Inside the PowerPC 970: Part II

[Sun Baked]

I have to say, I feel that Hannibal is being a tad unfair to Apple when he blames them for the lousy bus interface on the G4

>Chryx

Yes it is a bit unfair about Hannibal & Co. blaming Apple for some of the motherboard problems, but Apple used to be on the bleeding edge of pushing the memory, chipset technology, and new hardware ideas in the pre-PCI days -- even if this led them down some dead ends.

But lately Apple has been lagging the market by quite a bit in adopting the next generations of current standards, and you can't blame Motorola solely for that.

 

[Rincewind42]

Re: Re: Re: Re: Not a Sure Thing

Originally posted by Blackcat
IBM distanced themselves from the original Altivec SIMD because they wanted one which did not require special coding. This is the main reason why there were G1/G2/G3 chips by Moto and IBM, but G4 chips only by Moto. IBM felt it was the wrong approach and AIM went all tits-up. (I'm over simplifying here)

So if IBM just wanted a SIMD unit on a Workstation POWER chip, I think they would use a more refined VMX without Altivec compatibility.

Due to the architecture of the PowerPC (and most CPUs in general) it would be impossible to add vector extensions to the instruction set without having to code for it specifically.

The reason that IBM & Motorola went in seperate directions was that they needed different things. IBM wanted to focus on multicore chips for their 4th generation, while Motorola wanted to focus on vector operations. Since multicore G4s never materialized (embedded systems would rarely use two G4s so why bother with the expense) there would have been no reason for IBM to be interested in the design. Conversely, once Motorola had sold Apple on Altivec there would be no reason for Apple to switch to a multicore chip without it (there have been times when Apple added hardware as The Next Big Thing then took it away - developers have learned not to use a new Apple technology until it hits second generation - sad but true).

Finally, VMX IS Altivec. It's the original name of the technology. And IBM still isn't using it in their operating systems, which makes it pretty obvious that it was put there for another customer - Apple.

 

[eric67]

Originally posted by jayscheuerle
... Now sorry, but that's no lame system, and you can bet their warranty is better than Apple's. Not everyone out there wants an all in one system and not everyone out there is running Final Cut Pro. Apple is alienating a huge segment of the population by not offering a competitive low-end system.

Bring 'em in, and bring 'em in on target. - j

it is so true that I still do not understand why it has not been done yet. Why can we not by a low end model then update it with time, let say change processor, better graphic cards, more RAM (stupid Apple certified RAM).....
this is why PC are popular, because people can order/built their own configuration from A-Z, not us...
maybe it is time for a change soon....

 

[benoda]

If the 970 does come to Apple, how is Apple going to deal with creating an optimized version of OS X along with pushing software creators to optimize for 64-bits etc. while still having G4's and G3's in the lower end? Seems like a bad idea to say "program XYZ is made for a powermac 970, but you other mac users are SOL."

 

[Mr. MacPhisto]

Re: Not a Sure Thing

Originally posted by alset
Again, I have to ask why everyone is so hell bent on the idea that this is a sure thing? Why take so many rumors as the word of god?

Lots of people setting themselves up for a fall, right about now. Rumors have let us all down before. If Apple doesn't deliver there will be people screaming for blood, and it won't be Apple's fault.

Dan

Yes it will be. Here we have IBM coming in with a new chip based on their POWER4 series. They want to expand their Blade market, but why would they just make this chip for Blade servers see as they don't sell a ton @ this point?

And I do KNOW Apple is using these things based on what friends at IBM tell me. In fact, I know that Apple is the one that specifically asked IBM for help.

 

[nicmac]

Apple employee discount

Wasn't there information recently that apple was offering Power Macs to their employees for a large discount.
Or do they always offer large discounts?
Or was that just a rumor?

 

[Mr. MacPhisto]

Originally posted by Mudbug
Please don't shoot the messenger, but he's right. There needs to be a single processor low end 970 machine with all the simple goodies at a price point around $800-$1000 to get home users to buy. Don't worry about including displays, but just offer discounts on Apple models if purchased together, much like they do now. And include 3 years of AppleCare for free with the purchase of a new or refurbished system.

