This is for Ed M.

[js33]

Hi Ed,

Just thought you would like to know that 64 bit windows is almost ready for AMD Opteron.

64 bit windows for Opteron (http://www.gamepc.com/labs/view_content.asp?id=amd64xp&page=1)

You did have a point about drivers but those are being worked on as well. Microsoft has a lot of 64 bit drivers ready and are working on more. I imagine this will be an issue for OSX 64 bit as well. But you were so set on 64 bit Win not being ready for AMD until what like 2008 or something. :D

Cheers,
JS

[dfc]

I've heard reports that this OS was already beta...that must not be true if the link you gave was accurate (he's using the latest build as of 9/5/03 and that's a pre-beta build.

But, if they plan on shipping it sans much support/drivers..etc...that will certainly cut down on the testing time :)

Interested to see if they can actually storm thru the beta test of MS first desktop 64bit OS in a few months.

Especially considering how long the other MS OS revisions were in develpoment (beta, etc).

Is that good or bad?

[js33]

Well it's coming along pretty fast. Maybe it was already in Beta by the time he finished his test with the Prebeta and posted his article.

Ed was predicting we wouldn't see 64 bit Windows for Opteron until the end of the decade. :D

But it will probably have sketchy 64 bit driver support until it gets adopted widespread. Also the need for 3rd party drivers is minimized if it's like XP. XP has drivers for almost any hardware imaginable included. Of course most of them were developed by the 3rd parties for their hardware but MS does a good job of including them all on the XP disk.

Cheers,
JS

[Ge4-ce]

I think that what Ed ment is "when will there be a 64 bit Windows version available?"

Not when is there a version of windows available that supports the AMD 64 processor.

You can probably compare this version of Windows with Mac OS 10.2.7 and Panther. It takes notice of the 64 bit and can handle stuff but the system itselves is not 64 bit..

For Apple we also have to wait until ... ?

10.4? 10.5 for true 64 bit support

[Ed M.]

My timeline arguments start here:

http://vbulletin.newtek.com/showthread.php?s=&threadid=9600&perpage=15&pagenumber=29

The first key excerpt from the Microsoft Project Manager was stated here:
http://www.eetimes.com/sys/news/OEG20030320S0053
(Also in my original post)

That comment was made on March 20th, 2003 (this year). I doubt that things have changed much in 5 months. So, cutting down the time frame of release from years to months seem dubious. Why would the project manager lie?

It's also worth keeping in mind that Microsoft isn't likely to release a 64bit desktop version to support Athlon64 (Opteron's little brother that appears to be stillborn) before they release a version to support Opteron64. Besides, it's reasonable to assume that the Opteron 'SERVER' version will cost just as much as the Itanium version. They start at around the price of a dual-processor G5 and you won't be able to run any apps current apps on it that will take any advantage of the 64bit features. It's simply not built into AMDs design. You'll need a 64bit OS and 64bit apps (and drivers).

What makes anyone think that there would be so much support for this thing if Itanium hasn't gotten anything meaningful at this point (and Itanium has been out longer!)?

And it still keeps people strapped to the aging x86 architecture for much, much longer.

In a recent August 4th article of this year (2003) here:

http://zdnet.com.com/2100-1104-5058990.html

Microsoft makes it clear that the server project is in a severe state of flux. From the article:

But, once again, Microsoft's plans for a new server version of Windows are in flux. Last November, Microsoft said it was scrapping the server version and would move straight to the next generation version, code-named Blackcomb, in 2006.

Earlier this year, Microsoft said Longhorn would ship in 2005, but now the company is saying only that Longhorn is a major effort and will ship when it's ready.

So, it's very clear that Blackcomb will ship AFTER Longhorn. Microsoft said so just 4 weeks ago. It's now Sept. 9th, 2003?

Are we to believe js33 when he says that a release is imminent? Again, I find it dubious that the 64bit version for Opteron will ship much earlier than both AMD and Microsoft officials have stated.

Remember, it's not *me* who's saying this stuff. It's AMD and Microsoft representatives. I'm just here piecing together the timeline. Other can do the exact same thing given the information that AMD and Microsoft has provided us with in the forms of quotes from articles and interviews.

