Intel's 2006 Processor Roadmap

[N. Electron]

Processor Cooking Ability

In Taiwan, some testes were done to compare the AMD processor against Pentium M already.

They should have included the G4. My testes turn out well done when using the G4 PowerBook laptop. Especially in the absence of pants.

 

[faintember]

That's too early! You should wait for a Rev. D MacIntel! Now those are gonna be awesome.
C'mon, obviously only people that have a current revision PowerBook can wait before buying a new x86 PowerBook. There's lots of users with money burning holes in their pockets just waiting for what they think is a worthwhile upgrade.

Wow, i really hope that you are being sarcastic in all of your comments.
I think my Ti PB will hold me over until rev b, and as for the "lots" of users that have money to burn, i am not one of them. For me it is a matter of keep my Ti book and not take out a loan to pay for my grad. work this year, or buy a new PB and go into debt for my last year in grad school.
For me, ill try waiting until my doctorate programs to go into debt! 😛

Changing APIs via an OS change is FAR more difficult for a developer than a chipset change. In fact, chipset changes are dramatically easier. In many cases its a recompile. In many others, its an intern's day job for a few weeks. For a select few, its a few months of intense work.

OS changes, on the other hand, could involve a few man-years of development. This should be one of the easiest transitions, frankly.

Thanks for the info akac. I really hope that you are right. During the OS 9 -> OS X transition i had to boot into os 9 to get any real work done, and i for one hate os 9's interface with a passion (especially since most of us are used to os x now.) I hope that i do not have to keep my current PB running for three more years just to get work done, that is if it can even keep up by then. Once again, i really hope that you are right akac!

 

[Village]

If you don't NEED something now, then wait. (ALWAYS a good plan with technology--later means better!)

Intel PowerBooks will be much faster per power consumption/heat than today's G4s. Today's PowerBooks are great, with a great OS and apps, but they fall short in raw speed.

Software-wise, Intel Mac early adopters will get the short end of the stick--a few apps may not be ready (or at least not running to their full potential) at first. PPC Mac owners will have no such issues to worry about. BUT I don't think it will a big issue--developers have lots of warning, and Rosetta can pick up much of the slack.

Still, it's one reason to choose PPC and be a "late adopter" of Intel Macs. It's a personal call.

Thanks nagromme and digitalbiker for the replies. I forgot to ask, on the flipside of the transition to Intel code is the continued support for fat binaries - How long do you think vendors would continue supporting PPC stuff? Once code is completed for Intel chipsets, can the developers just flip a switch to generate PPC-compatible code?

My thinking is this: why suffer the hassle and initial performance hit of being an Intel early-adopter when you can run a a solid PPC-based Powerbook without performance hiccups for the next three years?

It would take at least three years for enough Mac users to move to an Intel platform before the vendors could even consider dropping support for fat binaries.

 

[oskar]

Wow, i really hope that you are being sarcastic in all of your comments.
I think my Ti PB will hold me over until rev b, and as for the "lots" of users that have money to burn, i am not one of them. For me it is a matter of keep my Ti book and not take out a loan to pay for my grad. work this year, or buy a new PB and go into debt for my last year in grad school.
For me, ill try waiting until my doctorate programs to go into debt! 😛

Ha, yes I was being sarcastic on both sides. First for those that say that when PowerBooks with Intel processors are released, they will wait for a Rev. B. As someone else already pointed out, that makes no sense at all. If you doubt the quality Apple will release in a product, then something's definetly wrong.
As for those whiners that probably aren't even thinking of buying a PowerBook and whine anyway, was the other money comment.

 

[BGil]

Ummm there is a reason why Apple's mobile solutions are so cheap. Their hardware is crap at this point. Remember the 17" PowerBook originally came in around $3200. Expect prices to climb back into the price range to take into account new up to date hardware. Also I want to know if those prices are in bulk purchases or are per unit. If these are per unit expect the price of bulk purchases to be MUCH cheaper. And its a good bet that Apple is getting as sweet of a deal as Dell on this migration so in all likelyhood they are probably getting, as Steve would say, an insanly great deal.

