SuperSite Blog

You know, Paul, as soon as you bring up Apple, math, and market share we can expect to see around 100 "Paul hates Apple" comments.  Oh, well.  They are kind of fun to watch.

3...2...1...

It is really sad to hear that. I thought for a second that x64 was actually getting somewhere. I really hope it does.

Don't worry about it, yert. In the case of Vista 64, the outcome is easy to predict. A few (let's say two or three) years from now, 64 bit

will inevitably be the dominant Version.

Just wait for the time that more than 4 GB of memory are Standard in Joe Users PC he bought for 300 bucks in his Best Buy around the corner...

"If you start from almost zero it's easy to triple," said IDC analyst Al Gillen."

Paul, you should realize that it's ok to use that logic when talking about Vista as a way to put its growth in context.

Doing so with Apple is considered completely wrong and is a sign of hate for that company.

I mean, do you really think that these guys in the tech media dont apply a double standard when it comes to Apple? Any journalist that dare use a sentience, like the one from the article, to describe anything  from Apple, would be crucified.

As I'm sure you will be for even making this post.

I will definitely be going to x64 for Windows 7.  I debated with myself to do it for Vista, but decided I did not need it in the end.  There going to be many applications and/or games that use more than 4 gigs of ram in the near future.

I believe x64 will reach 50% market saturation by SP1 of windows 7.  

Mathmatically correct, but practically speaking, people will only continue to buy if the product is good. Now, many if not most advertised PC's in Sunday's ads at CC and BB are for 64 bit Vista. It's the future...today! This is good. If 3rd party hardware makers want you to buy their product, they will be forced to write good drivers and apps if 64 bit is ubiquitous, as it is becoming. So IMO, 64 bit market share has less meaning than what's being offered in product lines. People desire it as it works well and is better. This works on both platforms and will improve computing.Even many average Joe's know about it at work. Life gets easier and faster. Go 64 baby! 64 was my football shirt number too!

Ruh Roh...I hate to do this, but I think you just broke cardinal rule by quoting an 'analyst'. :p

Let's see 64 Bit processors appeared on mainstream desktops, about oh, eight years ago, a couple more years and a MS will have a 64 bit os on its hands that's ready for prime time.  

Keep pushing the envelope MS.

Regards

Joe Dokes

"Let's see 64 Bit processors appeared on mainstream desktops, about oh, eight years ago, a couple more years and a MS will have a 64 bit os on its hands that's ready for prime time.  

Keep pushing the envelope MS."

And the desire for most people to have a 64 bit OS, eight years ago was what?

Um, Joe...

There's just about nothing you got right in "Let's see 64 Bit processors appeared on mainstream desktops, about oh, eight years ago, a couple more years and a MS will have a 64 bit os on its hands that's ready for prime time."

First off, the first "mainstream desktops" running x64 (and we can pretty safely assume that you don't mean Alpha or Itanium desktops are manstream) were based on the AMD Athlon 64 which shipped not "about oh, eight years ago" but in September 2003.

Microsoft's first 64-bit version of Windows was set to go in 1999 (for the DEC Alpha processor) but didn't ship due to DEC killing the Alpha PCs.

There have been 64-bit versions of Windows client operating systems since 2001 (for Itanium) and 2003 (for AMD-64) which is actually a little before the "mainstream desktops" based on the Athlon-64 even shipped since they could be used on specialty desktop hardware based on the AMD server chips.

Pretty amazing that you could get that many items wrong in one phrase.

You're right 2003 not 2001 my bad, that being said five f-ing years for ms to fully support 64 bit, my main point stands.

MS has never been a technology leader.

joe Dokes

BTW: Anybody actually interested in knowing something about modern Windows should know something about its design rather than just spewing what some friend of theirs made up...

I'd recommend going to a good, used book store and tracking down a copy of Helen Custer's 1992 edition of Inside Windows NT. Amazon has 70 used copies available starting at $.01 so it shouldn't be hard to get or too expensive.

Windows NT (which is the basis for XP and Vista) was architected from the start as a 64-bit OS and dialed down to 32-bit to support the existing hardware. (In the same vein, it was architected at the NSA "B-level" for security and dialed down to "C-level" so it would be usable by people ourside of 3-letter agencies)

That architectural legacy is a reason why a PC OS that first appeared in the early 1990s had a 64-bit file system that could handle files up to 16EB...

Gee, joe, you really should read the whole post.

