Demented speculation...
While setting up a PowerBook for a user today i was thinking about Jaguar and what it means for future hardware. Then I thought about something interesting. What if Apple is planning on abandoning Altivec? Maybe not entirely but look at the focus this WWDC. Jaguar. And Quartz Extreme... it won't use Altivec for accelleration but the onboard graphics engine in the AGP card. Suppose for a moment that Apple has been clamoring to motorola for faster processors but Motorola has been turning a deaf ear to Apple, instead concentrating on useful embedded features like low power. IBM is a full partner in PowerPC. They have some high end designs in the pipeline. Apple can augment their (IBMs) research and development by collaborating on these designs. We know IBM has a SIMD unit in the wings. Whether or not it is as powerful as Altivec is not important. What is is the rest of the processor and if present Altivec code can still be processed on the SIMD unit (reverse engineering).
IBM can design a fast workstation processor for Apple. It needs a fast bus (RapidIO) with fast int and fp units. And lots of them. With Core-Connect they can design a processor in less time than what conventionally be done.
iMacs, iBooks, and maybe Xservers can still use the G4 because it has low power feature attractive to that market. But IBM might not be out of the running yet.
IBM can design a fast workstation processor for Apple. It needs a fast bus (RapidIO) with fast int and fp units. And lots of them. With Core-Connect they can design a processor in less time than what conventionally be done.
iMacs, iBooks, and maybe Xservers can still use the G4 because it has low power feature attractive to that market. But IBM might not be out of the running yet.
Comments
[dream]
G5, G5, G5
[/dream]
I think all moto needs is a kick in the rear to get them going. I'm not sure how Apple's gonna do it, but they gotta do it sometime. Arn't those Apollo's scale-able to like 1.3 ghz or something? That sounds like something doable.
OMG <img src="graemlins/surprised.gif" border="0" alt="[Surprised]" /> Big idea <img src="graemlins/surprised.gif" border="0" alt="[Surprised]" /> Just came to me!!
It seems reasonable to use these 1.3 ghz and below G4s in the powermacs, and then unvail a new line packed with G5's. Make them like super-duper pro computers. Apple would have to scale down the price though on the Powermacs. Does this seem reasonable? (i know this idea doesnt really belong here, but i had to type it ) <img src="graemlins/embarrassed.gif" border="0" alt="[Embarrassed]" />
I don't know much about this oft-mentioned Power4 Chip, but it doesn't sound like that's likely destined for the Mac any time soon.
The only area where the Mac holds a speed advantage now is Altivec. If they abandon it, they won't even be able to win those Photoshop bake-offs.
I guess desparation breeds wild speculation.
<strong>Why does anyone think IBM will be better than Mot? The 1GHz G3s have been discussed now for over six months but are nowhere to be seen, whereas Mot is already shipping a 1GHz product.
</strong><hr></blockquote>
Its probably because the current iBooks are running at 600MHz. Apple ususally bumps them up 100MHz per revision so assuming 6 months for a speed bumps figures out to 2 years for the 1GHz G3 in the iBook.
I'm still hoping that they put the G4 in this next rev iBook along with a 32Meg graphics board so that it'll run OS X much better.
However it pans out, I expect to see IBM playing a larger role in Apple's future. But I'll go out on a limb and speculate that Motorola does not have the stranglehold on the Altivec IP that everyone supposes. I vaguely recall hearing that the original spec came from Apple. It would certainly be nice for us if IBM came out with some Altivec-compliant SIMD goodness. And I can't help but think that it would be a sensible move for IBM as well. Apart from Apple, I'm sure that Cisco would be interested in such a product.
There's so many theories and counter theories on the net that I'm punch drunk with them.
I liked the intriguing tone of Outsider's post.
Maybe the 'Raycer' buyout will result in a tech' implementation that will bear fruit in 'Power'macs soon. Acceleration of key multimedia tasks like 3d and digital video.
Perhaps IBM has a processor that can replace current offerings? There's been smoke and speculation that IBM will surplant Moto in Apple's high end.
Most AMD/Intel offerings are ahead of any G4 altivec or not, most of the time. Any decent new chip makes Alivec redundant. If it's still a G4 they'll stay with it.
They'll probably stay with 'it' or some other SIMD type of implementation because anything that shares the load is takes pressure of the main cpu, G5 or not.
The G5 is either 'killed' or 'imminent' depending on which rumour site you read.
There's the Apollo moves to .13 micron and we get 1 gig to 1.4 gig processors which would be the final speed bump until we move to a 'G5' type processor?
