IBM's Power 970

MacRumors

macrumors bot
Original poster
Apr 12, 2001
46,727
8,945
Arstechnica walks through some of the technical aspects of the upcoming IBM 970 as compared to current processors:

In the following article, I'm going to step through the stages of the 970's pipeline, much as I did in my previous G4e vs. P4 articles. I'll talk about instruction fetching, decoding, dispatching, issuing, execution and completion. I'll also cover some of the other interesting elements of the 970, like its 900MHz DDR bus.
 

gaomay

macrumors regular
May 28, 2002
116
0
Scotland, UK
Heavy stuff

Wow, that's pretty heavy stuff! I think the upshot of it all is that this nis a dman fast processor, designed specifically for SMP. Can't wait to see it in a MAC!
 

wdlove

macrumors P6
Oct 20, 2002
16,568
0
Preliminaries: die size, power consumption, and clock speed
Process Die Size _Transistors Core Voltage Power Dissipation
PowerPC 970 1.8 GHz 0.13um 121 mm2 52 million 1.3v 42 Watts
Pentium 4 2.8 GHz 0.13um 131 mm2 55 million 1.525v 68.4 Watts
G4e 1 GHz 0.18um 106 mm2 33 million 1.6v 30 Watts


As you can see from the table, the 970 at 1.8 GHz is much closer to the G4e than to the P4 2.8 GHz in terms of power dissipation. This means that Apple will be able to use this chip in the kinds of innovative enclosure designs that make their hardware continually appealing, regardless of how it performs. Furthermore, a 1U, 970-based version of the XServe is not out of the question. And if you consider the fact that the 970's power consumption at 1.2GHz is a mere 19W, it's almost certain that we'll see a future notebook from Apple based on the new chip.

I hope it comes to the desktop Power Mac at the same time & soon!
:)
 

TheT

macrumors 6502
Jan 5, 2002
485
0
Germany
I realize that in some sense it's not "fair" to compare a recently announced 64-bit processor that won't ship for at least a year to two 32-bit processors that are currently on the market.
At least another year... great. And that'll be 1.8GHz. Compared to 2.8GHz today. What will a Pentium be in a year? 5.6GHz (Intel actually does twice the speed every year, not like Motorola every 3 years...)
I love macs, but this is just BS!
 

locovaca

macrumors regular
May 14, 2002
187
125
Iowa
It's a good article. Just like other "parallel" style processors with shorter pipelines, the one question will be how high it clocks. No matter how well the 1.8 GHz performs when it's released, if it hasn't moved up a bit in a year then Apple will be well behind once again. Hopefully IBM will prove to be better in this than Moto, but this is by no means a simple processor design like the P4!
 

Caravaggio

macrumors newbie
Mar 10, 2002
6
0
Ars is one of my favorite sites. Its wasn't long ago that they didn't cover much Macintosh issues at all. Now they do them in excellence. Props to those guys over there (not to take any away from MacRumors, of course!)
 

Shrek

macrumors 65816
Jul 23, 2002
1,118
0
Nashville, Tennessee USA
Originally posted by TheT

At least another year... great. And that'll be 1.8GHz. Compared to 2.8GHz today. What will a Pentium be in a year? 5.6GHz (Intel actually does twice the speed every year, not like Motorola every 3 years...)
I love macs, but this is just BS!
Take a look at this:

The Intel® Itanium® 2 processor is uniquely architected for demanding enterprise and technical applications. Itanium® 2-based platforms enable businesses and organizations to maximize their investments by delivering industry leading performance at lower cost with greater choice than proprietary technologies.