I know you folks are going to say "well, home users aren't the market we're going after" - but they are. You want to use the same machine at home as you do at work, and the only way to persuade enough people to switch their offices over is if they have access to a mac that's better and cheaper at home than thier wintel box at work. Fortune 500 companies don't have thousands of P4 hotrods on desktops, they have netware boxes that are basically just end nodes of the network. What about just a simple network "appliance" (again, don't shoot me) for the enterprise that fits needs but doesn't overachieve?

Agreed. I also believe Apple needs to make ALL their machines upgrade friendly. That means making it easier for people to purchase chips, motherboards, etc - even if only from Apple.

As reported by my sources, I reported in an earlier post that IBM would be making workstations and (later) consumer-end OS X based systems. Apple needs to lower their pricepoint so they can allow for a limited amount of cloning and give the consumer (and the business) more options. If Apple's machines are better looking, more innovative, and the same price as the competition, then what is their to worry about?

I've also had a thought about bringing back the cube on the low-end. The iCube maybe? Only question is how can it be made upgradeable. I have some ideas and drawings on how to do this with the eMac, iMac, etc cause I think this is extremely important.

If I were Apple, I would allow for upgrades and license the ability to manufacture the upgrades out to several companies. Heck, I might even license out MoBo designs, giving the Apple user more options when checking out a MoBo. Just one of several ideas I have.

 

[areyouwishing]

Originally posted by eric67
it is so true that I still do not understand why it has not been done yet. Why can we not by a low end model then update it with time, let say change processor, better graphic cards, more RAM (stupid Apple certified RAM).....
this is why PC are popular, because people can order/built their own configuration from A-Z, not us...
maybe it is time for a change soon....

This might be of interest to you...
Core Crib (Barebones Apple Gigabit Machine that you can build from the ground up.)

The one drawback to this is that the processor runs on the older 100mhz bus (gigabit model), but you get the opportunity to upgrade anything you want...even to Dual 1.2ghz G4 processors.

 

[pi radians]

Originally posted by eric67
it is so true that I still do not understand why it has not been done yet. Why can we not by a low end model then update it with time, let say change processor, better graphic cards, more RAM (stupid Apple certified RAM).....
this is why PC are popular, because people can order/built their own configuration from A-Z, not us...
maybe it is time for a change soon....

Upgrading a system and building one from scratch are two totally different things. You will never be able to build a complete Apple computer from scratch. Upgrading a Mac is just as easy as upgrading a Wintel system.

While the cost may not be at a point you like, that is the cost of wanting a product with only 3% of the market. If Apple had a 95% marketshare, x86 companents would be as rare and expensive and the Apple compatible components would be cheap. It has nothing to do with Apple and everything to do with the market economics.

 

[mcs37]

Re: RIP G4

Originally posted by Blackcat
Who will buy a 1.4Ghz G4 iMac at $1300 if a 1.4Ghz Powermac at $1500 decimates it?

It's conceivable that Apple would do a complete shift-over: PowerMacs immediately, PowerBooks within a few more weeks, and the iMacs within a month. Not sure about the eMacs.

 

[Phazer80s]

Lower prices WILL increase market share!

.
.
.
The final paragraph from the Ars Technica article:

As with all things Apple, though, the big question is price. Will Apple drop its margins drastically and sell these machines at a competitive price point in order to increase market share, or will it continue to price itself into the increasingly non-existent luxury/lifestyle computing niche? I'm hoping for the former, because I'd love to give my TiBook and iPod some company with a 970-based PowerMac. If the answer is the latter, though, my Apple products will likely find themselves interoperating with an x86-64 Windows box._

A very good point. Makes you wonder how Apple will play this next move, no matter what new processor they choose. The number of people (professionals included) who can justify Apple's HW prices is shrinking rapidly. It's like they just don't get it.

