Jump to content
Sign in to follow this  
Followers 0

Will SSE3 be optional for Tiger x86?


Swad

It appears that Apple hasn't quite made up their mind about SSE3 yet. Or at least, they're not done with the Intel version of OS X.

 

Those who have studied the new x86 version of the OS have reported that the necessity for SSE3 in the GUI has been established as fact. However, as this Apple document on the Intel transition states, "SSE3 is an optional hardware feature on MacOS X for Intel. If you wish to use SSE3 features, you must detect them first, similar to how you are required to check for AltiVec." The paper goes on to reveal that, "SSE is not available in any format for MacOS X for PowerPC and AltiVec is not available for MacOS X for Intel. When writing code for Universal Binaries to run on MacOS X, you should conditionalize your code using appropriate symbols like __VEC__ and __SSE2__ to prevent the compiler from seeing vector code for unsupported architectures for each fork of the universal binary."

 

So, will the "official" version of Tiger on Intel be optimized for SSE2 as well?

Sign in to follow this  
Followers 0


User Feedback

Recommended Comments



why would they? I would assume all of the MacTels will be equipped with SSE3, and it's not like Apple's trying to cater to the needs of both MacTel users and those who want to hack OS X so it can be run on their Northwood Pentium 4 processors.

Share this comment


Link to comment
Share on other sites

I think you are reading too much into this SSE3 detection guideline. This parallels what already should be (is) done for AltiVec and just re-establishs a good coding practice as new x86 extensions are possible in the future.

 

The alternatives are either (1) Apple plans to build SSE2 hardware or (2) Apple plans to release OSx86 to the currently installed base of PCs. Considering that I can buy an SSE3 Intel chip for $75 and that Apple probably can get it for half that price, I think we can rule out (1). While Apple may have some big plans for OSx86 in the future, I think we would at best see some kind of an official clone program way before Apple opens up the flood gates, ruling out (2).

 

Steve likes to be in control and he does not want to lose it during this transition. Right now, SSE3 is one of two things that actually are holding back OSx86 from the masses. The other thing is graphics card drivers, not TPM.

Share this comment


Link to comment
Share on other sites
Steve likes to be in control and he does not want to lose it during this transition.  Right now, SSE3 is one of two things that actually are holding back OSx86 from the masses.  The other thing is graphics card drivers, not TPM.

 

So true.

Share this comment


Link to comment
Share on other sites

Sad... because I think with the impending release of windows longhorn, and the many users who will not want to have anything to do with it, that Jobs could gain an incredibly large user base by properly timing a general x86 computer release of Mac OS X

Share this comment


Link to comment
Share on other sites

Apple doesn't want to become a software company though. And essentially, since their hardware is pretty expensive anyway and barely anyone would buy it, that's what they'd end up being.

Share this comment


Link to comment
Share on other sites

most agreed, but even if they become more software :mellow: the companywill still make money. And most of all bye bye MS.

Share this comment


Link to comment
Share on other sites
its two years... technology moves on..

its not two years do the simple math the transition will end before the end of 2007 thats two years or so but the fist intel macs are early/mid 2006 its mid 2005 its under a year or a year max for some hardware and osx

Share this comment


Link to comment
Share on other sites

The reason they haven't commited to requiring it yet is because Intel's mobile Centrino platform doesn't support it yet. Notebooks are where they're going to start this transition. Since I've never seen any software reap large gains from SSE3, it's probably lower on their list of things to include then some other features.

Share this comment


Link to comment
Share on other sites
Notebooks are where they're going to start this transition.

technically right but wrong also there starting the tranidtion with consumer models which is macmini,emac,ibook and technically the might even start with imac but i wouldnt be suprsed if they left imac a little later

then pro models powerbook,powermac,xserve

thats 1 notebook to 2-3 desktops soit doesnt look like notebooks are the starting point

Share this comment


Link to comment
Share on other sites

GUYS!!

The reason SSE is optional is that the app should be able to run on PPC and X86 using Altivec acceleration OR SSE3 acceleration for all the nice stuff.

It doesn't mean that apple will release SSE2 machines (they will probably not)

Share this comment


Link to comment
Share on other sites

People are forgetting something. SSE3 must be optional in universal binaries, but as far as the OS itself is concerned (Rosetta), there is no reason to make it optional. EVERY x86 mac shipped from now on will have SSE3, so there is not, nor will there ever be, a reason for the OS itself not to require SSE3.