Available Speeds: 1 GHz, 900 MHz
Cache Level 3: integrated 3 MB or 1.5 MB
Level 2: 256 KB
Level 1: 32 KB
Features
Based on EPIC architecture
Enhanced Machine Check Architecture (MCA) with extensive Error Correcting Code (ECC)
Operating system support: HP-UX*, Linux*, Windows*
System Bus 400 MHz, 128-bit wide
6.4 GB/s bandwidth
Chipset Intel® E8870 chipset, OEM custom chipsets
 

jefhatfield

Retired
Jul 9, 2000
8,803
0
Originally posted by TheT

At least another year... great. And that'll be 1.8GHz. Compared to 2.8GHz today. What will a Pentium be in a year? 5.6GHz (Intel actually does twice the speed every year, not like Motorola every 3 years...)
I love macs, but this is just BS!
sure, intel will be way, way ahead in clock speed, but the ibm 970 may be an option for apple and better suited than the next generation motorola processor

we will just have to wait and see and see which direction apple goes with this information
 

tychay

macrumors regular
Jul 1, 2002
219
29
San Francisco, CA
Nice flamebate

At least another year... great. And that'll be 1.8GHz. Compared to 2.8GHz today. What will a Pentium be in a year? 5.6GHz (Intel actually does twice the speed every year, not like Motorola every 3 years...)
I love macs, but this is just BS!
Hmm, a traditionally pro-PC site writes a highly involved article on the design tradeoffs of a chip that won't be out for another half year and a mac user says that this chip is "BS".

In the next paragraph, the author explains that the comparison is done to compare architectures not performance. The assumption here is that Intel's offering on the desktop will only be a faster rendition of the Pentium 4, else he would have choses the more relevant Itanium or Opteron/Hammer or Pentium 4 XEON. The author wished to discuss design tradeoffs against other desktop CPUs that the readers might be familiar with.

Intel does not "twice"[sic] the speed every year. I think it is closer to doubling every year and a half. The problem is that Motorola's 25% speed boost in almost a year is pathetically slow by industry standards; not that Intels clock speed gains are incredibly impressive. Because of its extra long pipeline with spare stages thrown into the P4's design to account for anticipated wire delay, it would be surprising that IBM would have a chip that was as fast as a P4 in terms of clock speed, especially since it uses the same process, with smaller stage pipelines, and will be out in quantity in a year.

These speed gains do not come without a penalty (which you'd realize if you read some of the author's referenced articles). This is why the P4 was deservedly panned when it first came out--it was a worse performer than the PIII. The P4 design gave it room to grow (by increasing clock speed) and now it isn't so panned. It's not universally loved, either.

As a supposed mac lover, you'd also realize that clock speed is not the end-all/be-all of performance. While it is a factor: if you read the article, you'd see that PowerPC designs are designed more like the inside of McDonalds (many short lines served slowly serving multiple people at ocne) than the Pentium-4-like drive thru (one long line served very fast). The interesting thing here is IBM is trying have it both ways (multiple lines served fast).

How fast is an Itanium? (much slower than a Pentium 4). How much does it cost? (more)? Yet despite the slower clock speed, Intel seems to think that it is a good enough performer to justify the premium.

Every day, R&D departments are buying Athlon computers for HPC even though they're "slower" than the Pentium 4. People still buy Pentium III's for low-end servers despite the fact they are slower (though I believe this will flip to Pentium 4 XEONs next year). Right now, much smarter people than you an me spend $60k+ on n-way pSeries servers from IBM with a chip with a similar design to the PPC970 running at <1.5GHz. Ask yourself why?

As to whether or not this will compete with Intel P4's or AMD Opterons--only time will tell. Right now, such a comparison is "BS" just like saying "the IBM PPC 770's 1.8GHz is BS" is itself a bunch of "BS".

Nice flamebait. Try again.
 

ogun7

macrumors regular
Sep 20, 2001
187
57
Originally posted by TheT

At least another year... great. And that'll be 1.8GHz. Compared to 2.8GHz today. What will a Pentium be in a year? 5.6GHz (Intel actually does twice the speed every year, not like Motorola every 3 years...)
I love macs, but this is just BS!
Once again, we have a "Mac fan" that has TOTALLY missed the boat in what the power of this processor really means to desktop computing and is completely blinded by MHz. If 64-bit processing, 2 Altivec units, 200 instructions moving through the processor at any given time and up to 8 way SMP aren't enough for you, then get a SGI or Sun workstation, and stay away from Macs.