An article written on May 7th, 2003 reinforces a previous article about the ship date for Blackcomb:

http://www.eweek.com/article2/0,3959,1061875,00.asp

Now, remember, according to the information provided by Microsoft and AMD representatives, the timeframe now looks like the following:

- Blackcomb is coming AFTER Longhorn
- AMD64 support is coming AFTER Blackcomb
- Longhorn was just delayed *again*
- Desktop version coming AFTER server version

MOST RECENT:

Sept. 2, 2003 (Windows Longhorn Postponed Yet Again)
http://www.macobserver.com/article/2003/09/02.4.shtml

Sept. 1, 2003 ('Longhorn' Rollout Slips)
http://www.eweek.com/article2/0,3959,1234525,00.asp

So where does that put the earliest release date for the *server* version of Windows64 for Opteron? What apps will be there? Are we to believe that AMD and Microsoft are lying about the release dates? are they years ahead of schedule? What would this OS cost if it were available for Opteron?

Other things to keep in mind:

- Athlon64 not out yet
- Windows64 for desktop coming AFTER server version
- Windows64 for desktop (Athlon64) not out yet
- Opterons are the second slowest-selling processor on the market behind Itanium.

You guys can to the math and figure out the timeline for yourselves. I've backed-up and supported my conclusions and I expect that others will reach the same conclusions given the information that's been stated in public by Microsoft and AMD representatives.

--
Ed

[toby]

from JS33's link :
" Of the companies who have taken on the challenge of moving to 64-bit thus far (Intel, Sun, and Alpha),"

IBM?

"AMD's goal is to succeed where all of its competitors have failed"

IBM?

[DaveW]

Ed, in the first article you posted the MS exec just said that x86-64 support would come after the 32bit Win2003 server and BEFORE Black Comb, not after. An April press release from MS (http://www.microsoft.com/presspass/press/2003/Apr03/04-09AthlonOpteronPR.asp ) says they are making a Windows XP x86-64 version (which is in beta now). MS would not release Longhorn and Blackcomb and then release another version of WindowsXP, it just doesn't make sense. It would be like them developing a P4 optimized version of MS-DOS. Which link said that the desktop version of Windows64 was coming after the server version? As far as I can tell, that is an assumption of yours that you are stating as fact.

[js33]

Originally posted by toby
from JS33's link :
" Of the companies who have taken on the challenge of moving to 64-bit thus far (Intel, Sun, and Alpha),"

IBM?

"AMD's goal is to succeed where all of its competitors have failed"

IBM?

Toby,

I think he was only speaking of PC hardware. SUN had/has a version of Solaris that runs on Intel hardware.

Cheers,
JS

[js33]

Originally posted by Ge4-ce
I think that what Ed ment is "when will there be a 64 bit Windows version available?"

Not when is there a version of windows available that supports the AMD 64 processor.

You can probably compare this version of Windows with Mac OS 10.2.7 and Panther. It takes notice of the 64 bit and can handle stuff but the system itselves is not 64 bit..

For Apple we also have to wait until ... ?

10.4? 10.5 for true 64 bit support

Windows XP 64 is 64-bit and also runs 32 bit software on the Opteron.

Cheers,
JS

[js33]

Ed,

Well I've read your timeline posts. But things change. Do you really think these big companies are just going to sit on their hands and do nothing? Of course it will take awhile for everything to go 64 bit but that doesn't mean there will be no support for AMD for 5 years. That link I posted already shows a 64 bit version of XP. It will probably be ready when the Athlon 64 comes out.
Also keep in mind that desktop OSs are usually less than $200 and the server OSs can cost several thousand. Server OSs usually have support for 32 CPUs or more. These are 2 totally different markets. Noone running a desktop CPU will ever buy the server version of the OS unless they happen to have a 32 processor machine laying around. XP right now comes in 2 versions. Home and Pro with the main difference being the Pro supports 2 CPUs and Home only 1. The Home version is about $89 and Pro about $189 so that pricing scheme is likely to remain about the same for 64 bit (Maybe a little higher initially but not much).

Cheers,
JS

[Ed M.]

Like I stated earlier, the information is all there. Do the math. Everyone should read through my links and see if I've added things up incorrectly. Remember, the first post implied 64bit for Opteron, not Athlon64.

