IBM 64-Bit PowerPC Details

[MacRumors]

Details of IBM's new PowerPC chip to be released in late 2003 are trickling in. The official Press Release is found at IBM's site, and provides little additional information on the upcoming processor.

Mr.Hey provides this article from Silicon Strategies which provides more detailed information on the 32-bit/64-bit implementation of the processor:

IBM's approach to implementing a 32/64-bit architecture appears straightforward. The 970 supports full 64-bit registers and addressing. When a flag bit is sent it triggers a 32-bit mode in which the high-order words on an arithmetic logic unit and on memory addresses are ignored. In either 64- or 32-bit mode, the processor issues up to eight instructions per clock cycle.

According to the article, 32-bit PowerPC OS's simply need to support new data structures and interrupt handlers, but 32-bit PowerPC apps would run unchanged.

The Register provides details from the Microprocessor Forum conference... and reports projections of SPECint2000 and SPECfp2000 scores of 937 and 1051 respectively at the 1.8GHz speed. According to this chart the 2.8GHz Pentium 4 currently has SPECint2000 and SPECfp2000 scores of 970 and 938 respectively.

 

[scem0]

UGGGGGGHHHHHHHH................ A year is too long. They expect me to be able to sit here on this 400 MHz AGP powermac for another year! This comp sounds to good to be true, but I still can't wait that long.

 

[ennerseed]

score projections...

well i guess thats a projection... and i hope its low, by the time it is released the pentium will probably be higher... athough i am sure it will rock my world

 

[MacAztec]

NOT SO GOOD

The Power4 and the Intel have just about the same scores. But the thing is, that Intel is OUT, and the Power4 is not. That means the Intel will just get better and better, and surpass the Power4. Same thing with AMD

 

[vniow]

But Intel and AMD don't have Altivec.

 

[e-coli]

Yes, but wouldn't those benchmark scores related to the Power chip be based on 32 bit software and commands?

Those scores should go way up with 64 bit addresses.

 

[FattyMembrane]

the pentium beat the power4 in one test and was only a little behind in the second. this is not good news. if the best ibm can do in a year is 1.8ghz with this thing, we're screwed. intel is not going to take a year long break so that we can catch up. i really hope that apple has something else up it's sleeve.

 

[e-coli]

Again, you can't compare 32-bit chips to 64 bit chips. These benchmarks are most likely based on 32 bit integers and executions.

Anyone here ever used a Flame / Flint / Inferno? Well, those chips aren't even 1 GHz 64-bit processors, and those things do real time high-definition renders. they're absolutely insanely fast.

Just be patient. These chips should me much faster than benchmark scores indicate. I think these types of scores will be deceptive when compared to real world performance.

 

[Nipsy]

Everyone seems to be missing the fact that those are processing only numbers.

What this processor is likely to have is vast bandwidth, which means it will help enormously in memory throughput, which is a bottleneck far more often than processor.

 

[gbojim]

Originally posted by Nipsy
Everyone seems to be missing the fact that those are processing only numbers.

What this processor is likely to have is vast bandwidth, which means it will help enormously in memory throughput, which is a bottleneck far more often than processor.

Add to that the 2 SIMD units and 16 way SMP capability. If they do it right, system performance with this CPU will be amazing.

 

[vniow]

Also remember that this chip is likely going to be used in Apple computers and we all know how they like to keep secrets.
I wouldn't be suprised if this chip is used just for the low end PowerMacs, Xserves & possibly some of the 'i' line.

Initial quantities will debut at 1.4 to 1.8Ghz, with 512kb of Level two cache. In his presentation, he (Peter Sandon) described these as "conservative" estimates.

IBM's and Apple aren't stupid, they know that Intel is oing to be way above 3Ghz by the time this gets released so I wouldn't be shocked at all if they have somthing a bit quicker up their sleeve.

 

[Hawthorne]

Given the new architecture in the PowerMac,

I would expect to see this chip sooner than later.

But that may just be the booze talking.

 

[solvs]

SPEC scores don't mean as much as most people think. I just want some real world performance by the time I'm ready to spend. The faster the rendering, the better.

 

[748s]

Originally posted by e-coli


Anyone here ever used a Flame / Flint / Inferno? Well, those chips aren't even 1 GHz 64-bit processors, and those things do real time high-definition renders. they're absolutely insanely fast.

.

i think those chips in sgi computers[that run flame etc.] just hit 500mhz. [are they 128 bit though?]. will a 64 bit mac finish off sgi?

 

[scem0]

Originally posted by edvniow
Also remember that this chip is likely going to be used in Apple computers and we all know how they like to keep secrets.
I wouldn't be suprised if this chip is used just for the low end PowerMacs, Xserves & possibly some of the 'i' line.





IBM's and Apple aren't stupid, they know that Intel is oing to be way above 3Ghz by the time this gets released so I wouldn't be shocked at all if they have somthing a bit quicker up their sleeve.

That is good thinking. I never considered that. And now that I do - I think you are right. We were all expecting more then 1.8 GHz, I think apple knows this and is going to suprise us. Well, we have a year to wait and see...

 

[vniow]

Originally posted by scem0


That is good thinking. I never considered that. And now that I do - I think you are right. We were all expecting more then 1.8 GHz, I think apple knows this and is going to suprise us. Well, we have a year to wait and see...