I guess I took the bait.
 

MrMacMan

macrumors 604
Jul 4, 2001
7,002
11
1 Block away from NYC.
Well again the next generation processors come a little late to the computing party...but heck it will be a good show that macs too can makes these processors. :D
Wait How is the processor gonna support 900MHz DDR ram??? :confused:
What is the current ram speed, aint it much slower? Aint system ram up to like 533 now, a far cry from 900...:rolleyes:
 

eric_n_dfw

macrumors 68000
Jan 2, 2002
1,507
55
DFW, TX, USA
Originally posted by MrMacman
Well again the next generation processors come a little late to the computing party...but heck it will be a good show that macs too can makes these processors. :D
A little late for what?
 

Hawthorne

macrumors regular
Jul 1, 2002
198
0
In front of my Mac
Re: Nice flamebate

Originally posted by tychay

see that PowerPC designs are designed more like the inside of McDonalds (many short lines served slowly serving multiple people at ocne) than the Pentium-4-like drive thru (one long line served very fast). The interesting thing here is IBM is trying have it both ways (multiple lines served fast).
That is absolutely the best and easiest way I've ever heard to describe the differences between PowerPC and x86 architecture. Very nicely done, tychay.
 

MrMacMan

macrumors 604
Jul 4, 2001
7,002
11
1 Block away from NYC.
Hm... the real speedy processors we see coming before this processor.
In real life situations the P4 smokes The G4 for example. Don't go all Bull**** Photoshop on me...:eek:
The processors coming may match the Power 970 but I just hope we are not late to get some of the market share coming off this.
 

JamesDP

macrumors member
Jul 24, 2002
52
0
Nicely put, tychay. I think a lot of people miss the point when it comes to the whole Apple vs. PC debate. Yes, PC makers and Apple are both vying for the personal computer market, but they do it in entirely different ways.

In a market as competitive as the PC market where everybody and their grandma can put together a white box and sell it to Joe Q. Public, then you need to come out with newer, faster computers every couple of months that are above and beyond the previous version enough that someone's gonna look at the specs and say, "Hey, I need a faster computer." Truthfully, they don't need a faster computer - they fall prey to marketing, advertising, and insidious software makers who continue to bloat programs with unnecessary "features" and poor coding that require a faster computer to make sense of the muckety-muck inside.

Apple, while selling to the "PC Market", is essentially competing with no one, hence the less frequent updates to computer speed, which they don't really focus on in the first place. Apple's focus is on integration (although not in the twisted sense that Microsoft does) of software and hardware, a seamless user experience, features (in particular, their "i" programs), and the "digital lifestyle". And they are good at what they do. I switched two months ago and I wouldn't go back to PCs for anything. The joy of computing just isn't there. I don't dread doing work on my Mac the way I did for my home PC and the way I still do for my work PC. PCs just don't offer that experience, no matter what the pseudo-IT-genius-uber-geek-"well-you-must-be-doing-something-wrong"-crowd say. I'm a 20-year PC user. I know what I'm doing. PC makers and PC software makers, however, do not.

And personally, I think trying to compare Macs and PCs, while no doubt fun for your average troll or philosophical type, is a waste of time. While they do similar things, they go about them in dissimilar ways. And trying to compare speeds between the two based on something like MHz/GHz is not even apples and oranges, apples and dog crap or apples and skyscrapers. If Macs ran PC software natively the same way PCs did and the only difference was the CPU and internal components, then I could understand people being outraged over such a huge gap in MHz/GHz. But they don't. Macs run Mac software, and they run it well. While there is a large numerical difference between PCs and Macs, the apparent speed (at least in my experience so far) is similar, if not equal, and that's all that truly matters. End of story. Benchmarks are meaningless. It's what you see on the screen that matters. Some Photoshop filter takes 20 seconds longer on a Mac than it does on a PC? Even if you do it dozens or hundreds of times, do you notice? Are you calling up Adobe and asking for those 20 seconds per filter back? Are you that ridiculous? Let's hope not.