Microsoft had 64-bit XP out before the Athlon-64 shipped. And 64-bit OS out for years before that for other architectures you probably didn't buy.

How much more do you want from them? Telepathy?

Paul, you are truly obsessed.  Absolutely, ridiculously obsessed.  You have 100% jumped the shark on market commentary.  Give up and spend more time on Windows.  After all, Apple marketshare has no real impact on the Windows world, right?  Isn't that your point?  Doesn't that make your obsession that much more pointless?

WOW!!! Papola limited his verbal diarrhea to one paragraph. It's a bloody miracle.

I have had 64bit for quite some time, but out of curiosity what the hell are the Program Files(86)  and then the other Program Files for? I thought it was to seperate 64 bit programs from 32bit but all my 64bit programs install in the X86 folder...

@Weedmonk,

Do you really want us to get on this merry-go-round again?  Paul is a flamebaiter here.  He just keeps ringing the same bell and I'm tired of making the same (correct) rebuttals.  He's entitled to have no clue how to analyze market information.  It's the reason nobody really pays attention to his commentary outside this site's active community.

"won't translate into significant market share gains" - we all know Apple's made significant gains in consumer share, so that statement is just a lie Paul keeps repeating.  Whatever.

Someone said the definition of insanity is repeating the same thing again and again and expecting a different outcome.  Paul's a lost cause, and I don't feel like acting insane.

Enjoy the rest of your weekend!

John:

That insanity definition applies to you. If you answer the same thing everytime Paul talks about marketshare, why are you expecting any other response except for him to ignore you yet again?  Why don't you simply just ignore him? He's not going to stop... it would be insane to believe he would, right?

When I saw the post's title I was looking forward to see Paul's explanation of this long held theory. And what do we get?

"I've repeatedly mentioned ... snip...that Apple's phenomenal year-over-year quarterly growth in Mac sales won't translate into significant market share gains."

Absolutely right. But how does that explain how easy it is grow when you are small? Answer: It doesn't

Then we get the HP vs Apple comparison. Why not just use the real figures Paul,

In the last quarter the whole PC industry grew by 16%

HP grew at 17%. Just 1% better.

Apple grew at 41%. That would be.... er .... 24% better.

" It’s simple math" Yes but nobody here has told me why it's so "easy".

@DRWAM  yes in my local adds it was all Vista 64, most with 3gigs of ram (so much for dual channel) and on top of that its back to school weekend in the state I live in so no tax on computers purchased.

So joe user goes into the Bestbuy, does not want a mac and so joe gets a PC....Vista 64bit.  I guess if joe really knew he was getting 64bit and wanted 32bit he would have to have a geek squad person to install it at a cost, so I am going to say its safe to say that joe is not really choosing 64bit Vista he is choosing a PC and that is what he gets unless he wants to pay $$$ to have a 32bit version put on it.

Pretty easy to grow those Vista 64bit numbers once vendors stop giving you a choice at the retail outlets.

Anyhow Paul...........what happened to actually writing something interesting.  This is serious yawn material.  I guess the Microsoft employees here will dig it, but brother this kind of stuff is getting boring and fast.  

I mean how about writing an article about how the sun will up tomorrow, and HP will sell more PC's than Macs, and Apple will sell more iPod's than Zune...obvious 101.

Whats the advantage of 64-bit Windows vs. 32-bit Windows?

Pingback from  Lost Drive Blog » re: It's easy to grow when you're small

@ggolcher,

Did you actually READ my posts?

@Ocean, the OS can access more than 4gigs of RAM, or between 3-3.5 gig is what you see most of the time with a system with XP or Vista 32bit on a system with 4gig.

I am sure someone will come hear and tell you its faster at times.  Which could be true but I bet it depends on the app, and most of the time only a bench mark would tell you its faster.  I have read more than a few comparisons and they are mixed.  Some cases faster, some no difference, some actually slower.

I would say its useful for about 2% of applications.  That said 64bit is the future and MS should have just dumped 32 bit...hopefully Vista R2 or umm Windows 7 will be only 64bit.

@ guys wondering about x64 growth

I read Ed Bott's post about x64 versions of vista on HP's(i think?) website I looked at the Sunday paper Circuit City insert.  I was extremely surprised to see that about half or more of the computers in it came with x64 versions of Vista.  More than a couple had 5 gb of RAM.  And the funny thing was these things were cheap.  Under $1000.  I knew RAM was cheap but WOW.  If that many mainstream pcs come with x64 then there's going to be a lot of x64 pcs come December.