 

Think about it. All new desktop processors from Intel (Pentium 4) and AMD (Athlon 64) support SSE3. The only processor that anybody cares about that don't currently support SSE3 is the Pentium M.

 

However, by the time Apple releases x86 macs, the next Pentium M core, Yonah, will be available, and it features SSE3 support. It is rumoured, in fact, that Apple will be using the Yonah-based Pentium M processor in desktop x86 macs; it will be out in time.

 

So what is my point? Simply this: if OSX requires SSE3 now, there is zero reason for Intel-only parts of the OS (Rosetta) to remove the SSE3 requirements.

 

Sorry, but you guys are out of luck; if you want OSX on an x86 right and proper, you need a recent Pentium 4 or Athlon 64 processor.

Share this comment


Link to comment
Share on other sites

I am utterly amazed by the totally clueless wishful thinking in parts of this thread - by the time Mac/intel systems will ship, the relevant processors that Apple will be using will all be using a minimum of SSE3. The described consideration applies to universal binaries -- i.e. to properly branch to Altivec or SSE3 using code, while keeping the universal code free of any dependencies on such specialized instructions. Common sense, really.

 

Furthermore, while it's certainly nice to see patches and hacks to be able to install and run the current 10.4.1 based developer kit on regular x86 hardware, you *have* to be aware that this won't last (and that this was an intended outcome on Apple's part) - after all, notice the total lack of lawsuits or legal threats to sites and individuals that managed to crack Marklar. This isn't accidental.

 

This has provided Apple with valuable insight into the process of hacking what they released, while at the same time setting Mac OS X 'free' into the wild until the next revision. Essentially, it allows them a pretty effective viral marketing campaign, by letting you folks do the hard work (undoubtedly, totally self-installing DVD is in the offing soon from the same sources), getting OS X out into the open, only to finally yank it away with either the next system update, or the release of Leopard.

 

Just a few patches won't be enough to get 10.5/Leopard to run on the same standard hardware - it seems everyone is blinded by Apple's developer kits, thinking (stupidly, I might add) that these will be close to what Apple will ship in 2006.

 

Think again.

 

While Apple certainly will be using as much as possible Intel CPUs and support chips, you can bet that these motherboards will also include a couple of custom Apple chips (as well as firmware) that will drive I/O architecture, and which Mac OS X will not just look for, but depend on. This isn't news, as that has pretty much been what Apple has always been doing (though, admittedly, on the PowerPC platform, they needed to create their own support chips, as neither Moto nor IBM ever really had any worth their salt).

 

Would it be possible to circumvent this, and write custom kernel extensions to bypass these dependencies, and provide drivers for standard PC hardware instead? Certainly -- but it won't be as easy (comparatively speaking) as hacking Marklar to work 'in the wild'.

 

Apple's not stupid, and Steve Jobs knows exactly what he is doing -- that's the one thing you can rely on - so, enjoy the current situation while it lasts, but look forward to raised bar when 10.5 is released (though you will also see drastically new hardware, at some dramatic new pricepoints, which may further erode the whole 'price' arguments from PeeCee zealots - though we're not quite at the point where we'll see DELL selling Mac/intel systems... and we won't be for quite some time.

Share this comment


Link to comment
Share on other sites

"notice the total lack of lawsuits or legal threats to sites and individuals that managed to crack Marklar. This isn't accidental."

 

Are you aware that *every single* software title that is released - be it operating system, game, or general application is cracked and released on the net? How many lawsuits come about as a result of that? Almost none compared to the number of releases. Why would apple be any different?

 

Furthermore you actually have to know WHO to sue. How does apple know who to sue? It's not as if the leaker/cracker left a name and address in the release. It could take months to track that person down if ever. Most of these people have been doing this for years and are experts in covering their tracks.

 

Then you also have to consider the location. If the cracker is located in a nation that does not have strong (or any at all) intellectual property laws apple can't do anything. For example, if the person who cracked it is located in say China or Nigeria the courts in those nations don't really care if OS X is cracked. Nor do they care about finding the person who did it.

Share this comment


Link to comment
Share on other sites
Are you aware that *every single* software title that is released - be it operating system, game, or general application is cracked and released on the net?  How many lawsuits come about as a result of that?  Almost none compared to the number of releases.  Why would apple be any different?

 

OMG, are you serious? Every single piece of software is cracked and pirated? No way - now, there's some news. Quick, call the press...

 

DUH!

 

You're new at this, aren't you?