Hardware price is probably their greatest weakness given the innovative and cost-effective software they've introduced lately. They need to make Macs popular!!! Truely competitive prices will make Apple's quality stand out all the better from other PCs.

Get it right, Apple. We're rooting for you (thin wallets and all.)

 

[gwangung]

Many people take the elitest viewpoint that Apple shouldn't focus on the low end, shouldn't devote the R&D, which is absurd.

No, it isn't absurd.

What I find absurd is people who don't know the first thing about business, spouting off about going after this market and that market, without knowing the facts and figures. The only figure they talk about is price....which is only part of the equation.

I know of NO company that can be all things to all people. A company that tries, particularly coming from a position such as Apple's, is going to go down the tubes.

 

[Rincewind42]

Originally posted by benoda
If the 970 does come to Apple, how is Apple going to deal with creating an optimized version of OS X along with pushing software creators to optimize for 64-bits etc. while still having G4's and G3's in the lower end? Seems like a bad idea to say "program XYZ is made for a powermac 970, but you other mac users are SOL."

Apple may like to push developers to 64-bit immediately (because this will sell their hardware...maybe) but they will likely not. Why? Because 99% of the software out there today has no need to be 64-bit. The software that does need 64-bit is software that is currently bumping into the 4GB memory ceiling of 32-bit CPUs and using hacks to get around it (if possible). For example, you won't see Apple release 64-bit iLife anytime soon, because there is little in the applications that can use it (yes, iMovie could take advantage of 64-bit CPUs, but that would relegate it to 970 based systems - something that I doubt Apple wants to do in the near term).

As far as optimization goes, yes there are chip specific optimizations that you can do for the 970. But most of them have nothing to do with the fact that it is a 64-bit CPU and these kinds of optimizations won't break compatibility with G4s & G3s (which 64-bit usage will).

And it's not new for software to require specific hardware to run. FCP 4 will only run on G4 based machines, as does other software that requires the Altivec Unit. But this is generally the exception, not the rule.

 

[jayscheuerle]

Originally posted by gwangung
No, it isn't absurd.

What I find absurd is people who don't know the first thing about business, spouting off about going after this market and that market, without knowing the facts and figures. The only figure they talk about is price....which is only part of the equation.

I know of NO company that can be all things to all people. A company that tries, particularly coming from a position such as Apple's, is going to go down the tubes.

So you're saying that it takes additional R&D to take out a SuperDrive or put in a lesser video card? I'm glad you're not my mechanic! Price may be only part of the equation, but Apple shouldn't hold onto the promise of their fans paying over 50% more for comparable systems to those in the PC market. When money's tight, people demand to get something for it and Apple is not making this "something" obvious enough. Integration with iApps is nice, but isn't enough to justify such inflated prices.

Apple doesn't need to be all things to all people. Unfortunately, they are nothing to most people and unless they can attempt to address that, their precious 3% can drop to 2%.

 

[Blackcat]

Re: Re: RIP G4

Originally posted by mcs37
It's conceivable that Apple would do a complete shift-over: PowerMacs immediately, PowerBooks within a few more weeks, and the iMacs within a month. Not sure about the eMacs.

Going 100% 64bit would be a marketing coup.

 

[Frobozz]

Originally posted by benoda
If the 970 does come to Apple, how is Apple going to deal with creating an optimized version of OS X along with pushing software creators to optimize for 64-bits etc. while still having G4's and G3's in the lower end? Seems like a bad idea to say "program XYZ is made for a powermac 970, but you other mac users are SOL."

Good point, but I think it's more simple than you think. I suspect the G3 will be nuked in favor of a G4... in due time. Maybe not initially. In either case I do believe they will continue to use G4's in the consumer space for a while until the .09 970's are available.

As far as the OS goes, the 970's will run any existing stuff just fine-- and faster. Most programs will have fat-binaries that execute code in 64bit or 32bit chunks based on which machine you're using-- just like they would use Altivec only if you have it. In other words, the transition is much less presecent to the user. In fact, only certain applications will really want to process data in 64bit chunks, as there is no speed boost unless you're dealing with complex math or with large chunks of memory.