Not really. PPC's from IBM are simply cheaper than X86's from Intel and AMD. Part of the reason is because the chips are much smaller (which in turn makes them harder to cool). That's why Microsoft is using IBM PPC's in the Xbox 360-- to make the machine smaller and save on chip costs.

Those prices are in lots of 1000. There's no way Apple is goijng to get the same price as Dell when Dell sells 10x more computers, tons of PDA's, and uses Intel branded mobos for everything.

And your numbers are off.....



So there will be a 2.17Ghz chip in the first quarter of '06 not 1.83. Might want to go back and read the source of your link which is.....The Register

No, my numbers are fine. I didn't say there wouldn't be a 2.17ghz proc at launch, I said it would be too expensive and hot to use in any of Apple's current form factors (if Apple wants to retain their current margins). Even the 1.83 dual core Yonah is too expensive for a Mac Mini or iBook (and probably a PB too). The 17 inch is going to be the only hope for dual core unless Apple is big into slashing margins and battery life.

 

[sjl]

I know all that very well, my point was that the developers of apps heavily using Altivec would have to port all that Altivec code to SSE by January (which is highly unlikely, and even then SSE probably won't be as effective as Altivec, certainly not initially) OR a lot of apps will run mightily slow on these shiny new IntelBooks in January.

My apologies; the way I read your comment, I didn't think you were aware of that point. If they wrote the Altivec code in a sane manner, they'll have documentation on the algorithms being implemented, and how they're being used. Still a pain to re-implement, but better than a rewrite from scratch -- as I understand it, SSE3 and Altivec share a lot of common ground in terms of what they can do in silicon.

Note too that a lot of apps use libraries to do the vector maths, rather than having hard-coded assembly; in those cases, it's up to Apple (or the other library vendors) to re-write the assembly.

Is it possible sic payable for Apple to put in an extra Altivec unit? It certainly would solve a lot of initial problems.

Possible? Maybe. But considering that this would be a special-case CPU, only really used by Apple ... I can't see Intel doing it at a low enough price point for Apple to be willing to pay for it. It would also be a short-term bodge job, and I doubt Apple would want the risk of vendors relying upon it being there in the long term. No, much easier overall is to say "Altivec is dead; use SSE3 on Intel instead." I'd stake my mortgage on Altivec not making the jump over to x86 land in the opcode-compatible sense.

 

[AidenShaw]

Celeron is just a name. 🙄

Yonah = Pentium M = Good CPU = Damn good Mac update that blows away the G4.

Yes, it's a name for the same chip with smaller cache and maybe a slower FSB.

Gee, that'd mean that the Mini would only get 2 MiB of L2 (yes 2) instead of the 4 MiB of cache, and would be stuck with a 533 MHz FSB instead of 667 MHz.

I wanna get Dirty !!

 

[AidenShaw]

waiting for x64 has merit

I agree partially with an earlier poster who said that Apple may wait and go totally 64 bit. Maybe not the Mac Mini, it has no need yet. But PowerMac and iMac continue to be 64, obviously! and the PowerBook to join.

Note that if Apple ships 32-bit x86 chips, then when they add x64 chips the "universal binary" will need to have 3 separate images - PPC, x86 and x64. This is probably the plan, unless the development machines are only generating x64 code in the universal binaries. (Actually, the reports of OSX86 running on laptops would seem to say x86 + PPC.)

There's no added value to 64-bit on the iMac - it can only use 31-bits of memory today. Nothing would be lost with a 32-bit iMac.

Apple's 64-bit is a bit of a joke, actually. No Cocoa, no GUI, terminal apps only? I'd be embarrassed to emphasize "64-bit" in the adverts.

 

[AidenShaw]

"Release" means volume shipment for Intel

Ah dont be so excited yet. January is only the time when these CPUs will be announced, so WWDC is the earliest time of introduction of new Macs

Intel's history is that on the formal release date *systems* are on the shelves from the major manufacturers.

Dell and others had dual core systems ready to ship the day that Intel "announced" the processor.

Development samples ship months before to the manufacturers, and truckloads ship out in time to have the systems in the stores on announcement day.