But also remember that Apple is a customer of IBM and they don't usually lie about these sort of things to stockholders and analysts, but I suspect that at Apple's request, IBM is keeping at least some secrets about this chip.

 

[bidge]

The exciting wait begins, I am too happy

 

[DaveGee]

If you wanna look at SPEC and I'm really not sure it's a good thing to do because benchmarks like SPEC don't usually tell too much but that being said.

Lets use SPEC number from CPUs we can wrap our minds around.

SPECint_base2000
=================
242 G4 800Mhz
259 G4 867Mhz
306 G4 1000Mhz
937 GPUL 1800Mhz
=================

SPECfp_base
=================
147 G4 800Mhz
153 G4 867Mhz
187 G4 1000Mhz
1051 GPUL 1800Mhz
=================

Sure looks good to me!

Dave

 

[dricci]

Here we go... People basing their entire lives and decisions on one stupid benchmark that doesn't even include the processor we're talking about, only it's big brother. Why are we putting down a processor that isn't even out yet? Screw SPEC, let's wait to see some real world performance when these things come out next year.

 

[ddtlm]

dricci:

How doesn't SPEC include the PPC-970? It includes all processors.

If you want to know how fast a computer can crank RC5, you run RC5. For Photoshop, you run Photoshop. For performance in general, you run SPEC.

Depite your refusal to accept it, SPEC remains the dominant cross-platform benchmark. Within SPEC, a whole lot of different algorithms are run that stress many different things. No company conrolls SPEC, and SPEC is not written for any particular processor.

 

[Telomar]

It does 18M key/second in RC5 I believe and that is with code that wouldn't be optimised for the chip.

Keep in mind that that is a conservative starting point. I believe early overclockers will find a lot more performance and more importantly it should scale considerably better than the G4 has.

Finally early expectations for the Itanium 2's SPEC results were significantly below the actual numbers that were seen upon release.

 

[etoiles]

Originally posted by edvniow
IBM's and Apple aren't stupid, they know that Intel is oing to be way above 3Ghz by the time this gets released so I wouldn't be shocked at all if they have somthing a bit quicker up their sleeve. [/B]

they might not be stupid, BUT: Apple manages to sell 'slow' computers now, and apparently they are among the few companies making profit in those difficult times.

so hold your breath, because they might just be 'lazy', meaning they lack incentives to really get their act together on the processor front. People buy them anyway, apparently for other reasons than raw speed.

I secretly hope to be proven wrong in the next couple of months...

 

[RogueLdr]

Originally posted by scem0
UGGGGGGHHHHHHHH................ A year is too long. They expect me to be able to sit here on this 400 MHz AGP powermac for another year! This comp sounds to good to be true, but I still can't wait that long.

They probably expect you to buy a current PowerMac now.
Although I agree that it would be incredibly frustrating to purchase a G4 based machine now, and have your doors blown off in a year's time. It's too long to wait for me too (400MHz Gigabit Enet), so I got a PowerLogix 800MHz upgrade as a stop gap. It's amazing how most of my complaints about OS X performance went away when I moved to a current processor!

RL

 

[3G4N]

Originally posted by ddtlm
How doesn't SPEC include the PPC-970? It includes all processors.

Ummm, Maybe because... uuhhh..
THE PPC-970 DOESN'T FRIGGIN EXIST YET!! Jeezus.
Where *didn't* you read the chip won't go into production
until late 2003??? Does IQ need to become part of the
registration here?

Originally posted by 748s
i think those chips in sgi computers[that run flme etc.] just hit 500mhz.
[are they 128 bit though?]. will a 64 bit mac finish off sgi?

Current IRIX-based SGI's (O2's, Octantes, etc.) use 64-bit MIPS
R14000A processors. And yes, they go up to 600mhz, with 4mb of L2 cache.
They whoopass and I would love one, fully loaded. SGI knows hardware,
and Apple could take some design cues from them (or friggin buy them out!)
It's just AMAZING what you can LEARN by going to http://www.sgi.com/

Thanks DaveGee for putting this processor in perspective.
I'm sooo damn tired of people whining about this. Go buy a P4 and jump off
a bridge if you're so damn upset. And I really wish people on these forums
would take just 2 minutes to research a topic before waxing and whining
ad nauseum, going off on crap that they don't know what they are talking
about. The product of American education.

Yeah, it'll be close to a year until this PPC970 is in your little hands.
Be patient!! Pull up a chair, a cold beer, and a bucket, and start yer droolin.

End rant.
Boy, do I need a job!

 

[moki]

Re: NOT SO GOOD

Originally posted by MacAztec
The Power4 and the Intel have just about the same scores. But the thing is, that Intel is OUT, and the Power4 is not. That means the Intel will just get better and better, and surpass the Power4. Same thing with AMD

You're right! They should stick with the G4!

SPECint_base2000
=================
242 G4 800Mhz
259 G4 867Mhz
306 G4 1000Mhz
937 GPUL 1800Mhz
=================

SPECfp_base
=================
147 G4 800Mhz
153 G4 867Mhz
187 G4 1000Mhz
1051 GPUL 1800Mhz
=================