OSNews

Possibly, many of the lowpower chips such as the VIA C7 are 32Bit only and may still be that way in 3-5 years

What stops You from running 32 bit OS on 64 bit chip? I'm doing it.

A 32-bit OS has a constraint of being able to address only 4GB of memory.

In 3-5 years, entry-level machines will have more memory than that.

http://catb.org/~esr/writings/world-domination/world-domination-201...
http://catb.org/~esr/writings/world-domination/world-domination-201...

The 4GB limit applies to each process. Even the original i386 processor could directly address 64TB of RAM.

The 4GB limit applies to each process.

Not in vista, 32-bit. XP could use PAE, but 32-bit Vista won't, which seems to be an artificial limitation.

Unfortunately I can't find a good link, but search the net and you'll find lots of people who have been trying to get 4GB to work in Vista.

edit: found a link.
http://help.lockergnome.com/vista/Vista-32bit-recognize-memory-ftop...

Edited 2007-07-23 09:57

The original i386 could only address 4GB RAM. It had only 32 address bus lines.

The first x86 to break this barrier was the Pentium Pro, which had 36 address bits, hence 64GB RAM.

I don't know of any processors that can address 64TB of RAM.

i386 processor could directly address 64TB of RAM.
OMFG!!!

"Will there even be 32-bit chips for general sale then?"

For sale in general? As a new product? Surely not. But will there be 32-bit chips still around that are capable to run this new "Windows"? :-)

Theoretically yes; practically, no.
It's like saying you can run Vista on 800 MHz. You can do it; in the same way you can run XP on 8 Mhz http://www.winhistory.de/more/386/xpmini_eng.htm

By the time '7' is out, 32-bit machines may be considered unbearably slow, who knows.

64bit isn't any 'faster' unless you are doing stuff that needs more than 4GB or RAM, and again it's only faster because you can but more than 4GB in a 64bit system, where as 32bit is limited to 4.

The actual 64bitness of the processor does not make it any faster.

That's not what I meant. 32-bit systems won't be shipping at all soon. So by the time '7' comes out, computers with 32-bit chips will be old and possibly considered slow (by GHz / cores) compared to whatever things we have in the future.

Not in and of itself, but most 64-bit processors, since they are by definition freed from the need to directly handle 32-bit architecture, they can organize themselves more efficiently and offer more advanced functions compared to their 32-bit predicessors.

CPUGuy wrote:
-"64bit isn't any 'faster' unless you are doing stuff that needs more than 4GB or RAM, and again it's only faster because you can but more than 4GB in a 64bit system, where as 32bit is limited to 4. "

well, in terms of possible optimizations 64bit has the upper hand, which may very well translate to faster code.

the 32bit x86 cpu has 8 32-bit registers available to the programmer, and out of those, only 6 can really be considered general purpose, since the base and stack pointers impose restrictions when used. this forces the programmer/compiler to push and pop values to/from stack when they run out of registers which can be very detrimental to speed in time critical code. the 64 bit cpu's has 16 64-bit registers, of which 14 are general purpose and thus offers better possibilites for code optimization.

That's proveably crap, I've got lots of written software that has 2 different code paths, one for 64 and one for 32 which the 64bit machine is substantially faster with.

Granted, the software needs to take advantage of the extra address space to do it, but 64bit is currently an enabler for faster technology.

And as the number of cores increase the ability to run within the arbitrary stack + heap limitations with 32bit will squeeze software developers.

But will there be 32-bit chips still around that are capable to run this new "Windows"? :-)

There aren't a ton of those chips that can comfortably run Vista, to say nothing of Windows 7. They might exist, but will likely be completely unsuitable if the jump in requirements between Vista and 7 is anything like the jump from XP to Vista.

In the embedded space things could be different, but there's hardly a reason for a desktop OS scheduled for release 3 years from now to be 32 bit. Why would they delay this increasingly necessary transition even more?