They're just computers, people. Tools. They won't make you feel better about yourself, they won't make you more attractive to women, they won't make your junk bigger. They're just computers, and frankly the need of some people to cause contentions over something so purely meaningless to our lives as processor speeds or benchmarks between Macs and PCs or whatever is so sad that vocabulary to describe it escapes me.
 

springscansing

macrumors 6502a
Oct 13, 2002
922
0
New York
Originally posted by JamesDP

They're just computers, people. Tools. They won't make you feel better about yourself, they won't make you more attractive to women, they won't make your junk bigger. They're just computers, and frankly the need of some people to cause contentions over something so purely meaningless to our lives as processor speeds or benchmarks between Macs and PCs or whatever is so sad that vocabulary to describe it escapes me.
Get off the soapbox man. I am an audio engineer, and I need to do REAL TIME processing. The slower my computer, the less instruments and effects I can run in REAL TIME. Speed doen't matter for photoshop... 20 seconds is no big deal. But when you need to run things in real time, it DOES make a BIG difference. I had to buy a new 867 because my dual 450 simply couldn't run what it needed to in real time at 44100hz.

Audio is a world where its often can do or can't do. Not can do or can do 20 seconds later.
 

tychay

macrumors regular
Jul 1, 2002
219
29
San Francisco, CA
Re: Re: Nice flamebate

Originally posted by Hawthorne


[PowerPC is like the inside of McDonald's while Pentium4 is the drive thru analogy] is absolutely the best and easiest way I've ever heard to describe the differences between PowerPC and x86 architecture. Very nicely done, tychay.
I really wish I could take credit for the McDonald's analogy, but I read it somewhere a year ago. After a bit of hunting, I found it in an article written by the same author! (scroll to the bottom) This page should be required reading for anyone who has had to defend their platform choice based on performance against Intel-bigots. Even if you're like me and have more Intel boxen than Macs.

Has anyone else who read the article found it interesting that IBM has chosen to "crack" some of the PowerPC operations into smaller ones? After all, the whole point of RISC was to have all instructions be one cycle and avoid the one criticism the author has with the CPU (grouping of uops creating nop bubbles).

It's amazing that twice in one day I find myself referencing Ars Technica. If you told me this four years ago, I wouldn't have believed you.

Take care,

terry
(some of my best friends work for AMD or Intel)
 

copperpipe

macrumors regular
Jul 9, 2002
127
0
and...?

what do you think of your new dual 867? IS it fast enough?...

One thing in these articles that should put any speed freak at ease is that these new powerpc chips are designed to get together (can you say quad?) and have a party, while the x86's are not. Or at least that's my understanding of it. And OS X will be there to take full advantage of it. So do you really think whatever intel has out in a year can take on two of these monsters working together, much less the possibilty of four of them? I very seriously doubt it.
 

Catfish_Man

macrumors 68030
Sep 13, 2001
2,579
1
Portland, OR
Re: and...?

Originally posted by copperpipe
what do you think of your new dual 867? IS it fast enough?...

One thing in these articles that should put any speed freak at ease is that these new powerpc chips are designed to get together (can you say quad?) and have a party, while the x86's are not. Or at least that's my understanding of it. And OS X will be there to take full advantage of it. So do you really think whatever intel has out in a year can take on two of these monsters working together, much less the possibilty of four of them? I very seriously doubt it.
Correction: Intel's desktop chips are not designed for multiprocessing. AMD's chips do a moderately nice job. Their new chips are going to do a very nice job. The 970 will also do a very nice job, but will be hard to make a northbridge for. I wouldn't be worried about the Pentium 4 if we get dual processor or dual core 970s (dual core seems like it should be very practical when they switch to .09 micron). I would be more worried about AMD's chips (assuming they're still in business).
 