I find it entirely reasonable that Apple will announce Yonah PBs at MWSF and be shipping them in volume before the end of January.

Note that Intel's been talking about Yonah for a year, and has been showing Yonah protos (on 90nm tech) for half a year.

What's Apple gonna do if Dell, HP and Thinkpad announce Yonah laptops in mid-December?

Will they still wait for "one more thing" in mid-January? How anti-climactic (and The Steve loves to climax)!

 

[faintember]

First for those that say that when PowerBooks with Intel processors are released, they will wait for a Rev. B. As someone else already pointed out, that makes no sense at all. If you doubt the quality Apple will release in a product, then something's definetly wrong.

Humm i remember a fair amount of people having issues with the rev a Al PB, and those problems were consequently fixed in rev b and c. Remember reading a gajillion threads about "10.4.0 made my mac jump off a cliff!!!!111OMG111!!!!" Some people (not me) are still experiencing problems with 10.4.2.

You may be right, something may be wrong, and maybe it is Apple's quality control when it comes to rev a products.

Besides, i don't want to see what my software runs like under emulation in Rosetta when it pushes the limit of my 1ghz PB. By rev b most of the bugs should be worked out, we should have dual-core by then (at least) and most of the applications that i use should be universal binary by then (and it also gives me more time to save enough money to upgrade all of my software!)

Now if the rev a is dual-core, has other-wise great specs as we will expect in a new PB, the early hands-on reviews are good and the price is lower or constant with where the PB's are today, then i may just buy a rev a. Sorry, but Apple is going to have to convince me on this one! I don't part easily with $2500! I really want Apple to prove me wrong on this and release a great rev a. Either way they will get my next laptop purchase, it is just a matter of when and what i can afford at the time.

 

[aegisdesign]

Apple's 64-bit is a bit of a joke, actually. No Cocoa, no GUI, terminal apps only? I'd be embarrassed to emphasize "64-bit" in the adverts.

You don't really need a 64bit GUI though.

What you do need is 64bit memory access for databases or possibly for image processing - which all happens off-screen anyway.

 

[AidenShaw]

...and why should I...

You don't really need a 64bit GUI though.

What you do need is 64bit memory access for databases or possibly for image processing - which all happens off-screen anyway.

And why should I have to redesign my program to have a GUI process and a back-end CLI process? Because OSX has a half-a$$ed 64-bit implementation??

Windows 64-bit doesn't make me do this - virtually all APIs are 64-bit, so I just recompile and test for 64-bit clean and sell my app!

 

[shamino]

Intel will not build a custom board just for Apple sans PS/2 etc.

Not gonna happen.

You're right. But nobody said that Intel will be making any boards for Apple. Apple has very nice contracts in place with several manufacturers for making PPC boards. All of these companies also sell PC motherboards. It's a no-brainer for Apple to come to them and say "we need these same user-level features on an Intel board". And they'll get it, because these companies already make custom motherboards for everybody else in the industry.

 

[shamino]

Hell if Apple wanted to they could have boards that have SCSI, SP/2, Serial on it and simple disable the hardware in the firmware and ship it without the connectors and you wouldn't even know there was anything on there. (Which they have done in the past. Witness the mini's spare firewire port contacts on the mainboard that is sitting there without anything connected to it.)

Or the original iMac's solder pads where you can attach a socket for an Apple-standard floppy drive.

 

[aegisdesign]

And why should I have to redesign my program to have a GUI process and a back-end CLI process? Because OSX has a half-a$$ed 64-bit implementation??

Your code must be pretty half-a$$ed if you've NOT separated your business logic and data from your front end or it's an otherwise dinky small program. Perhaps you slept through the 80s and 90s but everything from Smalltalk through Java Swing, MFC and Cocoa encouraged the use of a Model-View-Controller design. That's why it's not such a big deal if Apple take their time to upgrade Cocoa to 64bit since most Cocoa programmers are up on MVC already. Since Universal Binaries are pretty much restricted to Cocoa apps, also not a big deal there too. GUIs aren't performance critical generally.

Windows 64-bit doesn't make me do this - virtually all APIs are 64-bit, so I just recompile and test for 64-bit clean and sell my app!