Snake

Dumping 32-bit at this time would be foolish. You're already saying it only benefits 2% of applications (I'm assuming that's just a wild guess but let's go with it for now). You're saying that it would be a good idea to sacrifice backward compatibility and incur transition inefficiencies for 98% of the apps to benefit 2% which can already get that benefit by those users going 64-bit. (Personally, I think the apps that benefit is more than 2% but that's also a guess)

There is an inherent cost to running 32-bit apps in a 64-bit OS in the same way that there was in running 16-bit apps in a 32-bit OS. That cost is offset in many ways but whether it will or won't be better overall right now depends on the mix of apps and the system.

Additionally, Microsoft cares about backward compatibility. To break with no transitional period would require that older hardware would be stuck in the older systems without the other benefits of an updated OS.

Since 64-bit has been available on Microsoft client operating systems since Windows XP, the choice of 32-bit or 64-bit is now up to the system builder and the consumer. To my mind a much better choice than dictating a mandatory transition and leaving users out in the cold.

Now, on the server side, you've already seen server apps that have gone 64-bit only. In those dedicated use machines which already push the hardware, it makes more sense since the benefits to that one app are solidly known.

uhh. iguess i should've read all the comments before posting everyone else already said it.

runner

There's no question that 64-bit is taking off. When it will pass 32-bit on new systems is still up for speculation but I'd say we'll see it within the year.

Now, that won't mean that the market share of 64-bit will pass 32-bit because of all that existing 32-bit Vista and 32-bit XP out there but those systems will age out in the next few years and we'll certainly see no more need for a 32-bit client OS in a few years in the same way that we no longer saw 16-bit client OS sales several years after 32-bit platforms came online. In fact, more likely the transition wil be faster since the architectural changes are less major since the Windows NT architecture wasn't tied to a single architecture the way that Windows 3.x and 9x were.

2% meaning there are not many consumer/workstation applications that need more than 4gigs of RAM, and I would bet that not many would even use it.  Heck most apps still wont use more than 1 core on the consumer side of things.

However probably 98%of mainstream 32 bit apps can run on Vista 64bit....a guess on my part.  Vista was a loooooooooooong time coming so everyone would have had a long warning had they chose to announce early on 64bit only.  They had 2003/XP 64, they should have just dropped Vista as 64bit only, same for 2008 and phased out XP maybe 6 months later, end of this year for the 32bit crowd  Hardware has been 64bit ready for some time now.  Of course the Anti-MS crowd would have eaten them up in the press.

Either way I guess, I mean it looks like the fall season, which is the biggest time of the year for PC sales will see 64bit go mainstream if vendors keep it up, I can see why not.

@runner, RAM has been stupid cheap for at least a year now.  The wife just got a new iMac (her first mac) and I picked up 4gigs of Crucial Memory on newegg for $91.

Mike,

Yeah, It really surprised me though how many in the ad were 64-bit.  Hopefully the consumers never know they are running 64 bit and the transition is almost transparent to them.  All they need to know is that "Wow I have 5 gigs of RAM"  That would be the goal wouldn't it.  Most people I know dont even know what version of windows they are using.  I think it will be decently smooth but time will tell, of course.

One other thing.  I've sorta been wondering for the last couple days.  How well do 64 bit Vista/linux etc.  run on the first generation of 64 bit processors(athlon 64, pentium 4 64 bit models)

Snake,

Yeah I have 4 gigs in my computer and it was $80 without the rebate and 50 some with.  

Maybe "Easy to grow" is the wrong phrase...how about if I were to say "A company with fewer customers can post a higher percentage growth by selling fewer of a given product than a company with more customers"? Would that make it more accurate? I believe that is what Paul is trying to say...he is just saying it badly. I am pretty sure that all of you that have been harping on this know that already, though.

Paul - can you tell us all if I am going in the right direction here? It seems that many of your readers are waiting to hear from you on this.

--tayme

Runner

It isn't so much the generation of the processor, it's been the lack of drivers for peripherals that's been the slowdown. There's very little motivation for a company selling a printer to write a driver when their business model is to give the printer away and sell the ink. They really don't want to invest in a product that's being replaced or has been replaced when they feel they're already subsidising the product.

There's been a push to get new drivers done and a big push to make sure that apps and drivers are tested for 64-bit as well as 32-bit over the last couple of years and I suspect that we're seeing the results of that push now in the mainstream adoption of 64-bit versions of Vista on average grade computers.