I don't understand why they're planning to continue shipping separate 32 and 64 bit versions? Apple has already proved that universal binaries that run on both platforms are quite possible, and it won't be long till we have the full operating system on 32 and 64 bit in a single package.

Splitting it up like this just seems like an unnecessary hassle for Microsoft's customers. Why should they worry about which edition to purchase? It should just work.

"I don't understand why they're planning to continue shipping separate 32 and 64 bit versions?"

I'm pretty sure a decision like this was actively made by Microsoft, and more than likely it's cost effective for them to ship separate media according to 'bitness'.

Also, don't forget that all SKU's of Vista come on either a Retail/OEM 32-bit, or Retail/OEM 64-bit DVD; the edition which is installed from the DVD is determined by the license key entered, so it's not like there are 12 different DVD combinations they have to press. From a manufacturing standpoint, I'd say it seems pretty efficient.

How is it efficient? Sun for over a decade shipped cd's with 64bit and 32bit binaries on the same media - Microsoft chooses NOT to do that or even offer BOTH at the same time because it means it forces the customer to purchase a new licence when they want to move from 32bit to 64bit.

It has nothing to do with 'efficiency' and everything to do with money extraction. If they wanted to make it easy, the choice would be automatically made for the end user when the OS is installed - like it is with Solaris x86.

Apple is running a 32-bit OS, so they have different constraints. It's a bit strange, but Apple has been supporting a set of 64-bit userspace programs on a 32-bit kernel, so it's not even necessary for them to have two packages for the OS. It's not so easy with 64-bit Windows, which is a clean implementation.

I hope they finally replace the kernel with something good in OS 11.

As for maintaining two versions on the Windows side, I agree with you that having only one will simplify hardware choices. But the embedded market will suffer a bit if they discontinue 32-bit Windows. I believe Microsoft announced that 2008 is the last 32-bit version of Windows Server.

What's wrong with the current hybrid microkernel kernel?

It has slow system call speeds and much lower concurrency than the other two mainstream OSes. The design is not that clean because there are two ways of doing many things in the system layer (the Mach way and the BSD way).

OS X is good enough for displaying pretty graphics and running compute-intensive SMP programs, but when you have anything that needs to call into the kernel often, it's not that fast.

OTOH, you should take this with a grain of salt... I haven't actually dived into the Darwin sources anytime recently, and I haven't owned a mac. I'm basing what I'm saying on benchmarks I have read and general overviews articles of how the Mac kernel works and some interesting postings on Apple's developer lists.

Unless 64 bit chips can run on a lower power diet than 32 bit ones can (that seems unlikely) embedded products that use Windows aren't likely to go to 64 bit chips for that reason, even if the prices of 64 bit chips aren't higher (another reason why not) and it's logical that Microsoft will transition their embedded versions over to the most recent client codebase for the base code.

In addition, as odd as it may seem to the tech geek that insists on replacing their systems every couple of years, there will still be quite a few people in the world that are using 32 bit desktop machines from this time period, because they still work perfectly fine for their usage, and they don't want/need anything higher-end to work with, and in a lot of places, it comes down to budget: not everyone has nearly as much disposable income for electronics as many of the higher-income industrialized countries. Granted, from reading on anandtech about some of the latest video games having issues with the 2 GB userspace limit, the people still running 32 bit processors at that time aren't likely to be the hard-core gamers that run these applications, but then again, those gamers aren't likely to be using systems that old, either, if they can avoid it.

Also, in the business arena, again, there will be many machines out there that are 32 bit processors, because they don't feel a great need/desire to upgrade to the latest machines, because most business users (people using word processors, spreadsheets, other more typical clerical stuff) frankly don't need even as much as current typical 32 bit machines offer, and why spend money on a new machine if the old ones are working well?