If there is no 64bit OS for Opteron yet, how can we expect there to be a 64bit *desktop* version for a processor that hasn't even been released yet? Same goes for apps. Furthermore, using the posted links with the pertinent information, one has to wonder what the *win* would be for developers. No one has migrated to Itanium yet and that's been out for a loooooong time. I think AMD64 is falling into the same category. I posted the link with the sales figures and it's pitiful.

You see, what you've fail to realize is that big companies like Dell, HP, Intel, Microsoft and others that rely on the *Windows* market are dependent on those "BIG sales" to sustain their revenue stream. If the sales aren't there, products and sectors will be ignored. Judging from the recent figures released on Opteron sales, it doesn't look encouraging at all. Developers are disheartened and less inclined to offer support. The exact reason a lot of PC people like to hold in the face of Mac users.

So, the question is, what percentage of the 95% of users already on Windows systems are eager and willing to shell out the $$ for a new AMD64 platform? On the same note, what percentage of those users will have a definite, significant and actual *need* for AMD64 systems that the current 32bit Wintelon systems can't already provide?

We'll see.

--
Ed

[Ed M.]

Another interesting AMD article that everyone might be interested in:

http://www.digitimes.com/NewsShow/Article.asp?datePublish=2003/09/04&pages=A7&seq=42

I mean even if these systems were to be released tomorrow, next week, next month.. 6 months... What are users going to run on them? Nothing that will take advantage of the 64bit capabilities. They might run Linux, but there isn't a significant number of apps that a large majority of users might use in the various submarkets. It still leaves AMD competing with Intel in the 32bit Windows world and if their machines can't show a big enough advantage over the Intel offerings, it's likely that they aren't going to bother creating a completely new app for the *same* Windows platform.

It's starting to add up. Just watch. Key delays are falling into place.

AMD and Microsoft stated that 64bit support would come AFTER Blackcomb ships, but Blackcomb is to ship AFTER Longhorn (which was delayed again) AMD releases Athlon64 chips as AMDXP processors with no 64bit support essentially making them 32bit chips. What's more is these systems will for the large majority of things be running on Windows-32bit making them nothing more than the same 'ol 32bit systems that are currently being sold in droves to users. People and companies that purchased 32bit systems and 32bit software 2 years ago aren't going to scrap them and suddenly rush out and buy these new systems; especially when said systems are likely more computer than they already need, not to mention the fact that there is no native support from Windows or Windows apps yet.

It would seem that there's going to be slow adoption for AMD64 systems.

--
Ed

[mlinde]

Originally posted by Ed M.
So, the question is, what percentage of the 95% of users already on Windows systems are eager and willing to shell out the $$ for a new AMD64 platform? On the same note, what percentage of those users will have a definite, significant and actual *need* for AMD64 systems that the current 32bit Wintelon systems can't already provide?

Just pushing your worlds back at you, to make you think. What percentage of the 3-5% of users already on Macintosh systems are eager and willing to shell out the $$$ for a new G5 platform? On the same note, what percentage of those users will have a definite, significant and actual *need* for G5 systems that the current G4 systems can't already provide?

Real-world feedback suggests that the vast majority of Mac owners still use OS 9 as their primary OS. Although the suits at the mother fruit want us to believe that everyone is embracing OS X, walk into your local print/pre-press shop and see if they are running X or 9. 90% of the time it's 9, and they aren't buying G5s. How about the local school district? Many of them are still running Performas with OS 8.5 (at best), if they haven't been swayed by the Dell/Microsoft greasing of palms.

A friend of mine owns a small digital print business, and he's not planning on buying a single G5 in the next fiscal year, and in his staff, only one employee is shelling out the cash for a personal G5 system. Apple needs a compelling argument to get people to double-switch, from OS 9 and older machines to OS X and the G5. A TV ad with a guy smashed against a tree outside his house, with inaccurate information isn't that argument.

[Ed M.]

Just pushing your worlds back at you, to make you think. What percentage of the 3-5% of users already on Macintosh systems are eager and willing to shell out the $$$ for a new G5 platform? On the same note, what percentage of those users will have a definite, significant and actual *need* for G5 systems that the current G4 systems can't already provide?

That's funny considering that well over 100-THOUSAND systems have already shipped. That more than the 5000 thousand Opterons that shipped since it was introduced. lol Way to funny, but nice try.

The G5s have been a hit!