matznentosh

macrumors regular
Apr 11, 2002
144
0
Re: Nice flamebate

Originally posted by tychay


How fast is an Itanium? (much slower than a Pentium 4).

what makes you say the Itanium is slower to desktop Intel chips? I thought most sites consider the Itanium to be superior at processing power (not superior in cost or heat generation of course).
 

tychay

macrumors regular
Jul 1, 2002
219
29
San Francisco, CA
Re: Re: Nice flamebait

Originally posted by matznentosh


what makes you say the Itanium is slower to desktop Intel chips? I thought most sites consider the Itanium to be superior at processing power (not superior in cost or heat generation of course).
Thus proving the point. As another poster mentioned, the Itanium runs at 900Mhz or 1Ghz. Yet you mention that most sites consider the Itanium to be supperir at processing power. Let's put this together as an inconsistent triad:
  1. Speed is equal to performance (or that speed is the most important indicator of performance).
  2. The Intel Pentium 4 is faster than the Itanium (3x the clockrate).
  3. The Itanium is superior at processing power.
    [/list=1]

    One of these has to give. If I work at it, I can come up with a benchmark (albeit contrived) to "prove" (3), but isn't it easier if we relax (1) by adding "work done per clockcycle is just as important" and perhaps "optimizing for different tasks lead to different design goals"?

    Recall my original point was, if (1) is true and the IBM PPC970 running at 1.8Ghz is "BS", then Intel should not be selling processors today that are 1/2 the clockspeed of the PPC970 for more than the price of a Pentium 4. Therefore (1) is naive and it is deliberately misleading to compare two different types of processors based on speed.

    The problem is IBM already has a processor for the Itanium's market (the Power4). So the PPC970 must be targetted for some other (lower end) market. Those markets (and the CPUs targetted at that market in a year) are the low-end server market (Intel Pentium 4 XEON/AMD Opteron/Motorola G4e+updates), the desktop market (Intel Pentium 4/ AMD Athlon/Motorola G4e), and the notebook market (Intel Pentium 4/Motorola G4/IBM G3 Sahara/Transmeta Crusoe). The spec sheet says it can run in any of these three markets, price and availability will determine which markets it enters into.

    Is this so unreasonable?
 

alex_ant

macrumors 68020
Feb 5, 2002
2,473
0
All up in your bidness
Originally posted by JamesDP
In a market as competitive as the PC market where everybody and their grandma can put together a white box and sell it to Joe Q. Public, then you need to come out with newer, faster computers every couple of months that are above and beyond the previous version enough that someone's gonna look at the specs and say, "Hey, I need a faster computer." Truthfully, they don't need a faster computer - they fall prey to marketing, advertising, and insidious software makers who continue to bloat programs with unnecessary "features" and poor coding that require a faster computer to make sense of the muckety-muck inside.

I can't tell whether or not you're taking apart your own argument here. You say that few people really need these newer, faster computers, but then you say that software developers continue to add bloat to their programs which requires progressively faster and faster computers. Of course a 486 will run the lean-and-mean Windows 3.1 and all appropriate software just as well as it did eight years ago, but the problem is that nobody wants to use Windows 3.1 anymore (nevermind that it isn't even available). They want the newest, most featureful software. Software really has improved over the years, as it continues to, and the fact that it continues to will mean that those who want to take advantage of ever-improving software will be needing faster and faster computers for the forseeable future.

Apple, while selling to the "PC Market", is essentially competing with no one, hence the less frequent updates to computer speed, which they don't really focus on in the first place.

Apple is competing as hard as it can with the PC market. The less frequent updates to their machines that you see are due not so much to complacence but to 1) the fact that they really can't improve their computers to the point where it makes financial sense to do so, in part because they haven't got fast enough chips from Motorola and IBM, and 2) they have a large, passionate, loyal customer base who is willing to live with this complacency.
Apple's focus is on integration (although not in the twisted sense that Microsoft does) of software and hardware, a seamless user experience, features (in particular, their "i" programs), and the "digital lifestyle". And they are good at what they do. I switched two months ago and I wouldn't go back to PCs for anything. The joy of computing just isn't there. I don't dread doing work on my Mac the way I did for my home PC and the way I still do for my work PC. PCs just don't offer that experience, no matter what the pseudo-IT-genius-uber-geek-"well-you-must-be-doing-something-wrong"-crowd say. I'm a 20-year PC user. I know what I'm doing. PC makers and PC software makers, however, do not.