One of the things I'm a little dismayed with in the move to Intel is that they didn't take the chance to jump into 64bit intel _only_ but I guess that was too big a transition! There's definite performance advantages for 64bit intel because 32bit Intel is so lame whereas there aren't any advantages on PowerPC bar memory and a few esoteric maths functions. I suspect Apple will follow a similar path with Cocoa on Intel going 64bit last even though it's got a more potential on Intel than PPC.

I'm still shuddering at the thought of *thunking* being back in my vocabulary even if it's not the horrible 32bit->16bit thunk overhead from the WinNT v3.11 and OS/2 days. 🙁

 

[willyjsimmons]

I know this.

You're right. But nobody said that Intel will be making any boards for Apple. Apple has very nice contracts in place with several manufacturers for making PPC boards. All of these companies also sell PC motherboards. It's a no-brainer for Apple to come to them and say "we need these same user-level features on an Intel board". And they'll get it, because these companies already make custom motherboards for everybody else in the industry.

I know this.

We went over that already.

The problem is, boards not made by Intel, usually don't have actual Intel controllers.

The video,LAN,sound,SATA,on-board SCSI, are usually third party or a mix with Intel.

You don't want promise, broadcomm, or other 'shwag' controllers on your Apple.

You won't be happy.

Trust me.

Trying to get a non win32, 'officially supported', driver out of some of these vendors, is like pulling teeth, if not impossible.

I would question whether or not third parties will be able to produce a stable OS X driver for their controllers, by Q1.

A hodgepodge of drivers from different sources is just asking for a tech support headache.

Just look at the retarded number of drivers for Dells and what not.

Different controllers within the same product line, produced within the same year?

Sometimes creating a cluster-duck 😱 is a talent.

You know what I mean. 😀

 

[LSlugger]

Your code must be pretty half-a$$ed if you've NOT separated your business logic and data from your front end or it's an otherwise dinky small program. Perhaps you slept through the 80s and 90s but everything from Smalltalk through Java Swing, MFC and Cocoa encouraged the use of a Model-View-Controller design.

MVC is a design pattern. It has nothing to say about bridging the gap between 32- and 64-bit code. In common implementations of the pattern, including pretty much any Cocoa app., the model, view and controller use function calls to communicate.

On the PowerPC architecture, a 32-bit GUI is a space-saving optimization that, even an Apple apologist must admit, is an inconvenience. (Do you know of any 64-bit Mac apps.?) On the AMD64 architecture, a 32-bit GUI is just daft.

 

[AidenShaw]

I'm still shuddering at the thought of *thunking* being back in my vocabulary even if it's not the horrible 32bit->16bit thunk overhead from the WinNT v3.11

Thunking in NT is 16-bit to 32-bit....

There was no NT 3.11, it jumped from 3.1 (initial release) to 3.5.

Windows 3.11 was 16-bit DOS-based only....

And thanks, LSlugger, for pointing out that good modular design doesn't imply adding the complexity of using independent jobs and inter-process communication techniques instead of methods (or function calls, for older languages).

The .NET/C# code that I write uses methods (although some of the methods are implemented as inter-process calls, where appropriate - like to the database engine). My .NET code, however, compiles cleanly to both 32-bit and 64-bit targets without doing anything special due to a lame 64-bit implementation.

 

[AidenShaw]

details on new PowerMacs Tuesday after next

http://news.com.com/Intel+to+detail+new+chips+at+conference/2100-1006_3-5829313.html

Intel to detail new chips at conference
...
Intel plans to provide details on a new generation of multicore processors at its semiannual developer conference in two weeks--and in the process highlight how its chip families are coming back together.

The new chips will start to emerge in the second half of 2006 and sport, among other attributes, greater energy efficiency and better manageability, according to sources familiar with the company's plans. The chips are code-named Merom (for notebooks), Conroe (for desktops) and Woodcrest (for servers).

...
As at the last Intel Developer Forum in March, the dual-core and multicore concept will be tough to escape at the upcoming three-day event.