Of course, to return to the question, there is sometimes a lack of drivers for motherboard components in the early chips. It's not that the early 64-bit chips don't work but the onboard hard disk controller may not have a driver or the onboard USB driver, etc.

Snake

The multicore/manycore issue is actually a bigger one than the 64 vs 32 bit one.

Vista (like any modern Windows) will allocate threads across the different cores as needed to balance the load. That's not the problem. It's rare that you'll see one core in a dual core sitting idle while the other chugs along under load. Even if an old single-threaded app is hogging one core, all the system processes and background threads from the OS will be migrated to the other core.

The problem is that writing apps that take advantage of multiple cores efficiently is a hard problem and one that very few programmers know how to do well. And I mean very few. I'd be surprised if 10% of the programmers working in the industry today could write really good multithreaded code. And this is a huge problem facing the industry. Since we've peaked at processor speed we're seeing the push for faster systems being answered by multi core systems. When I was scanning the mainstream computers today in response to this thread I not only noticed a lot of 64-bit Vista, I also noticed a lot of quad-core processor systems. That's becoming mainstream. And next year, I'd be surprised if the average home system sold had less than 8 cores. Writing an app to deal with not only threading but processor affinity and stalling and race conditions is hard.

I know both Microsoft and Apple are working hard on tools to make this easier (Microsoft has some in beta now) but it'll be a while before it's either easy enough for the average professional programmer as they are today or for the average programmer to gain the needed skills. And many major applications have not been architected to make transparent multithreading something that wil be easy to do without redesigning the app from the ground up.

At any rate, that's the crisis that's really looming.

Oh, and one more note. I've seen people criticize an application for running almost entirely on one core rather than loading down all the cores on their system. They thought this was inherently a bad thing. It isn't. Splitting the work is often best done by keeping everything on one core. The overhead of synchronization or cache invalidation can make splitting into to many threads as bad a thing and not splitting them enough. Some problems break up well, some not that well, some not at all.

OT:

The NYTimes gives air to some of the Mojave criticisms:

www.nytimes.com/.../04vista.html

Ocean

OT: Why don't you go and put it on one of the existing Windows "Mojave" threads where it belongs rather than attempt to hijack an unrelated thread? Since we'll assume you'll do that, we can ignore it here and keep it in context.

You know, not to say too much nice about Apple, but when you buy Leopard it's both 32 and 64 bit and you get the right one for the machine you have, all automagically and behind your back.

"Someone said the definition of insanity is repeating the same thing again and again and expecting a different outcome. Paul's a lost cause, and I don't feel like acting insane."

What's truly ironic is that he thinks he's "bringing some sort of balance into the media", which, through his glasses, is so "pro-Apple and anti-MS".

So what does he do? He decides to compare the adoption rate of Vista with that of Mac OS X, because Mac is small and Vista, when it came out, was "small".

Vista, of course, has had a market share of nearly 100% of the pc market the day it came out because pc's just happen come with Windows preinstalled, so calling it "small" market-wise at any point in history makes this probably the stupidest thing I've ever read.

To me it's all quite entertaining, but don't any of you winfans feel the least bit embarrassed reading this stuff?

Pingback from  Six Sense Blog  » Blog Archive   » re: It's easy to grow when you're small

Come on Mum we all know that no-pro Windows backer here is ever disingenuous *dons flame suit*

John Papola, the fact that you proffer that all of your rebuttals are correct illustrates your arrogant buffonery. The answer is simple, go elsewhere if you don't like what is written.

@mikegalos - "Why don't you go and put it on one of the existing Windows "Mojave" threads where it belongs rather than attempt to hijack an unrelated thread?"

Isn't that what you have done here...this thread was regarding the statement that it is easier for a small company to grow...not 64 bit or multi-core processors. Just being fair, here.

--tayme

“The problem is that writing apps that take advantage of multiple cores efficiently is a hard problem and one that very few programmers know how to do well. And I mean very few. I'd be surprised if 10% of the programmers working in the industry today could write really good multithreaded code. And this is a huge problem facing the industry. Since we've peaked at processor speed we're seeing the push for faster systems being answered by multi core systems. When I was scanning the mainstream computers today in response to this thread I not only noticed a lot of 64-bit Vista, I also noticed a lot of quad-core processor systems. That's becoming mainstream. And next year, I'd be surprised if the average home system sold had less than 8 cores. Writing an app to deal with not only threading but processor affinity and stalling and race conditions is hard.”

