IBM Power4 PPC AltiVec

[arn]

PenguinPPC is reporting that IBM's new vector unit is AltiVec compatible and is called VMX.

No source is reported.

The new processor is reported to be under consideration by Apple.

[Ensign Paris]

That it is pretty cool, I hope we see these processors, although I don't see any specific MHz (or GHZ ) yet, maybe I need to look harder.

Ensign

[DannyZR2]

HELLZ YEAH!!!

[Jays]

this is gearing up to be the best news for us in a long time, I fear with these speculations the new powerMac's will be a bit disapointing, and it won't be befor 2003 that apple will catch up with the pro line.

I for one will not be buying another moto G4 for now...

[gotohamish]

Should the IBM chip be backward compatible with Altivec it would obviously be a great boost for Apple, but also the fact that Motorola and Apple use different names might help too:

Motorola have Altivec, Apple call it the Velocity Engine. The casual Apple fan who might not be processor-learned might not know Altivec, and if Apple switch chips and STILL have the 'Velocity Engine' on board, who's to know the differnce on paper?

Difference on the desktop is another thing, and I frankly can't wait.

[barkmonster]

When I finally ditched the Atari and got my first mac I went from a 16bit cpu to a 32bit cpu. It would be great if after the 4 years or so of having my current mac that my new one makes the leap to 64bit!

I can't afford a new mac this year anyway, assuming apple bring these out sometime next year, by the time I've got enough saved for the current entry level I could end up getting something really powerful for my money. I'm looking forward to seeing how these IBM rumours develop.

One thing that's impressed me most about the specs on the IBM chip is this;

The G4 is 4 x superscaler

The IBM chip is 8 x superscaler

This means a maximum of 4 instructions per clock cycle, it manages 2.31 in realword terms so with an 8 x superscaler chip we could be looking at nearly 5 instructions per clock cycle. That would mean apple can return to the days of the G3 and Pentium 2, understating the genuine pentium crushing performance and shaking up intel. from what I remember, when intel performed their own tests on the G3, it showed the powermac is even faster than apple claimed it was

[Beej]

This sounds great.

But VMX? Sounds too much like MMX for my liking...

[Blitzkrieg]

Originally posted by Beej
This sounds great.

But VMX? Sounds too much like MMX for my liking...

VMX was the original name of Altivec:
Vector Multimedia eXtension

And developed by IBM, by the way...

[sageenos]

Now maybe all the switch to x86 threads will die off...

[Rocketman]

Other links on the PenguinPPC site also go into various linux installs onto Powermacs and X-serve. One such link (http://penguinppc.org/) reports:

"Another One Bites the Dust
July 24, 2002

Ben Herrenschmidt has pounded 1U of Apple hardware into submission. The Xserve has booted and run
GNU/Linux!. There are still some kinks to work out, and your favorite distribution's installer may need some tweaks to install (check with them). Support in the kernel source trees is forthcoming. Word is, running in uniprocessor mode GNU/Linux returned data from mysql databases 4 times faster than OSX running in SMP mode. This was with the same datasets, same queries, same mysql version, and same compile options."

The point I am raising is Apple has apparantly not optimized the software yet for maximum speed since a single linux hacker got a recompile to run 4x faster in several days on a particular database call test.

This might indicate how Jagwire has achieved such substantial speed increases in several areas and an indication of lots of good things to come as
hundreds of elements of OSX are updated. One wonders if they are doing 64 bit versions at the same time as 32 bit so when a Power4 derivitive comes out they already have OSX64 10.3 (Warp 10) handy.

Rocketman

Flip those bits.

[ibjoshua]

Originally posted by sageenos
Now maybe all the switch to x86 threads will die off...

we should be so lucky!

there are some serious AMD fans lurking in these corridors.

josh

[G4scott]

This sounds good, seeing as how Apple has dedicated themselves to Alti-Vec, it'd be pretty stupid to dump it...

[Mr. Anderson]

Even though this info was mentioned before, its still great news. I'm really curious what the specs of the intended intially delivered CPUs will be - again, great news, but its all potential - I hope Apple doesn't drop the ball here.

D

[dhdave]

I just saw this this morning

Of particular interest:

"What I find is interesting is the fact that IBM can talk about it. If there was committed Mac design, you know (Apple CEO) Steve Jobs would have his hands around IBM's neck not to talk about this chip," said Kevin Krewell, a senior analyst at In-Stat/MDR. "The fact that IBM is talking about it indicates to me that it's not a mainstream Apple product at this time."

That to me, is VERY interesting. It is possible that Apple is just biding it's time, but then again...

The article goes on:

The current bleak economic climate, which has undermined Apple's sales and profits, and the likely high costs of introducing such a new platform may be deterring the company from making the leap from 32-bit to 64-bit computing.

"I believe it's possible that it's too expensive at the moment for Apple to commit to it at this point and time," Krewell said.

Nevertheless, IBM's disclosure will likely cheer supporters of the PowerPC design by showing the architecture evolving to more powerful implementations.

Count me among those. I love the PowerPC. I hope Apple decides IBM's is worth using. I for one would pay MORE for a better internal design. Screw the enclosure. They can use the quicksilver case for the next 10 years for all I care. Just give me a workable architecture with a roadmap of steady improvements in CPU speed and memory bandwidth that makes the Pentium look like the old, piecemeal design that it really is.

dh

[Bear]

I find it very likely that IBM would be talking about a chip Apple is contract to use.