--
Ed

[js33]

Originally posted by mlinde
Just pushing your worlds back at you, to make you think. What percentage of the 3-5% of users already on Macintosh systems are eager and willing to shell out the $$$ for a new G5 platform? On the same note, what percentage of those users will have a definite, significant and actual *need* for G5 systems that the current G4 systems can't already provide?

Real-world feedback suggests that the vast majority of Mac owners still use OS 9 as their primary OS. Although the suits at the mother fruit want us to believe that everyone is embracing OS X, walk into your local print/pre-press shop and see if they are running X or 9. 90% of the time it's 9, and they aren't buying G5s. How about the local school district? Many of them are still running Performas with OS 8.5 (at best), if they haven't been swayed by the Dell/Microsoft greasing of palms.

A friend of mine owns a small digital print business, and he's not planning on buying a single G5 in the next fiscal year, and in his staff, only one employee is shelling out the cash for a personal G5 system. Apple needs a compelling argument to get people to double-switch, from OS 9 and older machines to OS X and the G5. A TV ad with a guy smashed against a tree outside his house, with inaccurate information isn't that argument.

Yeah the argument that the AMD64 will have a slow adoption applies to the Mac as well. Not to mention the 3% market share so even if there is a fast adoption it won't be that significant.
As Milde says about most print people still on OS9 I can confirm that as well since most of the ones I know use Quark for a living and Quark just released an OSX version. Most of them couldn't switch to OSX even if they wanted to until very recently. Also how many Apple apps are 64 bit now? How long will it take and how big will the market be? The only one Ed has mentioned was Photoshop and that it had a plugin to make it 64 bit which still makes you wonder how much of the app really is 64bit.

I think it will take awhile on either side with the only ones adopting it early are the ones that always adopt new technology early.

Also the only compelling advantage is the ability to address more memory which is nice but how many people are even using 4 gigs right now when most computers only support 2 gigs anyway?

For me to switch the more compelling reason will be speed which isn't a guaranteed advantage just because it's 64 bit.

I'll sit back and wait for a lot of the dust to settle as I'm sure most people will as well.

Cheers,
JS

[js33]

Originally posted by Ed M.
That's funny considering that well over 100-THOUSAND systems have already shipped. That more than the 5000 thousand Opterons that shipped since it was introduced. lol Way to funny, but nice try.

The G5s have been a hit!

--
Ed

I think a lot of that is the 4 year pent up demand Mac users have had for more power rather than anything to do with 64 bit.
So a lot of Mac users probably haven't updated in 4 years so they have a good reason to upgrade now.

The PC users have had performance increases constantly for the last 4 years so most have less reason to upgrade right now.

Also they haven't ALL shipped yet. Most people that ordered the Dual G5 had their orders pushed back a month or more.

Cheers,
JS

[Ed M.]

Yeah the argument that the AMD64 will have a slow adoption applies to the Mac as well. Not to mention the 3% market share so even if there is a fast adoption it won't be that significant.

Duh! Let's just use the 100,000 G5's shipped figure. Some of those machines might well be the dual-processor configs. Whatever, let's just say that as of right now, there are roughly 100,000 G5 systems out there. Compare that to the 5000 or so AMD64 CPUs that have shipped and it's obvious. That's 2000% more systems already out there since it was introduced.

Also how many Apple apps are 64 bit now? How long will it take and how big will the market be?

What don't you understand? I've been trying to tell you that current MacOS X apps needn't be fully 64bit in order to gain access to the 64bit features (math and addressing) that the G5 offers *now*. Apple has done a great job with their API's and they are releasing more and improved versions as time goes on. This is significantly different from the AMD64 platform where in order to gain ANY access to the 64bit features of the chip, the OS AND the app *need* to be 64bit. That doesn't apply to OS X or the PPC platform since 32/64bit was designed into the PPC ISA from the very beginning.

The proof will be when apps become available that feature G5 support? How many Windows apps (or OSs for that matter) out there feature AMD64 support?

None.

The only one Ed has mentioned was Photoshop and that it had a plugin to make it 64 bit which still makes you wonder how much of the app really is 64bit.

That's the beauty part about it ... the entire app needn't be 64bit, just the parts that will make a difference. Why risk slowing down other parts of the app (if it went fully to 64bit) when you can surgically target specific sections and routines that can benefit instead?