Overall, what an intelligent post (except the 8-core comment for next year; I don’t think quad is mainstream yet, so maybe it will become so next year, I think even 8 logical units is pushing it for mainstream yet – quad+HT – and I think 8-core mainstream is 2-3 years away at best).

It’s funny how many “commentators” (read: dumb hack bloggers; read: Robin Harris - blogs.zdnet.com/storage) get it oh so very wrong.  He has no clue that OS can already load up the cores, and thinks you can take inherently unparallel code and somehow spread it over the cores at runtime efficiently.  I wish you would post there, so that the [idiot] (I apologize for the rude treatment of the said individual, but I actually asked him to define what the heck he was talking about and he still couldn’t) does some research.

And here’s the kicker.  Wouldn’t you agree that a lot of common computing tasks do not thread well and forcing threading may actually incur you a performance penalty?

Mike, no one is reading those threads anymore.  I put it where people will see it.  We want both sides out there, right there?

OT:

The NYTimes gives air to some of the Mojave criticisms:

www.nytimes.com/.../04vista.html

@Ocean - Keep in mind, that those old Folgers, and even the new Pizza Hut pasta adverts were and have been successful. Along with that MS has stated that this is not even a part of the advertising campaign that they have planned. Even if it is, it is just as valid as Apple touting their own product, not be demonstrating it, but by "poking fun" at its competition. See my post in the thread regarding Mojave and you will see that it is possible to get a thorough "test drive" of Vista at no cost, for up to 120 days.

--tayme

lehenbauer

When you buy a retail copy of Vista (full or upgrade) you get both 32-bit and 64-bit as well.

tayme

It's interesting to note that the "Pepsi Challenge" ads were done by John Sculley who was later hand picked by Steve Jobs to run Apple.

Sculley quadrupled Pepsi's ad budget and took Pepsi from being the number two cola brand to being the number two cola brand. :-)

It's interesting that, at least according to a widely circulated anecdote, when Scully took the test, he preferred Coke.

ibarskiy

If the industry pattern follows, I suspect we'll see 8-core processors by fall in the "Extreme" chips and they'll go mainstream by holiday 2009. But the failure of AMD to keep up with Intel may stall that out a year.

As for your last question. Absolutely (I thought I said that). There's no question that there are lots of things that do not get any benefit from parallelization and don't split out well so attempting to multithread them not only gives no benefit but incurs the overhead so overall they go slower.

ibarskiy

Wow. I just read that article and don't even know where to begin.

It takes real talent for a supposed professional to take multiple press releases from multiple sources for technology he doesn't begin to understand and first off have the nerve to think he should editorialize on it and second to get virtually every sentence wrong.

You'd think he'd have learned at least a little about the subject after reading those press releases.

btw: you might want to take a look at Microsoft's TPL (Task Parallel Library) and PLINQ (Parallel Language INtegrated Query) technologies. They're still in pretty early beta but they've done some fascinating work on parallelization for devs not in that very top tier who can do thread sync well on their own.

OK, I couldn't resist and did reply to one paragraph of that insanely bad ZDNet article. I'd have written more but I despise the "reload the ads to see each reply" interface that they use.

Here's what I said...

---------------------------------

Wow. It's hard to write a single article with so much inaccurate content. The mind boggles.

Let's look at one early paragraph just to keep it simple since correcting every single thing wrong here would take more space than the original article itself.

Robin Harris said:

"Apple is clearly a leader in implementing multi-core support, beginning with the first dual processor Power Macs 5 years ago, while the DayStar multi-processor Macs date back to the mid-90s."

1) The first Microsoft OS that had multiprocessor support (and multicore is a subset of that) was Windows NT 3.1 in 1993 and dual processor Windows NT boxes were available to buy at release.

2) The DayStar licensed Macintosh clones which were available for about a year before the System 8 contract busting killed both them and DayStar were based on a DayStar proprietary premptive scheduler that ran underneath Mac System 7. To have a multiprocessor aware app it had to be written to the DayStar APIs and apps written just to the Macintosh operating system got no benefit. In fact, all Macintosh processes including the OS and all the cooperative multitasking apps ran on one processor and any DayStar tasks you spun off were the only things that could run on the other processors. After DayStar was killed, that technology died with them and Macintosh remained single processor and cooperatively multitasking until OS X in 2000.

