps

[pershino]

I run gentoo on AMD64 which has 32-bit compatibility modules which allow running 32-bit apps without the need to chroot. Gentoo's portage also provides 32-bit binary versions of Firefox, Flash player and OpenOffice, amongst others. All works perfectly. Surely other distros do the same?

Re:The most important question-

[SA Stevens]

Solitaire is a relatively new component of Windows. Windows before 3.0 had a different compliment of games. Reversi was the main one I remember. Might have been the only one.

Re:still a 32-bit file system?

[man_of_mr_e]

Uhh.. NTFS has always been a 64 bit filesystem. FAT is 32 bit and only supports 2GB files.

Also, many C based apps only use a 32 bit file pointer, so that could be your problem as well.

Re:still a 32-bit file system?

[pegr]

I routinely work with files over the 2gb limit...

That's a filesystem limit, not an OS limit. Use NTFS (yuck!), and you won't have that problem...

Fat Binaries

[Anonymous Coward]

Not true. The system is based on Fat Binaries. These allow an executable package to contain code for multiple platforms. Theoretically, you could put versions of the same application for NextStep MC68K, OpenStep, OSX 32, OSX 64, Darwin x86 and several others in the same package and have it look like one program. Too big for you? Run lipo to remove versions you don't need. The whole system is based around this concept, allowing the OS to be fully 64bit on 64bit systems and fully 32bit on 32bit systems. Even XCode allows people on 32bit machines to design and compile applications for both platforms and release the compiled application in fat binary format.

Re:I'm at a loss for words

[DrSkwid]

We aren't talking quantum leaps in computing, and that's the problem.

Lol, you should go find out what a quantum leap actually is.

The jump from 32 bit to 64 bit Windows is precisely a quantum leap.

Unless you can show me the (infinite number of) versions of Windows that have 32 > bits 64 !

Re:what, only 16TB?

[pmjordan]

Well, if you need an int of a particular size, you need to typedef yourself a compiler/platform specific one anyway; this has always been that way. I believe 'int' used to be 2 bytes on most 16-bit DOS compilers. If you're talking about pointer arithmetic and array indexing, you should have been using size_t for those uses all along.

By the way, the GCC sizes for i386 and x86_64 are:
int: 4 bytes, 4 bytes
long: 4 bytes, 8 bytes
long long: 8 bytes, 16 bytes
size_t: 4 bytes, 8 bytes (I believe it's typedef'ed as an unsigned long)

Note that long long isn't ANSI/ISO compatible.

~phil

Re:Are there any 32-bit-only OSes left worth menti

[Queer Boy]

Mmm... well sort of. AFAIK, Tiger isn't fully 64bit. The only thing that is 64 bit in Tiger is its support for 64 memory for POSIX based apps...

The only real advantage 64-bit has over 32-bit for anyone outside of the supercomputing realm is the memory it can access. ALL applications in Tiger can access 64-bit memory if they are written for it meaning the backend is not written for Cocoa but for Darwin. [apple.com] BTW, Darwin is different than POSIX.

The true genius of Apple is that the data model for Tiger is LP64 which means source for Linux, SGI and Sun is easy to port to the G5 with Tiger.

Re:what, only 16TB?

[tomstdenis]

... arrg I was gonna mod in this discussion... but ...

"long long" is eight bytes on __x86_64__ platforms [e.g. AMD64 with GCC].

long long is also C99 compatible and has been available in GCC and most unix cc's for a very long time.

Tom

They had to do something...

[connah0047]

...their error report server could not handle the Explorer crash reports.

Re:what, only 16TB?

[Anonymous Coward]

Having programmed on numerous 64 bit OSes, I can authoratatively you're completely full of shit.
short = 2 bytes.
int = 4 bytes.
long = it depends.

So are you. (well, not *completely*, I guess)
short = at least 16 bits
long = at least 32 bits
int = at least as wide as short and no wider than long

So says the language spec.

long is 4 bytes on most (all?) 32 bit OSes. long is 8 bytes on most (all?) 64 bit OSes.

I don't know about that, but the amd64 native data size is 32 bits. Aside from that, Windows has a *lot* of legacy software that assumes that long is 32 bits. (can't use int, that used to be 16 bits on Windows.) Would you upgrade to an OS that stopped all your (broken, but works) code from compiling?