The company plans to provide updates on dual-core chips, code-named Yonah, Presler and Cedar Mill, that will come out in the first half of 2006. Although these chips precede the Merom generation convergence, they will also be made on the 65-nanometer manufacturing process, which will begin later this year.

and http://www.mercurynews.com/mld/mercurynews/business/technology/

A new flagship for Intel

ANALYSTS: DESKTOP CHIP WILL HAVE ROOTS IN PENTIUM M LAPTOP CHIP
By Dean Takahashi, Mercury News

In a move that could determine whether Intel will continue to dominate the $30 billion market for personal computer chips, the Santa Clara company plans to unveil a new flagship ``next-generation'' microprocessor for desktop PCs for the first time in almost five years.

How will Jobs do a "one more thing" now that we already know what he'll be unveiling?

 

[Websnapx2]

It's not just the chip, It's the package that would be unveiled by Jobs. We also are not quite sure what will be offered from apple or when. Thought this will give us a ballpark figure.

 

[AidenShaw]

the dynamics change completely

It's not just the chip, It's the package that would be unveiled by Jobs. We also are not quite sure what will be offered from apple or when. Thought this will give us a ballpark figure.

But the "one more thing" depends on Apple being able to control the announcement schedule - only now Intel will be setting the schedule, and the rest of the PC manufacturers will be there with Intel on announcement day.

If Intel starts shipping dual-core 64-bit chips next July, Apple won't be able to wait until Paris to say 'one more thing'....

The typical Intel announcement timeline has been:

• T-18/24 months : General roadmaps known
• T-12 months : Intel discusses new chips at a conference, using codenames (this happens in two weeks)
• T-6 months : small quantities of early samples are available to OEMs for design and test
• T-1 month : large quantities of chips start shipping to manufacturers, production ramps up
• T-2 weeks : a press event is called, and everyone knows that Intel will announce a nGHz blah blah blah on the Tuesday after next
• T-1 week : manufacturers are shipping the new systems to the channel, with "do not sell before next Tuesday" on them
• T-1 day : news reports in popular press (WSJ, SJMN, NYT...) say that "INTEL TO ANNOUNCE BLAH BLAH TOMORROW"
• T-0 : Intel announces BLAH BLAH, Dell, HP and IBM are onstage with them showing off new products with BLAH BLAH
• T+1 minute : walk into Best Buy, Fry's, CompUSA or wherever and walk out with a new HP or eMachine with BLAH BLAH

My question is really "how will 'one more thing' fit into that timeline?"....

 

[BGil]

You don't want promise, broadcomm, or other 'shwag' controllers on your Apple.

You won't be happy.

Trust me.

Trying to get a non win32, 'officially supported', driver out of some of these vendors, is like pulling teeth, if not impossible.

I would question whether or not third parties will be able to produce a stable OS X driver for their controllers, by Q1.

A hodgepodge of drivers from different sources is just asking for a tech support headache.

Just look at the retarded number of drivers for Dells and what not.

Different controllers within the same product line, produced within the same year?

We are in solid agreement here. It's been pretty difficult to obtain Linux and Win64 drivers for a lot of standard motherboard hardware these days so getting Mac drivers would probalby be equally as hard if not harder. Usually the most unstable part of any PC is the hardware supplied by the people not making the northbridge or southbridge. Apple needs to stay away from Promise, Silicon Image, and others. If they get Intel supplied parts and build their own mobo then they'll be fine. In fact, Intel mobos and processors are world-renowned for their stability and "just works"-ness.

 

[Websnapx2]

But the "one more thing" depends on Apple being able to control the announcement schedule - only now Intel will be setting the schedule, and the rest of the PC manufacturers will be there with Intel on announcement day.

If Intel starts shipping dual-core 64-bit chips next July, Apple won't be able to wait until Paris to say 'one more thing'....