On point worth mentioning if you feel like annoying people on the ZD thread would be this comparison since designing for a few cores doesn't really cover all you need to learn to do dozens:

Largest commercial OS X multiprocessor machines

2 sockets for a total of 8 cores

Largest commercial Windows multiprocessor machines

64 sockets - HP Integrity SuperDome series

(32 socket servers aren't unusual)

And to add another data point into the mix although the loosly coupled HPC model really solves different multicore issues than things like TPL/PLINQ or Robotics Developer Studio...

Largest Windows Server 2008 HPC Cluster

9,600 cores - NCSA "Abe" supercomputer

(currently ranked as the 23rd fastest supercomputer in the world and running on a beta release of commercially available Windows HPC software)

Mike, who do you write for?

Ocean

Right now I'm not. In the past I've done some tech magazine articles (Robert Scoble was once my editor back when he was just starting with Fawcette), did a couple of books and a few dozen courses.

Why?

@mike

When I was asking about the performance of the original athlon 64's etc. I was just wondering how they performed with 64 bit oses.  I guess it would be a nightmare finding drivers for those old mb's.

"Largest commercial OS X multiprocessor machines

2 sockets for a total of 8 cores

Largest commercial Windows multiprocessor machines

64 sockets - HP Integrity SuperDome series

(32 socket servers aren't unusual)

And to add another data point into the mix although the loosly coupled HPC model really solves different multicore issues than things like TPL/PLINQ or Robotics Developer Studio...

Largest Windows Server 2008 HPC Cluster

9,600 cores - NCSA "Abe" supercomputer"

-------------------------------------------------

Keeping it real in the multi-core world.  My day job is a Windows System Engineer for a LARGE company.  We have more than 14,000 Windows server to give you some perspective.

We have ZERO with more than 8 cores.  We have a few with 32gigs of memory.....a few.  We have a small, medium and large build, from 2 vendors, small = single dual core, medium = 2 dual core, large 2 quad core.

Only SQL, and Exchange servers run on the "large" windows servers.  I just built a new SQL cluster last week, 2 64bit 2003 R2 "large" servers with 16gig of RAM each.  

I would say the "large" servers make up about 10% with most servers at the small level, single CPU 2gig of ram for file, print, Domain controller, SMS servers, IIS servers.

Even on our ESX 3.5 servers we max out at dual quads with 64gig of RAM.  We are testing ESX on some servers with quad/quad 16 cores, with 128gig of RAM.

So by in large the vast majority of servers I have seen in my years of supporting Windows servers from NT 3.1 until today are of the 4 core, 4gig of RAM or less.

So long story short if a customer was serious about Apple server products, their current products have enough power for almost any job.

At the workstation level they are equal basically, looks like the Dell is more expensive....

store.apple.com/.../A

vs

configure.us.dell.com/.../config.aspx

I think, not sure in any way, that the author of that blog post meant that more consumer/workstation apps for OS X are multicore than in the Windows world???  I mean I am sure Final Cut Pro, which is used by something like 60% of movie makers can use the full power of a Mac Pro and nothing hammers a CPU like video work.

It seems like you mention one thing positive about Microsoft and people are ready to attack you. So Microsoft continues to find positive things to boast about with Vista, and someone for the love of God can't let them enjoy that. Its getting rather pathetic that we can't talk about something positive about Microsoft. Heaven forbid we find something good and positive to talk about.

I will definitely migrate to Vista 64 when I can. For the meantime, I'm enjoying the 32 bit version with still no issues. It has been some of the best computing experiences I've had.

Good article, Paul. Thanks for finally debunking the myth with facts. As for the rest, hate it or love it, Vista's here to stay.

Peace.

Snake

Lots of little servers is certainly one way to go. It is not, however, the only way.

For example, I was involved a while back in a very large database job that needed most of 32-processor SuperDome to run at a reasonable speed. (Yes, it would have run on a smaller box but the data would have been obsolete by the time it was done) And before the MS bashers kick in, switching this job over to SQL Server set new speed records on the same hardware from previous high-end SQL engine.

The point isn't that 32 or 64 proc boxes are the majority case right now, it's that they exist out there in significant numbers, are absolutely needed for some types of work, and Windows supports them very well.

BTW: With the rise of virtualization in the server space, don't expect to see 14,000 small, dedicated servers as the rule much longer. Microsoft, for example, has been shifting vast numbers of their small servers over to Hyper-V images for a while now and the savings have been huge as has been the growth in flexibility in shifting server loads around as demand changes.