 

While the hypotheticals that you are listing certainly make for nice academic arguments, you are (not surprisingly) entirely ignoring reality and facts.

 

a) Mac OS X x86 developer kit is not just any old piece of software - in fact, it's probably one of the most closely watched releases, and certainly THE most closely watched by Apple.

 

:) Apple has taken legal action over much less, and at the very least could go after the sites which are providing instructions (and links to downloads and torrents), on how to patch and bypass their protection - all of the various methods in use currently could be pursued and enforced under the DMCA, and it would pass.

 

c) While they will certainly have trouble finding phenix, or any other group releasing the dev kit, they most certainly could shut down sites like this one.

 

d) Lastly, let's see who'll whine when 10.5 gets released. Either way, when (if) an update to the dev kit gets released, it will be telling what Apple's position is -- if patched systems will continue to work after the update, then it is in Apple's interest to allow the proliferatin of OS x86 during this pre-release. If they will stop working, well, then we'll have our answers.

 

Easy - your wishful thinking won't change this.

Share this comment


Link to comment
Share on other sites
Think again.

 

Daffy, there's really no need to be condescending. And that's a *lot* of declarative statements from someone who, as far as we know, has no connection to Apple. This is all just speculation -- and let's keep it friendly.

 

I don't think it's a safe bet at all that Apple will continue to develop custom I/O hardware. With the state of modern x86 I/O being what it is, what would they possibly gain by doing this, other than making it somewhat more difficult to run OSX on stock PCs?

 

My thinking is this: if the OSX install CD won't install on stock PCs, then 99% of PC users aren't even going to touch it. End of story. Why spend that much money, time and effort just so that it's cracked in 3 months instead of a week, when only a ridiculously small number of people are even willing to go there?

 

I think Apple would love it if they could spend less time worrying about the nitty gritty of I/O chipsets and more time thinking up cool {censored}. Now they can.

 

But I do agree that Steve-o has something up his sleeve. For whatever reason, he's watching all this very closely, I'm sure.

Share this comment


Link to comment
Share on other sites

Flamewar aside, it is almost certain that OSx86 will always require SSE3. The chance of this changing are infintesimal. All Intel macs will feature SSE3, and Rosetta needs SSE3 for good reason. So there is no reason for Apple to remove the SSE3 requirement. Not gonna happen.

Share this comment


Link to comment
Share on other sites
Flamewar aside, it is almost certain that OSx86 will always require SSE3. The chance of this changing are infintesimal. All Intel macs will feature SSE3, and Rosetta needs SSE3 for good reason. So there is no reason for Apple to remove the SSE3 requirement. Not gonna happen.

 

I doubt that the SSE3 requirement is going away, but for a vastly *different* reason: except for Pentium-M, how much non-SSE3 product is Intel shipping to anybody (let alone Apple)? Exactly none! The P4 D line is all-SSE3. So is what's left of the P4 S478 line (which is about to croak). Even the *Celeron* line is all SSE3 these days. The only hole is in the M line, and Intel has a vested interest in getting rid of the hole, and it has nothing to do with Apple. SSE3 as a requirement for Intel-based Macs makes sense because Apple, as a customer, will *not* have any non-SSE3 product *except* PowerPC Macs, where the OS already runs natively.

 

So, basically, SSE3 will stick because Apple literally has zero reason to get rid of it.

Share this comment


Link to comment
Share on other sites
Daffy, there's really no need to be condescending.  And that's a *lot* of declarative statements from someone who, as far as we know, has no connection to Apple.  This is all just speculation -- and let's keep it friendly.

 

From one cartoon character to another - fair enough.

 

I don't think it's a safe bet at all that Apple will continue to develop custom I/O hardware.  With the state of modern x86 I/O being what it is, what would they possibly gain by doing this, other than making it somewhat more difficult to run OSX on stock PCs?

 

"...other than..." ??? I would think that this 'other than' is the prime and absolute reason for them to be doing so, considering that their stated goal has been an absolute statement of 'no OS x86 on non-Apple hardware". In order to achieve said goal, just relying on a custom ROM is not enough (hasn't worked since the late 80s); the TPM chip reliance didn't work in the dev kits (at least, with regards to just having a kernel extension do the work); and we know they are going to rely on standard intel BIOS firmware - so, this leaves them with the most reliable approach, which is a custom I/O layer -- which just might be as simple as a standard intel I/O chipset with Apple specific modifications to accomodate OS X (at Apple's volume, this is economical for intel to provide)

 

You're right - they would not gain much else than this one feature, but considering that is not only their penultimate goal, but also (according to Apple) critical for their hardware survival, I would consider this to be an important enough benefit.