The typical Intel announcement timeline has been:

• T-18/24 months : General roadmaps known
• T-12 months : Intel discusses new chips at a conference, using codenames (this happens in two weeks)
• T-6 months : small quantities of early samples are available to OEMs for design and test
• T-1 month : large quantities of chips start shipping to manufacturers, production ramps up
• T-2 weeks : a press event is called, and everyone knows that Intel will announce a nGHz blah blah blah on the Tuesday after next
• T-1 week : manufacturers are shipping the new systems to the channel, with "do not sell before next Tuesday" on them
• T-1 day : news reports in popular press (WSJ, SJMN, NYT...) say that "INTEL TO ANNOUNCE BLAH BLAH TOMORROW"
• T-0 : Intel announces BLAH BLAH, Dell, HP and IBM are onstage with them showing off new products with BLAH BLAH
• T+1 minute : walk into Best Buy, Fry's, CompUSA or wherever and walk out with a new HP or eMachine with BLAH BLAH

My question is really "how will 'one more thing' fit into that timeline?"....

I see what you saying, and with general computer companies, your 100% on the nose but there are two BIG factors that don't make it so cut and dry.

First is that Apple is known for pulling announcements out of their a$$es on little or no notice, take the U2 iPod of example. They sent out a media invite and announced it at it's own promo launch with the iPod Photo. Apple's product launches seem to have become more unpredictable now that they have the media's ear.

Second is the package idea I was referring to. Apple's product lines are entities unto themselves, almost sequels to the prior product incarnation, hence no reference to them by product numbers (PM 6548 and such) from majority of the user base. No other company can boast that about their entire line. Taking that into consideration, most mac users don't dwell on parts (this current thread being different because it's speculation on future items of a major product transition phase) but rather focus on the full revision of a products form factor, new technology features and general software enhancements. Take the G5 for instance. We all knew the 970's from IBM were coming out before the official announcement Jobs made for the G5, but the name, look, extra features (like Firewire 800, internal case design and methodology, optical audio) drew a definite line between the G4 and the G5. That helped make the "…one more thing" exiting. We can only compare Apples releases to other Apple releases. So to be honest, if and when Dell gets the same chips, It really won't/souldn't matter to mac users. I go to Apple for the fit and finish of the WHOLE final product, something no other computer manufacturer can claim. Features, software, parts and all create new incarnations of Mac products. The Mac experience is really more than the sum of it's parts, no matter who else has access to them and when. And BTW - if apple releases the same chips on the same day or even two weeks latter than say dell, ibm or even say sony, when it's all said and done who do you think will get more media exposure? Apples launches will have more attention than the other releases and I'll bet the exact chip used will be barely more than a footnote. Worth mentioning, but not the whole story.

 

[shamino]

The problem is, boards not made by Intel, usually don't have actual Intel controllers.

Usually. Not always. Anybody can buy Intel chipsets. And there are third-party companies that use them on boards.

Second, there are good non-Intel chipsets out there. You don't have to go with the cheapest junk.

The video,LAN,sound,SATA,on-board SCSI, are usually third party or a mix with Intel.

Since these are not (usually) part of the core-logic chipset, Apple can buy them from anybody. They can even use the same chips that they are using right now on their PPC systems.

Trying to get a non win32, 'officially supported', driver out of some of these vendors, is like pulling teeth, if not impossible.

Except Apple won't be trying to get a driver. They'll be getting chipset specifications, from which they'll write their own driver. Chip makers will give the specs to someone buying several million chips.

You don't seriously think all these companies were writing the OS X drivers for the chips in the PPC Macs, do you? With the possible exception of video drivers, I would be surprised if Apple didn't write every driver needed by the on-board hardware.

 

[heisetax]

The 1.5" - 2" PowerBook

I think that it is more likely that we will see updated (intel) powerbooks in june (WWDC) of 06. Thats what Jobs said. And think about it, that it is just over 9 months away... That means that Apple are able to update the PPC PowerBooks one last time, and still give it a decent life cycle.

I will be all over a low power, dual-core powerbook if that is the case (and I dont get one before then).

And I do think that we will get some sort of a redesign. It almost seems necessary to put something new and fresh out there to help distinguish the Intel Macs from the PPC Macs. Apple may very well stick with the Aluminum for the "Pro" line, but i think that there will be some change in the form factor.

Maybe we'll see the 1.5" to 2" thick PowerBook since Apple will be using the same processors as the Wintel machines.