According to Mosr (and like all rumour sites with 'experts' on them...including this one...take it with a sack of salt...) expect a 100% boost in performance and new cases. That doesn't mean dual G5s necessarily but perhaps a 1.4 gig Apollo with 4 megs of ddr level 3 cache? DDr? Bus speed also improved. So over hall improvement?
I think the latter is more likely for New York.
I wouldn't rule out Outsider's theory. Apple are always up to something. And most of the 'experts' on this board had ruled out anything like Quartz extreme. But when we got it there was screaming about how their mac didn't support it!
But hey, here we are.
Once dreamed about us having Maya...and now we got it. (...and I bought Lightwave 7 instead
With everything else unveiled it's ibooks and Powermacs for Newyork? What else? Some video based iPod? I dunno.
I'm guessing Apple have got something up their sleeve...
Lemon BOn Bon <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" />
[ 05-18-2002: Message edited by: Lemon Bon Bon ]</p>
<strong>
Most AMD/Intel offerings are ahead of any G4 altivec or not, most of the time. Any decent new chip makes Alivec redundant. If it's still a G4 they'll stay with it.
They'll probably stay with 'it' or some other SIMD type of implementation because anything that shares the load is takes pressure of the main cpu, G5 or not.
[ 05-18-2002: Message edited by: Lemon Bon Bon ]</strong><hr></blockquote>
Intel and AMD where the first to introduce SIMD unit in their chips : mmx, mmx2, SSE, 3Dnow, SSE2. I don't think it's only marketing, even if there was a lot at the beginning (the problem now is that there is too many different sort of SIMD unit for the developpers)
Nothing can replace altivec except a Altivec compatible SIMD unit. On their roadmap , IBM said that the Sahara2 will have a SIMD unit, but without any further precisions.
Developping an altivec compatible SIMD unit is possible : but it's a lot of money, and IBM break the AIM alliance because he did not want of it. And now IBM is supposed to support all the cost of the R&D of a altivec compatible unit, after have saying that this tech was useless ?
Many people say they want auto-AltiVec coding to happen because of the extra work they would incur in using it by hand. Others say its not applicable to their field. And then others simply don't realize that its already pervasive in OS X bits and pieces and even OS 9.
There are already high level APIs for signal processing using AltiVec in OS 9 and OS X, but unless you need that it wont matter to your app (anything not an MP3/DVD/Video encoder).
Scientific apps are really making use of the G4, its just such a vertical market it doesn't filter down to mere mortals so no one is aware of its real world use.
Theres your reasons "why". Lest you all forget Apple totally snubbed Sherlock the past few years and suddenly its a near web browser in Jaguar.
You cant please all of the people all of the time, welcome to reality.
<strong>
Intel and AMD where the first to introduce SIMD unit in their chips...</strong><hr></blockquote>
No, x86 was not first. The first SIMD implementation was SUN's VIS (Visual Instruction Set) in the UltraSparc series CPU.
<strong>There is no future for Apple without altivec : do you think that Final cut pro is very efficient on a G3 ? How many hours of developpement Apple have invest in the altivec optimization of FCP, is apple ready to say that they downgrade their soft , and leave behind some of their realtime effects ?</strong><hr></blockquote>
didn't anyone else find it interesting that when they introed the g4 they didn't call the unit "altivec" they called it the "velocity engine." my guess is that, besides marketing, there was a reason not to use the moto trademark, right? just in case, they wanted to call something else "velocity engine" without having to explain that whatever comes after it is not the same as "altivec."
In a few 'mono' tasks like 'blast' it's great. But for most of the time? It's not even being used.
In that case, you can 'drop it'. Or modify it. Or replace it with another 'velocity engine'.
It just doesn't necessarily have to be Moto's simd unit.
A dual core G3 would knock the snot out of the current G4 most of the time.
So, big deal altivec.
IF Apple can do Quartz Extreme for 2d/3d in the OS most of the time...Blast for scientists...maybe the Raycer thing is not about hardware but a realised software technology to more effectively use the power in 3D graphic cards. It was a patented tech'? If that tech' could boost 3D by a factor of 5 to 10 then I'd take that over altivec anyday.
There's something about the way Apple executed the tech' on Quartz Extreme and on Xserve that suggests they've got something up their sleeve for Macworld Newyork other than merely labouring to a 1.2 dual gig 'Power'mac.
Maybe, again, they've found a way to boost real world performance in 2D and 3D apps like Photoshop and Lightwave.
By the time the 7,500 comes in January (please...) then I think Altivec will be the icing on the cake, not the current G4 processor performance crutch that it currently is.
presently, take away 'altivec' and the G4 comes up embarrassingly short.
lemon bon bon