Remember, there are large parts of code that will actually SLOW DOWN if they were made 64bit simply because there would be much more data being shuttled over the bus. And this is what AMD64 developers have to contend with.

Let's say there is an app out there (the large majority?) that will not benefit from a performance increase at all from going to 64bit. In many cases, the 64bit version of that app might actually be slower than the 32bit version. The best that can be hoped for is a "same speed" situation where even though there wasn't a performance increase, they maintained parity.

The thing is that in some apps, there very well might be specific regions where 64bit will be a big help. This is where the G5 makes the difference because Apple allows access to the 64bit features of the chip whereas AMD says that the app and OS have to be 64bit before any of the new features can be exploited.

Also the only compelling advantage is the ability to address more memory which is nice but how many people are even using 4 gigs right now when most computers only support 2 gigs anyway?

Exactly. That's why Apple is at a significant advantage *now*.

For me to switch the more compelling reason will be speed which isn't a guaranteed advantage just because it's 64 bit.

Exactly. And there is a speed improvement with the G5.
Just wait and see.

--
Ed

[js33]

Originally posted by Ed M.
Duh! Let's just use the 100,000 G5's shipped figure. Some of those machines might well be the dual-processor configs. Whatever, let's just say that as of right now, there are roughly 100,000 G5 systems out there. Compare that to the 5000 or so AMD64 CPUs that have shipped and it's obvious. That's 2000% more systems already out there since it was introduced.



What don't you understand? I've been trying to tell you that current MacOS X apps needn't be fully 64bit in order to gain access to the 64bit features (math and addressing) that the G5 offers *now*. Apple has done a great job with their API's and they are releasing more and improved versions as time goes on. This is significantly different from the AMD64 platform where in order to gain ANY access to the 64bit features of the chip, the OS AND the app *need* to be 64bit. That doesn't apply to OS X or the PPC platform since 32/64bit was designed into the PPC ISA from the very beginning.

The proof will be when apps become available that feature G5 support? How many Windows apps (or OSs for that matter) out there feature AMD64 support?

None.



That's the beauty part about it ... the entire app needn't be 64bit, just the parts that will make a difference. Why risk slowing down other parts of the app (if it went fully to 64bit) when you can surgically target specific sections and routines that can benefit instead?

Remember, there are large parts of code that will actually SLOW DOWN if they were made 64bit simply because there would be much more data being shuttled over the bus. And this is what AMD64 developers have to contend with.

Let's say there is an app out there (the large majority?) that will not benefit from a performance increase at all from going to 64bit. In many cases, the 64bit version of that app might actually be slower than the 32bit version. The best that can be hoped for is a "same speed" situation where even though there wasn't a performance increase, they maintained parity.

The thing is that in some apps, there very well might be specific regions where 64bit will be a big help. This is where the G5 makes the difference because Apple allows access to the 64bit features of the chip whereas AMD says that the app and OS have to be 64bit before any of the new features can be exploited.



Exactly. That's why Apple is at a significant advantage *now*.



Exactly. And there is a speed improvement with the G5.
Just wait and see.

--
Ed

Hi Ed,

I'm not arguing that the Mac isn't on the right track. In fact I agree with you about the hybrid OS approach. 32 bit apps will work on AMD64 as well. Intel is the only company with the wrong approach right now. It will take years before any desktop is fully 64 bit. It's just too early for me to see who will be the right one to go with. I have a lot invested in PC software but luckily Lightwave gives you both versions. I will probably have to stay with the PC until some compelling must have reason to go all Mac comes to light. I do have an iMac and it is useful although I don't do any production on it. I am getting the FCP express since it is free with a Premier CD tradein. :D

I am content to watch for awhile to see what happens.

Cheers,
JS

[DaveW]

Originally posted by Ed M.

AMD and Microsoft stated that 64bit support would come AFTER Blackcomb ships


Ed, re-read the article you posted, x86-64 support is coming BEFORE Blackcomb, not after.

[policarpo]

js33,

Couldn't you have just PM'd Ed about this discussion instead on making an entire post about this?

This thread serves no purpose really, and just hangs people up who are on modems that might have thought it a worthwhile post.

Please consider this the next time you post something with malicious sarcasm being the intent.

Much appreciated.:D

[js33]

Poli,

It wasn't malicious sarcasm. Ed has filled many threads with long winded explanations about why Win64 won't be available for 10 years. This was just to set the record straight. That's all.