 

 

 

My thinking is this: if the OSX install CD won't install on stock PCs, then 99% of PC users aren't even going to touch it.  End of story.  Why spend that much money, time and effort just so that it's cracked in 3 months instead of a week, when only a ridiculously small number of people are even willing to go there?

 

Because, once that Genie is out of the bottle, you can't cork him back in there -- and Apple has every intention on not releasing him until a time of their choosing. Let's just leave it at that.

 

 

I think Apple would love it if they could spend less time worrying about the nitty gritty of I/O chipsets and more time thinking up cool {censored}.  Now they can.

 

They can do that either way, since intel engineers will be doing the thinking here, in that case, and not Apple engineers.

 

 

But I do agree that Steve-o has something up his sleeve.  For whatever reason, he's watching all this very closely, I'm sure.

 

In that, you are correct.

 

Let's just leave it at that - if 10.5 proves to be easily hackable (i.e. less than 60 days to an operational release by various groups), than we will know that Apple has a vested interest in seeing the beast in the wild (most likely to *really* take down MS a notch or two), and if that were the strategy, it would certainly work.

 

Then again, letting PC kiddies use OS 10.4 (dev kit) for the next year or so, letting them get used to it, and enjoying it, only to then pull the carpet out from underneath the thousands of users by then, will certainly also motivate sales of Apple hardware, when those folks want to maintain their 'fix'.

 

Apple can go either way, and with a solid offering of hardware on one end of the spectrum, and *very* solid professional software products on the other end of the spectrum, they do stand to actually benefit from a software sales strategy. The only item missing from that equation is an MS-Office type product, capable of dethroning Office -- heck, maybe they are even working on something like that right now (or maybe they will leave that market to MS' MacBU, despite them consistently producing {censored}).

 

Yep, Steve certainly has something up his sleeve. Keep buying iPods.

Share this comment


Link to comment
Share on other sites
Can someone remind me please if the new Universal Binaries ( PPC + x86 ) need SSE3 or just SSE2?

 

The universal binaries just need SSSE2.

 

On the apple developer pages is a note that says that SSSE3 is optional for Mac x86 hardware, so there is hope :). But they might change that in the future.

Share this comment


Link to comment
Share on other sites

Here's to the Crazy Ones!

 

And the chips made by Intel and AMD through the last 3 or 4 years are all SSE2 compliant, aren't they?

 

In 2006 almost every current OS X application will have Universal Binaries...and be SSE2 compliant. Only a small bunch of apps. probably the Pro ones ( like in the present they have Altivec for acceleration ) will need the help of SSE3.

 

So, in 2006 we have as OS and 90 to 95% of apps that can run on SSE2 computers!

 

This is crazy! But do you remember "The Crazy Ones, the misfits, the rebels. the troublemakers...."?

 

I could almost bet that Apple is going to the Great Prize : MS!

If not directly at least trying to undermine the success of Shorthorn/Vista and in the way selling even more with the free publicity.

 

As someone said, the only problem are the video card drivers.

 

But the Crazy Ones don't have "respect for the Status Quo" and "they change things. They push the Human Race forward".

Share this comment


Link to comment
Share on other sites
I doubt that the SSE3 requirement is going away, but for a vastly *different* reason: except for Pentium-M, how much non-SSE3 product is Intel shipping to anybody (let alone Apple)?  Exactly none!  The P4 D line is all-SSE3.  So is what's left of the P4 S478 line (which is about to croak).  Even the *Celeron* line is all SSE3 these days.  The only hole is in the M line, and Intel has a vested interest in getting rid of the hole, and it has nothing to do with Apple.  SSE3 as a requirement for Intel-based Macs makes sense because Apple, as a customer, will *not* have any non-SSE3 product *except* PowerPC Macs, where the OS already runs natively.

 

So, basically, SSE3 will stick because Apple literally has zero reason to get rid of it.

 

Vastly different reason? You just rephrased the exact same reason I gave :rolleyes:

Share this comment


Link to comment
Share on other sites

See, you're all basing your speculation on last week's news -- while Apple is basing their decisions on a roadmap they have seen, but which is only slowly being unveiled by intel.

 

Do the math!

Share this comment


Link to comment
Share on other sites


×