IBM is a much larger company than Apple and use the PPC chip in its own machines. Apple is Motorola's biggest PPC chip customer. So while Apple may get to control the news Motorola reports, IBM would never agree to delaying announcements for another company when it needs the chip for itself.

This of course is just my 2 cents on the issue.

[cyberfunk]

Well, if this is the G5 or it's replacement from IBM, I'd be most happy ! I think that these chips sound a lot like they could be used in Apple's boxes, quickly, unlike the Motorla embedded 7500s that dont have altivec, and lack many other things that are integral to a desktop computer. Hopefully Apple will adopt this almost "off the shelf" soluion, as it seems to need not that much tweaking to work with their stuff.

[cyberfunk]

Originally posted by Bear
I find it very likely that IBM would be talking about a chip Apple is contract to use.

IBM is a much larger company than Apple and use the PPC chip in its own machines. Apple is Motorola's biggest PPC chip customer. So while Apple may get to control the news Motorola reports, IBM would never agree to delaying announcements for another company when it needs the chip for itself.

This of course is just my 2 cents on the issue.

I agree completely

[Blitzkrieg]

Found this link yesterday:

http://www.ece.cmu.edu/~dianam/ics02/ics-3.pdf

A research paper (dated back to June 23, 2002) from IBM about an LPX processor (low-power with VMX - tapeout early 2003), about circuits running at 3.3 GHz up to 4.5 Ghz and so on. In one graph this processor compares to a Power4 - sounds interesting to me

[G4scott]

I know how this processor can get so much publicity and still be used by Apple.

When Apple introduced the G4, they were the only ones supporting it. Motorola didn't say anything about how fast it was. It's hard for Apple to 'sell' a processor that nobody's ever heard of.
If Apple is planning to use this processor, they're letting it speak for itself. By having IBM get so much publicity with it, people will think "Man, that's one kick @$$ processor!", and then Apple can put it in their desktops, and people will already know that it can really kick major pee-cee @$$...

This may be a new strategy from Apple. You don't see dell or gateway showing off how fast the p4 is. They just have to say "intel inside" and let intel do the advertising. It would make it much easier for people to see how fast these machines truly are.

A PowerMac with a Power4 will scream!!!

[jadam]

uhhhhhhh did you JUST call an Apple computer a pee-cee?? no you didnt.....................


anyways. Apple will use these chips, and IBM is tooo big of a company to listen to apple and not disclose details about this chip.. For gods sake, when did we know about the 1ghz G3s?? when did apple use them???? they havent, and IBM is supposedly coming out with 1.5ghz G3s also(with VMX maybe please IBM ... "keep on dreaming"....) Apple cant force IBM to shut up. Doesnt IBM make around 100billion a year?? they make 5x what MS makes for gods sake... and who knows how much more than apple.

and HAHAHA to all of you saying Apple is safe because it has a few single digits billions in hte bank, look at what IBM has.

Anyways, i need a powermac anyways, this ibook isnt the most amazingly super super fast thing yah know(but then again theres X.2) heh.

[luiss]

Given all we know (or really the little we know) about this chip, why would Apple NOT want to use this chip?
Of course, we don't know alot of things about this chip, but assuming...[list=1][*]Current software will not need a recompile to run[*]The vector units are Altivec compatible.[*]The chip is real (not still being designed/debugged)[/list=1]
Why would Apple not want to use this chip?
I don't know, just starting a thread...

[ffakr]

People, please read the Microprocessor report blurb again. It clearly states this is a POWERPC, Not a Power4 processor.
The posts, especially in this thread, seem to finally be recognising this but there are still people calling the new chip a Power4. It isn't.
It uses the PowerPC instruction set, with design cues from the Power4 processor.

This will be a HUGE difference when it comes time for Apple to implement this processor.
The Power4 ISA is a cousin of the PowerPC ISA. OSX and all of the apps would need to be recompiled to run on a Power4. This chip just needs hardware support and we should be able to use it.

... sorry for being grumpy...my back hurts



Other topic... If this thing is 8way superscalar, it is going to need a lot of load/store and excecution units. It is going to be a big honking processor.
Don't expect dualies any time soon... Maybe when IBM starts cranking them out on their .09 micron plant next year. Not that it will matter though... it will likely have twice the IPC (instructions per Clock) of the G4 and It will probably be clocked higher than the G4. yum yum

[Rajj]

I wonder what is the micron level on that chip and what are the processor speeds?

[Catfish_Man]

Originally posted by ffakr

Other topic... If this thing is 8way superscalar, it is going to need a lot of load/store and excecution units. It is going to be a big honking processor.
Don't expect dualies any time soon... Maybe when IBM starts cranking them out on their .09 micron plant next year. Not that it will matter though... it will likely have twice the IPC (instructions per Clock) of the G4 and It will probably be clocked higher than the G4. yum yum

Actually, with a G4+ you have:
2 simple integer
1 complex integer
1 floating point
4 vector
= 8 way superscalar
The G4+ is a 4 (or maybe 3 I can't remember, I think it's 3) issue 8 way superscalar processor.
btw, why do people care about this thing's clock frequency? It's got 6.4GB/sec of memory bandwidth, we don't need clock frequency!

[drastik]

yeah, but the masses want a high clock, and I say just give it to them. Steve comes out and gives a wink to the loyal and knowing and then tells the world they can get a 4 GHz Mac, it would be a good thing. I don't sweat hertz, but ma and pa kettle do.