Cheers,
JS

[dfc]

There was some discussion going on in the Digi forums about should Digi recompile Protools for G5 to address more than ram..etc...ie.

The response from the Digi engineer was..that protools, the app, in and of itself...didn't need more than 512 megs of ram to run. That more ram than that..with a lot of plugs..etc..could help ..but 4 gigs of ram (32bit) was more than overkill for this.

He also mentioned that OSX (10.2.7 and panther) that they were testing..still has a 32bit ram limit (4gigs) for apps or a single process. Huh?

So, I went to do some checking..and I found this

http://www.osnews.com/story.php?news_id=4009

It turns out, that thru all this, at least with 10.2.7 and coming Panther, there will "still" be a 4 gig ram limit for each application or process on the G5.

The idea, that one could recompile a 32 bit app..to address 64bit ram addressing (and the 8 gig or beyond limit) is not correct. ..at least for now.

This, according to this article, is similar to how windows NT functions.

At this time, there is no way to write an application for the G5 in such a way that "one" application or process can address more than 4 gigs of ram (the 32bit limit).

The OS itself can see/use all ram. In that way, it's good for running several apps at once. According to apple.."parts" of OSX use 64bit etc. But I could find anything on apple dev site to show that applications could be recompiled to address more than 32bits in memory. (4 gig limit)

So much for that theory. (that is..if I understood it right...ie..the 32bit apps could be recompiled to address more than 4 gigs of ram). According to what I found out..that's not possible.

That brings us back to AMD. Whenever the AMD 64bit windows comes...it WILL be able to have a single app or process address more than 4 gigs of ram by writing it in full 64bit. And my point is..it appears...this is exactly the way it will have to happen under oSX and the G5 as well. (a future full on 64bit version of OSX that also supports legacy 32bit apps)

There will be no single apps or process that can address more than 4 gigs of ram until some future update from apple for OSX (beyond Panther) and as well..until applications are completely rewritten in 64bit for it.

In that "one" regard...it's no different than AMD.

Ah well...4 gig's per process..still leaves you with the capability to address "more apps" or processes...up to the higher 8 gig limit (and beyond...the G5 supports 2gig dimms..so...once those are available..the limit will be 16gigs).

Perhaps it was my own misunderstanding Ed? I thought the premise of this was..that the 32/64bit hybrid OSX under G5 would be that applications could be recompiled to use the math libs and memory addressing at 64bits and 32bit apps could address more than 4 gigs of ram? And that's one reason specifically why it was a better implimentation strategy than the AMD roadmap.

Maybe that was just my misunderstanding of what you have described.

Still trying to get a handle on it.

[Ed M.]

js... ahh.. I'm away for awhile and I see that you are still at it... tsk tsk. tsk

First of all, I presented the timeline. You can go back and lay it all out for yourself. All the information was taken from AMD and Microsoft representatives, not me; Therefore they are the ones you should be complaining to. However, let's take a look at a few things...

Even if we were to agree that 64bit Windows was coming to Opteron a few things still stand in the way of it's adoption.

- Time. When will we see it? Every Windows-based Opteron rig out there is only a 32bit machine as it stands right now. Let's agree that it will come *before* Blackcomb.. So what, when is Blackcomb slated for release? The answer to that is AFTER Longhorn (As stated by Microsoft) It's even on their website stating as much. Longhorn has just been delayed *again*. I find the comment about the "64bit version of Windows for Opteron" being rather significant, not to mention strange. If availability and delivery of such a system was imminent (or at least years sooner than PC people are clamoring for) then why didn't the Microsoft and AMD representatives say it would be *before* Longhorn instead of Blackcomb? I ask because those comments seem to imply only that it will ship sometime before Blackcomb. That's a suspicious buffer of a LOT of years. Blackcomb won't be out until around 2007-2008. It could be that the OS could ship tomorrow or on Dec 31st 2006. That's really narrowing it down, eh?

- Price. Even if the OS was available TODAY, how much will this Server OS cost? Hint: more than the top G5 system! Are the masses of everyday users going to leap out and buy it just so they can run it on their spanking new Opteron server-stations? What will they do with it?