 

[Frobozz]

Hard evidence of price

Can anyone give hard evidence of the price for a 970 chip? Or, for that matter, of the new systems?

A lot of people are worried that, although these systems will run fantastically fast, they will be more expensive than they are today. I, for one, do not agree with this. I've heard (conjecture) that the 970 will be, or currently ARE, less expensive than the G4's. I'm sure an arguement can arise about when that price threshold will actually come about, but the point remains. I think Apple knows full well that price is a concern and will keep their prices in line with what they have today. I supposed you could argue that they'd need to recoup some R&D costs, but that's not the case. You don't pass on R&D costs to the consumer like that unless you're looking to go out of business.

 

[jayscheuerle]

Re: Hard evidence of price

Originally posted by Frobozz
You don't pass on R&D costs to the consumer like that unless you're looking to go out of business.

But Apple's been going out of business for years!

 

[tcmcam]

Will this grow market share?

I think we would all like to see the 970 have a positive effect on Apple's marketshare.

What is preventing market share growth now?

1. Price. Yes, Macs are expensive. It's no coincidence that the Apple Stores are put in the "fancy" malls (next to Tiffany's) not the low-end next to Sears. As Steve Jobs says "Dell is the Walmart, we're not". Only people with fat wallets can afford to buy Macs (especially in this economy).

2. Upgradeability. Except for PowerMacs (which are even more pricey), most Macs can only be upgraded externally. Grant you, the ease and elegance of adding a firewire drive is great. The positive here is that "non-tech" computer users don't get themselves in trouble. The negative is that it costs you a lot more to just add another 40 GB. I sometimes wish the iMac had a "hard drive" access port that made it easy to upgrade the hard drive. And let's ditch the "Apple Slot" memory, heck, even most Apple Store employees hate changing that out.

3. Performance. After paying a lot for your sexy Mac, you still know that some dolt who just bought a Dell for $800 is running circles around you in Photoshop. Let's hope the 970 fixes this.

4. Windows at Work, wanting Windows at home. Even though many users hate WinTel's complexity, if Mom or Dad use a PC at work, they think they should use one at home. We could even extend that to "Windows at School, windows at home" because so many school systems now have Dell boxes and have thrown out their Macs. This is why MSFT Office is still a CRITICAL application for the mac.

Finally, I don't think Apple can simply lower prices to increase market share. One thing Steve should be commended for is making Apple a profitable company. They have to be careful lowering prices too much because Apple lives off those hardware margins. The only great software you have to buy is Final Cut and the other high-end media stuff. To Apple, "software upgrade" revenue comes from a customer buying a new Mac, not from buying a slew of new iApps or OS X.

The 970 is great. I loved the article, but what it really says is that Apple will finally have parity (better at floating point, worse at integer) with the Intel/AMD world.

I think the only issue that the 970 solves is the performance issue, Apple has to figure out the other 3 issues because those are management and engineering issues.

 

[Tim Flynn]

An interesting point on teh Ars article was that to fully get the benefit of the 970, the code needs to be recompiled for the 970. I am assuming 32 bit only code.
Since the ISA is the same between a G4 and the 970 for 32 bit code, the G4 or G3 should be able to run 970 optimized code.
So my question is; what is the effective speed difference of running 970 optimized code on a G3 or G4?

Apple/IBM seems to have an advantage over the x86 world. If you optimize code for a P4, it may not run on a Pentium III or below.

The advantage to software vendors (& Apple) is that they just need to optimize for the 970 and get the speed benefit. The code will still run on older G3s and G4s albiet slower, but runs. One code base!
Perhaps good planning pays off

 

[cubist]

Originally posted by Mr. MacPhisto
... I've also had a thought about bringing back the cube on the low-end. The iCube maybe? Only question is how can it be made upgradeable. ...