The real question as to how quickly the Windows world will transition over to new stuff that's only 64-bit rests largely in the hands of the developers, combined with customers currently using 32-bit versions of Windows and transitioning over: if there's not much market for 64-bit only software, developers are often not very willing to make that jump, and the overhead for developing and supporting both 32 and 64-bit software isn't zero. For an example, look at how long it took to get many applications to target only 32-bit Windows: Windows 95 was 2 years after Windows NT 3.1, which had very few 32-bit applications available for the longest time, and even for quite a bit of time after Windows 95, people were still developing games that didn't use it for various reasons, using EMM, etc. so yes, I wouldn't be surprised if there's still a meaningful market for 32-bit Windows 3 years from now.

64b CPU won't of course use less power than 32b one, but I doubt that they will use much more power though.

>there will still be quite a few people in the world that are using 32 bit desktop machines from this time period,

Sure, but those people won't upgrade their OS too,
so for a new OS what matter the most is the new PCs..

Microsoft probably learned some lessons from the Vista development cycle. Maybe they'll apply them this time around. The biggest mistake they made was promising these bold new pillars, at least one of which was dropped. The remaining pillars were not nearly as exciting as Microsoft thought they would be, and it's not clear that third-party developers are chomping at the bit to use these new proprietary APIs.

But they did effectively sell really scary doomsday scenarios that made the released version of Vista seem comparatively cute and cuddly. Remember Hailstorm, the plan to turn Windows into a subscription-based remote application platform? Microsoft should announce a bold plan to make Windows 7 completely unpalatable to just about anyone.

For example, announce Microsoft Media Protector, a mandatory system service that Protects all media on your computer with special DRM required for playback on Windows 7. Then back away from this plan about a year before the release, citing licensing issues with copyright holders that don't want to Protect their users. Suddenly, Windows 7 won't seem so awful. It'll seem downright acceptable.

It has also been a giant lemon as well - if they had 'server' and 'desktop' edition; and if 'desktop edition' was what 'Ultimate' is, then it would be a pretty good bargain. The problem is that what they've sold is a castrated version of Windows which provides no benefits when you transfer from Windows XP Home to the equivalent version in Windows Vista.

As for third parties - personally they get what they deserve - they choose to neglect alternative operating systems, let Microsoft crush and bankrupt each company one by one and suffer the most painful humiliating defeats. The day they refused to create alternative operating system versions of their applications is they day I stopped caring - I can't wait till I see adobe go bankrupt given the products Microsoft is releasing, not only will it be better but cheaper.

the AMD chip i have in my computer right now can in theory run any X86 code, from 8-bit to 64-bit...

"Will there even be 32-bit chips for general sale then?"

NO, I'd be surprised to see any x86 32 bit processors being sold. AMD ( & INTEL? ) presently only sell 64 bit versions. VIA has 64 bit too but I'm not sure if they still sell their 32 bit cpus.

Why make 32 bit version of the OS?

Well, because lots of people will still be running 32 bit computers ( not everyone will upgrade systems ). Microsoft would lose those customers to Linux or running *older* Windows ( = less revenue ).

If your 32bit processor runs everything you want it to then why even upgrade to 64bit processor? ( Just because it is the latest & greatest doesn't mean everybody has to go buy into it ).

Edited 2007-07-22 18:43

Its interesting about the three years, there have been mixed messages - first it was three years, then 2011 was thrown around, then another Microsoft person said it will be three years but might be longer if it isn't 'up to scratch'.

To me, if I was a Microsoft customer, I would be deeply concerned that they're worrying about the next release before getting their Windows Vista and Server 2008 sorted out properly.

As an antique dealer friend said to me - "...concentrate on this deal. Forget the next deal 'til you've tied this one up."

Very true, very true. I do the same thing at work. People crap on about 'multitasking' when in reality they're doing more than one thing, but doing each of them very crappily - kinda like those multifunction mobile phones.