- Apps. What apps will be there? 32bit apps? That would make it run exactly like any 32bit system currently out there. There is no way for 32bit apps to get at the new features that AMD64 brings to the table. The apps AND the OS AND the drivers all need to be 64bit clean. This raises another problem for adoption and one that I'm sure Intel already ran a simulation on.

You would want an AMD64 machine (Opteron or otherwise) on which to test the app. Nobody likes to run code that the developers didn't test on the actual hardware. It's not clear that many developers are testing these things. Not many Opterons have sold since it's introduction. Roughly 5000 CPUs or so. And Athlon64 hasn't even been released yet.

To gain any advantage that AMD64 brings to the table with respect to 64-bitness, you have to recompile the app and do quite a bit of debugging to get the app 64 bit clean. Then you have to debug it alongside other 32bit apps to make sure that compatibility is maintained. Not only that, but some developers may see no huge advantage of moving their app to AMD64 just as many app developers haven't moved their apps to Itanium. The only difference between AMD64 has Itanium is that Itanium's emulation layer for 32bit apps bites. Developers will be facing the same questions. IF the decision is made to support AMD64, why not Itanium? Most developers will continue to offer only 32bit apps for some time to come. The vast majority of machines out there will be 32bit, and not many developers are going to want to support multiple versions of the same app for the same platform.

Even if all the developers that promised AMD64 support were able to buy AMD64 machines the first day they were out, fix their code, recompile and debug, check for compatibility and were willing to do so for free, there still remains a monumental distribution problem.

Users would have to download or otherwise replace their *entire* software library in order to get everything 64 bit clean. That is if they wanted to take advantage of the chips new features. This might not be a big deal for the slow-selling Opterons, but for Athlon64 machines that are expected to outsell them, this is important. In a real world, nobody wants that kind of trauma of having to replace their wares at one shot. So, even if the OS came tomorrow, expect the road for PC users to be exceptionally long toward all apps being 64bit clean. Sure, you'd be able to run 32bit apps along side (or so they claim), but they will not run any better than current versions of the same app on the same platform save for maybe speed and that's something Intel is probably banking on.

You need a nice smooth forward migration path, hopefully one that uses established distribution methods that everyone likes and can take advantage of on their own schedule. Until that process runs its course, sensible users demand that old software should "just work" on the new processor, and work better than on the processor it replaces. To achieve this, there needs to be a lot of fixing, debugging, recompiling, testing for interoperability between current and new 64bit apps and all those wonderful device drivers for the thousands of devices that are currently out there. This all takes time. The timeline speaks for itself. It's all been based on information provided by AMD and Microsoft representatives and I agree that it does raise a lot of suspicions.

So, no record has been set straight.. There is no 64bit support for ANY AMD64 processor with any version of Windows. There is also no Athlon64 processor for the desktop yet and furthermore, there are no 64bit apps (the ones that are needed) available to take advantage of the new features. Opterons are selling as slow as Itanium processors. I wonder why? It will be a lot longer than you think before Microsoft releases a 64bit version of Windows for the desktop. They only provide just enough vague information to make it sound more imminent than it really is. You guys can do the math.

--
Ed

[Ed M.]

dfc...

He also mentioned that OSX (10.2.7 and panther) that they were testing..still has a 32bit ram limit (4gigs) for apps or a single process. Huh?

So, I went to do some checking..and I found this

http://www.osnews.com/story.php?news_id=4009

It turns out, that thru all this, at least with 10.2.7 and coming Panther, there will "still" be a 4 gig ram limit for each application or process on the G5.

The idea, that one could recompile a 32 bit app..to address 64bit ram addressing (and the 8 gig or beyond limit) is not correct. ..at least for now.

This is not correct at all.

At this time, there is no way to write an application for the G5 in such a way that "one" application or process can address more than 4 gigs of ram (the 32bit limit).

Bull. Who told you that, OSNews?

Since the G5 is a 64 bit processor, it handles 64 bit pointers, it has 64 bit registers and it does 64 bit arithmetic. There isn't a single standard by which it is not a 64 bit processor. The OS is a hybrid. Plain and simple.

Apple *does* in fact have APIs that let you reference MORE than 4 GB of memory in a single app, if you really, really need it right now, this very instant.

Most developers should already be aware of this. So, let's put the issue to rest. It's not that it *can't* address more than 4GB per app, it's that developers must not see a need to do it now. The APIs are their for them if they want to address more than 4GB of RAM per app.