This is all true, very true
And personally, I think trying to compare Macs and PCs, while no doubt fun for your average troll or philosophical type, is a waste of time. While they do similar things, they go about them in dissimilar ways.

There are different levels on which to compare Macs and PCs. If you're trying to decide between buying a Mac and a PC, why can't you compare them? They're both intended to be used for more or less the same things, by more or less the same people. Of course they can be compared.

And trying to compare speeds between the two based on something like MHz/GHz is not even apples and oranges, apples and dog crap or apples and skyscrapers.
If Macs ran PC software natively the same way PCs did and the only difference was the CPU and internal components, then I could understand people being outraged over such a huge gap in MHz/GHz. But they don't. Macs run Mac software, and they run it well.

The problem is that Macs are running Mac software increasingly slower than PCs are running PC software. The fact that a person can run Photoshop 7.0 on a fast Mac and Photoshop 7.0 on a fast PC and watch the PC piss all over the Mac is evidence enough that the Mac needs improvement. In this case, it's not apples to dog crap, it's very clearly apples to apples. It doesn't matter what special "whatever" the Mac has under the hood - it matters how fast this gaussian blur takes to render. This is why these threads about IBM's "new chip that's going to save us all" are so popular. The Mac needs a faster chip and it needs it yesterday. Sure, a lot of people may not need a faster processor now, but for better or worse, it's the most demanding 5% of computer users who will be the ones who will be deciding where the computing industry will be tomorrow, not the other 95%.
While there is a large numerical difference between PCs and Macs, the apparent speed (at least in my experience so far) is similar, if not equal, and that's all that truly matters. End of story.

That's very good for you. Unfortunately there are a lot of people who will disagree with you, and that number is growing quickly every day.
Benchmarks are meaningless. It's what you see on the screen that matters. Some Photoshop filter takes 20 seconds longer on a Mac than it does on a PC? Even if you do it dozens or hundreds of times, do you notice?

20 seconds * 100 times = 2,000 seconds = about a half hour. Just to continue this, to use a graphic designer who does this as a living for an example: 30 minutes/day * 5 days/week * 4 weeks/month = 600 minutes = 10 hours, or slightly more than an entire work day each month, lost because the Mac can't keep up with the PC. And that's not just lost time, that's lost money. Now assuming you're a graphic designer, this becomes a question of economics. Does using a Mac and all its advantages (whatever they are) make up for the initial higher cost of that Mac plus the fact that it loses one day per month in lost productivity?

This doesn't only apply to graphic designers, of course - it applies to every job whether commercial or not at which Macs are used. Windows keeps getting better, PC hardware keeps getting faster, and as a result, the incentive to buy Apple is becoming less and less clear.
 

scem0

macrumors 604
Jul 16, 2002
7,028
1
back in NYC!
With the low power consumption we can see new enclosures, and multiprocessors. Yes, multi. :D What i am saying is maybe we will have the cube back (with dual processors), dual processor laptops, or quad processor powermacs. Hey, we have no idea what apple might make of the low power consumption - but I look forward to seeing.

Note - This processor isn't really going to outperform other companies processors by much, but it will come in multi-processors, making it faster. That is my view of this processor. The only difficulty is waiting for it......
 

Chryx

macrumors regular
Jul 8, 2002
248
0
With the low power consumption we can see new enclosures, and What i am saying is maybe we will have the cube back (with dual processors),

Nope, not gonna happen, the PPC970 is relatively low power for a processor in that performance catagory, but it's NO WHERE NEAR cool enough to stick in an 8"x8"x8" cube (well, maybe ONE of them with active cooling)