I get one thing done, finished, the move onto the next thing. Concentrate your full and undivided efforts on one thing at a time then you don't lose focus on not only the big issues but the small things that can cause the biggest problems for customers.

The problem with Microsoft, also, is their focus on the profit rather than than the product. The best CEO I remember hearing was, "focus on the product and the profits will follow' - the problem with Microsoft, its all about profit and cutting costs rather than developing a product then allowing the profit to naturally flow from the fact that people will want the product.

Personally, I set up a small network full of 32 bit Windows, server and XP clients because 64bit just isn't there yet, contrary to what AMD and Intel would have you believe.

Stability and compatibility isn't a problem at home, but when your network depends on reducing downtime, you need to go with what works, and that equates to staying 32 bit and forgeting you ever heard the word Vista.

I agree with your point about app compatibility, I think a lot of people are overlooking that, but I'm curious to know which laptops are you currently buying that come with a 64-bit OS?

"I can safely say that 75% of regular 32-bit apps work without a glitch on a 64-bit Windows OS. "WoW64" works very well. And most apps and drivers are already 64-bit, so a 32-bit OS is totaly unnecessary."

I can back that up as well. I have 3 x64 XP machines at home, all provisioned for different purposes and thus running different software (>100 3rd party apps total). I actually have yet to see any issues with application compatibility, and have only had issues with 3 drivers (an older printer, a scanner, and a sound card) out of around 60 total. For the drivers that don't work, it's not a big deal to purchase newer hardware since the benefits of x64 outweigh the cost of buying said new hardware.

Wow.....3 whole machines, huh? Sounds like your little home network really mimics that of a large corporate IT

"Sounds like your little home network really mimics that of a large corporate IT"

Corporate IT is a controlled environment...the vast majority of the machines are A) going to be very similar to one another and B) tested thoroughly before deploying any kind of software. Thus if they were to deploy an x64 solution, I'd expect it to have been tested enough that it would work on 95%+ of the machines in the organization.

I'd say my 'little home network' is very typical of the hodgepodge of older hardware and varied software that a lot of other home users who would need x64, i.e. power users who are software developers, or media producers, etc, and thus is actually a very good reference for the types of issues I'd expect to see out in the uncontrolled space of home consumers where machines vary widely.

So what was your point again?

The no backwards compatibility approach is what I've been really waiting for Microsoft to do, like Apple did when they transitioned from OS 9 to OS X. Apple proved that you can successfully make a move like that, as long as its handled properly.

"Isn't three years a bit soon to release the next version?"

For businesses with software assurance, it doesn't really matter, and seeing as how they were the most vocal (along with developers) about how long it took to get Vista out the door, this should appease them quite well. This is an example of MS reacting to market pressure, and I personally feel it's a good move on their part. A lot of folks were extremely unhappy about the amount of time between XP SP2 and Vista.

Really? My impression was rather that a lot of people were unhappy that, after taking all that time to develop the new OS, many of its most interesting features were dropped for the final release...

Meanwhile, most Windows users around me are not that excited to move to Vista...I have a hard time believing that a sizeable portion of the Windows userbase was indeed impatient about getting the new version. Geeks and power users, sure, but Joe User? That seems unlikely.

"Geeks and power users, sure, but Joe User? That seems unlikely."

I was speaking about businesses, specifically businesses with software assurance licenses since they are paying a ton of money per year so that they are guaranteed the most current releases of MS software. MS makes a lot of their revenues from these types of customers, and the majority of them were irate that here they were paying all that money only to have to wait for a new OS to get released.

Businesses (like in all other industries) comprise the bulk of sales, so that's the demographic they need to keep the most happy. A 3 year iteration (if they can stick to it) is a win win situation for everyone.

Edited 2007-07-22 22:38

Good point. I had forgotten about those.