Re:It has been out in beta for a while

[RichM]

No you don't.
I have been running the final release downloaded from MSDN for a couple of weeks and I'm using Firefox 1.0.3 to post this, while I listen to mp3s on Winamp and talking to my mates using Teamspeak.

I maintain a list of programs which do and don't work here:
http://www.cableforum.co.uk/board/article.php?a=64 [cableforum.co.uk]

Re:Are there any 32-bit-only OSes left worth menti

[pohl]

Just in case anybody was curious about what the term LP64 implies, or what the alternatives choices where, this page [wikipedia.org] describes them.

Re:Third party apps

[jdmuir]

• I miss things a lot of the little things (like the flash plugin) that were never compiled for a 64 bit system.

Run the 32-bit version of Firefox all of your plugins will start working again.

Re:Are there any 32-bit-only OSes left worth menti

[forkazoo]

I am not aware of a 64 bit port to SPARC. I know they had 32 bit build for alpha, ppc, MIPS. I have heard from reliable sources that there was a 64 bit port of NT 5 beta (eventually named Win2k) to Alpha. Apparently, a lot of the code from that port was used to make the 64 bit Itanium port.

Re:Fat Binaries

[Anonymous Coward]

Fat binaries have nothing to do with 64-bit suppor for GUI apps.

"It is important to note that in the Tiger release, the support for 64-bit programming does not extend throughout the entire set of APIs available on Mac OS X. Most notably, the Cocoa and Carbon GUI application frameworks are not ready for 64-bit programming. In practical terms, this means that the "heavy lifting" of an application that needs 64-bit support can be done by a background process which communicates with a front-end 32-bit GUI process via a variety of mechanisms including IPC and shared memory."

http://developer.apple.com/macosx/tiger/64bit.ht ml

Already available for download on MSDN

[Anonymous Coward]

Windows XP for 64 Bit processors is already available for download on the MSDN download site.

Re:Are there any 32-bit-only OSes left worth menti

[drsmithy]

BTW NT is little endian. Running on a big endian cpu would like change EVERYTHING..

NT is endian-neutral (else it wouldn't be able to run the new PPC970 Xbox2, as the 970 is only big-endian).

I'm not entirely sure *when* that happened, as the old Alpha, etc, versions of NT all ran the chips in little-endian mode, but it is certainly true today.

Re:Third party apps

[turgid]

This is a short-comming of the design of 64-bit debian systems. The way Solaris does it is to have 32-bit and 64-bit user-land libraries and utilities side-by-side so that you can run 32-bit and 64-bit binaries on the same 64-bit system at the same time. The debian people chose to break backwards compatibility when they went to 64-bit. I suppose from debian's ideological point of view "everyone should be using Free software and compiling from source" so it doesn't matter. However, in the real world, it does.

Now, if only someone would lend me an AMD64 machine I'd do that badly-needed Slackware AMD64 port, and I'd do it like Solaris...

Re:what, only 16TB?

[NearlyHeadless]

Well, if you need an int of a particular size, you need to typedef yourself a compiler/platform specific one anyway; this has always been that way.

For C99-compliant compilers (and gcc), there is <stdint.h> [opengroup.org] which defines:

int{N}_t uint{N}_t
int_least{N}_t uint_least{N}_t
int_fast{N}_t uint_fast{N}_t
intptr_t uintptr_t
intmax_t uintmax_t
INT{N}_MIN INT{N}_MAX UINT{N}_MAX
INT_LEAST{N}_MIN INT_LEAST{N}_MAX
UINT_LEAST{N}_MAX
INT_FAST{N}_MIN INT_FAST{N}_MAX UINT_FAST{N}_MAX
INTPTR_MIN INTPTR_MAX UINTPTR_MAX
INTMAX_MIN INTMAX_MAX UINTMAX_MAX
PTRDIFF_MIN PTRDIFF_MAX
SIG_ATOMIC_MIN SIG_ATOMIC_MAX
SIZE_MAX WCHAR_MIN WCHAR_MAX WINT_MIN WINT_MAX
INT{N}_C(value) UINT{N}_C(value)
INTMAX_C(value) UINTMAX_C(value)

Where {N} can be 8, 16, 32, and, if supported, 64

As somebody else noted, c99 also supports long long. Of course older compilers don't have stdint.h. I don't think Microsoft C does either, although I don't have the latest version.