The Cube IS upgradable. I have added RAM, upgraded my hard drive, video card and processor. See http://www.cubeowner.com

 

[jayscheuerle]

Originally posted by cubist
See http://www.cubeowner.com

$899 for a 450MHz G4, non-expandable (though upgradable!) system?

Somebody's paying an extra $500 for the form factor, but I guess that's why they were bought in the first place, eh?

Gorgeous machines. Obscene pricepoints. Apple should be able to introduce a new cube with a contemporary processor for $899 NEW!

Dream a little dream... - j

 

[-hh]

Re: Hard evidence of price

Originally posted by Frobozz
Can anyone give hard evidence of the price for a 970 chip? Or, for that matter, of the new systems?

Anyone with access to hard pricepoints is most certainly under a NDA.

A lot of people are worried that, although these systems will run fantastically fast, they will be more expensive than they are today. I, for one, do not agree with this. I've heard (conjecture) that the 970 will be, or currently ARE, less expensive than the G4's. I'm sure an arguement can arise about when that price threshold will actually come about, but the point remains.

I don't see it as a huge worry. Between the 970, G4 and the P4, the P4 is the largest die set (increases costs), and largest volume (lowers costs), whereas the G4 is the smallest die set (lowers costs), and smallest volume (increases costs).

The 970 is in the middle for both factors. Don't forget that its volume is going to include whatever IBM builds with it.

So as good of guess as any is "somewhere between the G4 and P4 for price", which means that it shouldn't be a major factor.


And insofar as recouping R&D costs, that applies to all Apple models, regardless of which chipset they use.


-hh

 

[Booga]

Keeping up with the Joneses

I see most people overlooked the paragraph about halfway through comparing the overall speed of the 970 against that of the P4. It more or less says that unless the 970 can keep within 2/3 of the clock speed of the P4, it will end up being generally slower, and 3/4 of the clock speed or better for integer. Considering the P4 is already over 3GHz, I sure hope the 970 debuts at 2GHz or more.

But at least it indicates the Mac will be back in the ballgame in terms of performance.

 

[Rincewind42]

Originally posted by Tim Flynn
An interesting point on the Ars article was that to fully get the benefit of the 970, the code needs to be recompiled for the 970. I am assuming 32 bit only code.
Since the ISA is the same between a G4 and the 970 for 32 bit code, the G4 or G3 should be able to run 970 optimized code.
So my question is; what is the effective speed difference of running 970 optimized code on a G3 or G4?

The advantage to software vendors (& Apple) is that they just need to optimize for the 970 and get the speed benefit. The code will still run on older G3s and G4s albiet slower, but runs. One code base!

The speed difference between optimized for 970 and not optimized for 970 probably won't generally amount to more than 5% on average. There are pathelogical cases (which is pointed out in the Ars article) but for the most part there it won't make a huge difference.

And you will get a speed boost on the 970 regardless of if you recompile the code or not. You will just get more of a speed boost if you tell the compiler to produce code optimized for the 970 instead of for the G4 or a generic PowerPC.

Originally posted by jayscheuerle
So you're saying that it takes additional R&D to take out a SuperDrive or put in a lesser video card?

That's not the problem. The problem is that if a user percieves that they can get more for their money buying the $800 system and upgrading it (even if this is not true) than they can from buying the $1500 system, then they will buy the $800 system. This will eat into Apple's margins, which leads to less R&D dollars, which leads to less new technology, which leads to Apple not being able to afford to innovate (opinions on how well Apple has innovated is beside the point here). Yes, Apple will be selling more systems selling them at $800 than at $1500, but it is what is called the death of plenty - Apple would sell many many more systems, but because the margins are so much lower they make less profit, to the point that they actually end up losing money while selling more systems.

It is better to make 15% margins on 3% of the marketplace than it is to make 1% margins on 10%. Compaq, HP and even Dell learned this the hard way when they were each selling $500 systems with razor thin margins (<1%). They sold so many of these systems that they were forced to discontinue them because of the amount of money they were losing. It really is true, there is just some business that you don't want.