I'd question that. Enterprises don't want to see major infrastructure turning over every 3 years. They can take a year or more just to evaluate/test/rfq/purchase/deploy applications. Remember how long companies held on to NT? Some of the banks are still running it. Many companies are still using 2K on the desktop. I know our company is imaging XP overtop of Vista on new systems, just as they imaged NT and 2K over top of XP previously. And I'd question Microsoft's ability to properly provide 10-year enterprise support cycles when flipping the product that much more frequently.

The issue in the past wasn't so much that companies were demanding a 3 year release cycle, it's that MS wasn't delivering on their commitment with Software Assurance. It's a similar but not identical issue.

Frankly, I think the 5 year release cycle is fairly sane as far as companies are concerned, but I think it's a strain for consumers, or more correctly, the hardware manufacturers that depend on Windows driving upgrade sales. But then again, that was the choice MS made when they decided to merge the consumer and commercial brands of Windows. That, I think, was a mistake.

Here's one possible way

1 year since Vista's release where Windows 7 spent time out of the spotlight in planning 1 year.

The Windows 7 announcement, and 3 years intense development: 4 years total.

1 year extra-long beta for catching bugs, since that program seems to have been proven successful thanks to Vista and Windows Home Server: 5 years total.

"Isn't three years a bit soon to release the next version?

NO.

Ok, looking at Microsoft's past actions.

They create Windows 98. Then the next *new* OS afterwards is Windows 98SE.

They develop Windows 2000 followed by Windows XP.

Simply said, Windows 98 & 98SE are very similar. Windows 2000 & XP are very similar. ( 98SE and XP are just improved/updated versions ).

Why did Vista take so long to come out?

Because they were throwing lots & lots of *new* stuff into Vista ( big changes ) and it was getting very complex ( with lots of code to work with ).

Windows 7 is very likely going to be a *polished* ( improved & updated ) version of Vista maybe with a couple of new features, add-ons, fixes and improvements. If they don't go crazy adding stuff to it then they'll be able to come out with next Windows version in 3 years time.

Edited 2007-07-22 19:00

I'm not quite so sure. If http://www.levenez.com/windows/history.html#01 is correct then:

They had 2 stable source branches running side-by-side for a long time, the 3.1/9x product, and the OS/2 NT (New Technology) tree.

When the 9x tech finally had too much instability/cruft/old-code to maintain properly, then migrated the NT source over to be more consumer-friendly, and created the XP branch.

I believe that the NT codebase is coming to a similar place as the 9x line at the time of ME. There's just too much old engineering in the works to be efficient.
Vista has lots of shiny, decent features, but there is just too much instability (on some hardware) and odd performance wierdness around for many people to trust the platform. It is reminiscent of the dreaded Word 6.0 days when you had to hit save every 5 minutes.

Unfortunately, they cannot just switch to another branch, as they did with NT (unless they fall-back to modifying Windows Mobile or XBox for desktop use), Singularity seems just too far off at the moment. So they are forced to battle with the NT code until they come up with something better.

This last bit is just my opinion BTW.

So they are forced to battle with the NT code until they come up with something better.

There's nothing wrong at all with the NT codebase. The problem with Windows is the userland on top, a whole bag of spaghetti code that is not easily changed due to Microsoft wanting to maintain backwards compatibility at whatever cost.

As I said in an earlier comment in this thread, and in an old editorial of mine [1], the only way for Microsoft to effectively go forward is to wipe the slate clean, and start writing the userland from scratch on top of the already-proven and well constructed NT kernel - without backwards compatibility in mind. They should move backwards compatibility into a VM, similar to how os/2 handles win3.x/dos applications, and how OS X for PPC handles OS7-9 applications.

[1] http://www.osnews.com/story.php/14412/Why-I-Am-Indifferent-About-Vi...

Edited 2007-07-22 22:24

You're right. I was (sloppily) talking about Windows as the kernel and userland package that is sold, not just the kernel, (however a lot of the backward compatibility issues are in the kernel anyway.)