Oh I know there are big servers out there with out a doubt.

I would say most of them are UNIX but there are some big Windows boxes as well.  It just that its not the "norm" by a long shot.

Yes visualization is rapidly ending the life of 95% of small Windows servers at my place of work.  We just replaced some 250 Windows 2000 IIS/web servers in a big farm, with an 8 way ESX cluster running 220 Windows 2003 IIS VM's.  So from racks of old IBM blades down to 1 rack of 8 DL580's.  If fact you need a reason not to VM and it needs to be approved.  With this economy and the massive drive to be GREEN, visualization is very popular these days.

Snake

I didn't say that big servers were the norm, just that they weren't unusual (at least up to 32-proc)

The real problem with virtualization that's showing up in the field these days isn't a lack of adoption, it's that it's so easy to move physical servers over to virtual that people are doing it without doing the planning and doing it right is not as easy as just doing it at all.

I agree, you cant just VM anything.  

I/O of a VM candidate is the key in regards to if it can be VM'ed and if so knowing its load on a VM host.  That and if you use VMware putting like OS'es on the same host is key as ESX can really maximize memory if all of the guest OS'es are same OS like all Windows 2003 32bit or whatever.  Mixing up different guest OS'es, Windows 2000, 2003, 2008, RHEL on the same host will negate this feature and limit the capacity of your hosts.

I think we can all agree about one thing:

32-bit applications aren't going away, just because of 64-bit adoption.

The key thing here is that Windows Vista x64 offers 32-bit compatibility via WoW64.

While nobody will dispute the fact that 32-bit compatibility is necessary for probably the next 5 years AT LEAST, there really isn't a reason to move forward with x86-32 "native" versions (I say native in quotes because even the x64 versions of Windows run x86 code natively - there is no emulation performed).

So WoW64 should still be included for the foreseeable future.  Native 32-bit versions aren't necessary on new Windows versions though.

I'd like to reiterate my previous statement:

Until Intel or AMD (more than likely either Intel or NVIDIA realistically) releases some new 64-bit-exclusive platform architecture for mainstream computing (think Itanium for home users), x86 will be synonymous with 32-bit application development.

Okay, I'll throw in some quick two cents.  Heh.  Why not.  I sent this (with a few tweaks) as an email to paul:

Here’s the problem.  You're assuming that it requires the same effort for Apple to sell an additional 250,000 computers as HP (or Dell or any PC vendor).  This is an incorrect assumption for the same reasons why you find it hard to change away from google for search. (a reference to a statement he made in the latest podcast)  Change is hard for people.  Very hard.  That's the main cause for grief in the Vista migration.  Not that Vista sucks. Just that Vista's got a lot of changes to the user experience.  And change is HARD.  Convincing people to change is even especially hard.  NOT EASY.  In a saturated market of Windows PCs, it’s far “easier” for a PC vendor to drive marginal growth by stealing users from other vendors than it is for Apple to steal the same users given the operating system change.  Apple's change case to a Windows user is a much harder sell.  They need to basically start over with new software and hardware.

So your simple math is correct but the accurate way to phrase this would be “the same increase in absolute marginal sales will result in a greater percentage improvement for the firm with a smaller base of sales”.  That’s unfiltered, unbiased, objective truth.  

But saying it’s “easy” is an unproven, untested opinion.  So you’re qualitative analysis is dead wrong and not defended by your argument.  It’s not “easy” for Apple specifically to add new customers faster than Windows vendors can when it involves convincing people to switch operating systems.  It's certainly not "easy" for apple to grow their business as they have been from existing Windows users.

Now go ahead gang.  Try to explain why I'm a crazy apple fanboy.

Mathematically correct, easy no. If so, Palm should be dooing great, since they sell low numbers. Heck anyone with small sales should do great, but they don't. There needs to be a market and people who want to buy. That said, the computer market is big and many buy, but they buy what they want, so if they are happy with Windows, then why switch. Two days ago, I went to the ATT store to get my wife a new phone and the rep there said Vista sucks [he uses XP]. I told him that I found it to be fine, and he replied that I was the first to tell him good things about Vista. He never tried it! [Yep, last week I posted that I had not recently heard bad things about Vista, so this today]. So I agree JP.

XP is still the preferred OS. www.infopackets.com/.../20080801_windows_xp_still_outselling_windows_vista.htm