Also, 64 bit math (i.e., adding two 64 bit values together) is possible, extremely easy and happens automatically in a 32bit app if developers compile with the appropriate flags that tell the compiler that the code doesn't need to run on a G4/G3, but rather a G5.

So, the question that many users are thinking about is: What happens if you load your G5, running MacOS X.2.7 with 8 GB (16GB as soon as 2GB modules are available) of RAM? Does the RAM get used? Yes, ALL of it!

Most developers are going to initially provide tweaked apps for G5 that will be limited to a virtual address space of 4 GB (((unless you use the APIs described above))), again, their decision, but this doesn't prevent multiple apps -- even old apps that never even dreamt of running on the G5 -- from together using more than 4 GB of RAM. Period!

How does that work? Virtual Memory.

This is how a developer at Apple explains it:

This means that every address that the program sees is virtual -- doesn't map directly to hardware -- thus, the address 0x00010000 from one app may point to a different chunk of RAM from the address 0x00010000 in another app. This means that even if you only have a 32 bit virtual address space, as most MacOS X apps currently do, you can still have one app use one 4GB chunk of memory and another app use a different 4GB chunk of memory ... assuming of course they each *needed* that much RAM. This way the whole 8 GB (later 16 GB) can be used and the apps will never trip over each other. If you have a really big job to do, one easy thing that developers could do is split it up among multiple apps.

Still don't believe that Apple offers APIs that allow developers to address more than 4GB RAM in a single app? Contact Apple and ask someone on their dev forums. They'll point you in the direction of what you are looking for.

That brings us back to AMD. Whenever the AMD 64bit windows comes...it WILL be able to have a single app or process address more than 4 gigs of ram by writing it in full 64bit. And my point is..it appears...this is exactly the way it will have to happen under oSX and the G5 as well. (a future full on 64bit version of OSX that also supports legacy 32bit apps)

Not true, it only has to support enough RAM as the user is able to physically install in a machine. The G5's will soon ba bale to utilize 16GB. The difference is that developers can do it NOW within existing apps. There will be no such advantage given to 32bit apps running on AMD64. Those AMD 32bit apps don't even have an option of seeing more memory nor any other 64bit feature. Apple offers APIs allow access *now*.

There will be no single apps or process that can address more than 4 gigs of ram until some future update from apple for OSX (beyond Panther) and as well..until applications are completely rewritten in 64bit for it.

Wrong. APIs are available. You'll have to contact Apple. The *truth* is that developers don't see a *need*, not that the OS can't handle it. If there is an app that is being planned for OS X on G5 and said app will require more than 4GB, rest assured that it will because it is possible, right now, for the reasons mentioned above.

In that "one" regard...it's no different than AMD.

Sure it is. You can have it now if developers decide to use the APIs.

Perhaps it was my own misunderstanding Ed? I thought the premise of this was..that the 32/64bit hybrid OSX under G5 would be that applications could be recompiled to use the math libs and memory addressing at 64bits and 32bit apps could address more than 4 gigs of ram?

Yes, you misunderstood. Contact Apple and ask them for information on said APIs that will allow an app to address more than 4GB or RAM.

And that's one reason specifically why it was a better implimentation strategy than the AMD roadmap.

Wrong. Haven't you read anything I've posted in previous threads? It's not *just* about 64bit memory, but about other features like 64bit math. A 32bit app running on an AMD64 system will not even know it is running on that system. It will have ZERO access to any 64bit feature that the chip offers, All it offers is being able to run "as is" with no performance hit.

On G5 systems with the appropriate tweaks, developers have access to all the 64bit math and memory and whatever else the chip offers through Apple's APIs. You don't have this option of making your app a hybrid app on an AMD64 machine. Refer to my post on describing the various compatibility modes of AMD64.

Developers programming for G5 can surgically and strategically target specific, current 32bit apps or sections of those apps that will benefit from the 64bit processor. In other words, they needn't make their *entire* app 64bit clean in order to take advantage of the G5. This will cut out any trauma associated with buying/upgrading all new apps. People can now have hybrid 32/64bit apps running on the G5.

Again, reread my initial posts.

--
Ed

[dfc]

Ed,
Some of the quotes used in that article "are" from apple employees off the Darwin List.