I don't see that MS would be able to restart the userland stack in the near future even if they wanted to . To announce such a dramatic shift in Business model would be seen as an admission of Vista's failure by many people which would effectively waste most of MSs investment in the product.

Yes they where throwing a lot of new stuff into Longhorn, making some very big changes. Then time and Apple brought them under pressure and Longhorn was skipped.
The Vista we know is a 2003 Server base system including a few shiny hacks.
For example the WinFS (Longhorn version) was at first 100% database based, but they never got it to work and it was skipped. Plan B was to make a metadata database driven add-on to the existing NTFS - sort of like what Apple had done. But now the klock was ticking (Or they couldn't get it to work properly) and finally that also was skipped.

Vista and most likely 7 (Vienna) is a transitional release. The strategic main purpose is to migrate people from the Win32 platform and onto the .NET platform, making the underlying operating system somewhat irrelevant. The backward compatibility of Vista is maintained using the technology bought from Connectix, also known as Virtual PC. Running Win32 apps within an VM also boxes them in, seperating them from the rest of the system - removing the security weaknesses of the Win32 framework as number one on the todo list.

After Vienna/7 I believe that MS could skip the NT branch, making an applemove and moving the .NET platform onto something BSD-driven.

Comment : And yes NT used to run on severel other architechtures, for instance the Alpha (You are missed), that for some time was development platform.

Edited 2007-07-24 05:43

The problem for Microsoft is that competitors like Linux and MacOS comes in frequent small incremtal upgrades. If Microsoft waits too long new users will be tempted to go with some other platforms than windows.

Other than that, you are probably right most existing windows users are probably quite satisfied with XP or Vista, so Microsoft may have an hard time getting people to upgrade before they have got some return of current investments. It have been that way since the release of win2k. The fact that the quality have gotten a lot better since then doesn't help Microsoft.

"The problem for Microsoft is that competitors like Linux and MacOS comes in frequent small incremtal upgrades"

It's not a problem for MS when those small incremental upgrades also come with small incremental breaking changes that cause problems with existing software, which is usually the case (especially with Linux). Normal users don't care about upgrading once a year, especially if that brings the risk of breaking backwards compatibility. Also in the case of OSX, the fact that each of those incremental upgrades costs a hundred bucks or more certainly doesn't work against Microsoft as all of their point releases are free.

It's not a problem for MS when those small incremental upgrades also come with small incremental breaking changes that cause problems with existing software, which is usually the case (especially with Linux).

That doesn't help Microsoft.

New users haven't upgraded yet. They will look at the current version of Microsoft windows and compare that to the current versions of its competitors. If the competitors are better at the moment of comparison, people may try them instead and that is something Microsoft want to avoid. Once you have started on one platform you are likely to stick with it.

As for incremental upgrades breaking things, I would say Linux isn't any worse than any other platform in this respect. E.g. Oracle designed to work on FC4 or RHEL4 runs just fine on Fedora 7 or RHEL5, Applixware designed to work on Red Hat 6.0 runs fine on FC5 (I haven't tested on any later versions as OOo is so much better). My old HP ScanJet IIc scanner still works on current RHEL5. On windows it stopped working at win2k.

Exactly. Binary-executable-file style backward compatibility is much harder to achieve than is source-code-recompile style backward compatibility.

To achieve the former you have to jump through all sorts of hoops, and put all sorts of kludges in the OS.

To achieve the latter, all that you need is control of the source code.

"How many times have you seen someone plunk down a chunk for XP home, just to upgrade to XP Pro?"

Damn close to never.

Twice, possibly three times here.

So I take it this Windows 7 will be based on that singularity project.

It most likely will not. Singularity is a research project.

I hope so, Microsoft needs to move on and get something from the ground up and stop issuing these half a** patched up Os releases.

There is absolutely no need whatsoever to discard of winnt as the base for Windows. It's well-tested, and fairly portable. What Microsoft needs to do is rethink its userland.