Re:Are there any 32-bit-only OSes left worth menti

[Detritus]

The Alpha was 64-bit, the Alpha version of NT was mostly 32-bit due to the architectural limitations of NT.

Re:64bit or 32bit

[Anonymous Coward]

It includes both:

C:\WINDOWS\system32>dir sol.exe
Volume in drive C has no label.
Volume Serial Number is D489-189D

Directory of C:\WINDOWS\system32

03/25/2005 05:00 AM 78,336 sol.exe
1 File(s) 78,336 bytes
0 Dir(s) 31,403,466,752 bytes free

C:\WINDOWS\system32>dir ..\SysWOW64\sol.exe
Volume in drive C has no label.
Volume Serial Number is D489-189D

Directory of C:\WINDOWS\SysWOW64

03/25/2005 05:00 AM 57,856 sol.exe
1 File(s) 57,856 bytes
0 Dir(s) 31,403,466,752 bytes free

In general 64-bit windows is laid out with 32-bit copies of everything (for interoperability w/ legacy 32-bit software) and 64-bit copies of everything. If you were to start Solitare from the start menu you'd get a 64-bit solitate. If you were to do start, run, %windir%\syswow64\cmd.exe and then run "sol.exe" then you'd get the 32-bit version (launched from a 32-bit command prompt). If you run %windir%\system32\cmd.exe and then run "sol.exe" you'll get the 64-bit version run from the 64-bit command prompt (yes, the folder with 64 in it's name contains 32-bit versions, and the folder with 32 in it's name contains 64-bit versions).

Another interesting example is Internal Explorer which is included in both versions and by default you get the 32-bit IE (probably because all the browser plugins are 32-bit).

Things get even more interesting when you're at the 32-bit command prompt. You start seeing what to 64-bit users is "syswow64" as "system32". And all the registry keys get re-mapped to a Wow6432Node (ie, HKLM\Software\Wow6432Node contains all the Software settings for 32-bit apps).

What all of this means is that there's no virtual machine needed. The 32-bit apps run on the 64-bit OS and are exposed a 32-bit set of DLLs that call into the 64-bit OS for all of their operations. So if you have a program that spends a lot of time in OS libraries then it'll potentially see some of the benefits of running on a 64-bit OS.

Re:why should I care?

[uarch]

Different concept.

The x86 micro architecture specifies 8 32-bit integer registers. (some of which are used for stack pointer, etc)
In x86-64 this was raised to 16 64-bit registers.

If we were talking address space it would be:
2^64 - 2^32 more virtual addresses

Every additional bit that we tack on doubles the address space. Adding 32 bits double's the possible values 32 separate times.

Re:still a 32-bit file system?

[Deviate_X]

Actually NTFS has supported files upto 16TB in size since 1992.

type Convert c: /FS:ntfs on the command line if u are using a non 9x OS

Comment removed

[account_deleted]

Comment removed based on user account deletion

Re:Windows bits

[Anonymous Coward]

Let's attribute this quote properly, shall we? Original quote here [bash.org]

Re:Are there any 32-bit-only OSes left worth menti

[Lost Race]

I used to run NT 3.51 and 4.0 on an Alpha, and I can confirm that those versions of NT were definitely 32 bit, not 64 bit. I did software development using the native Alpha version of Visual C++ and pointers were 4 bytes long (32 bits). There were no 64-bit versions of the Windows or NT kernel APIs at the time so the entire operating system, shell, and native applications were all 32-bit. I'm not talking about the x86 emulator, which was obviously 32-bit, I'm talking about native AXP code, which was also 32-bit. Alpha supported native 32-bit pointers and little-endian integers (probably just for NT).

I also used RedHat on the same machine, and its pointers were 8 bytes long (64 bits).

Re:Windows bits

[glitch23]

That isn't the original saying though. 32-bit shell on a 16-bit OS (DOS)...

Re:I don't know abou this...

[spworley]

Gates never made that famous "640K is enough" quote. It's an urban legend.

http://tafkac.org/celebrities/bill.gates/gates_mem ory.html [tafkac.org]