New IBM chip breaks barriers to double speed

[Dan1317]

http://www.ft.com/cms/s/c41f2d94-973c-11da-82b7-0000779e2340.html

Just an interesting article I thought some of you might like to read.

 

[General]

Clock frequency means nothing to me. Benchmarks do.

 

[spud]

http://www.ft.com/cms/s/c41f2d94-973c-11da-82b7-0000779e2340.html

Just an interesting article I thought some of you might like to read.

Oh IBM blowing their own horn yet again.

 

[second_derivative]

There’s nobody looking at anything like this. We have a more highly integrated chip that is multi-core and we are increasing the frequency – we are turning up both knobs at once when the industry is going the other way and turning [the frequency] knob down

you'd think they'd get a clue.... even intel stopped with the GHz-mongering.

 

[K8MAN]

You dont inturpret that very well. They are going multicore in addition to cranking up the clocks. It's all about finding the perfect equillibrium between the 2. The 3rd factor which they dont mention is IPC and that will be the deciding factor in this technology's success.

 

[exit2dos]

Strange, if this technology works as IBM says, that Apple didn't stay with PowerPC. Apple claimed among its reasons to switch to Intel that the PowerPC architecture didn't scale as well as Intel's. I'm sure IBM showed Apple this on their roadmap, just like Intel probably let Apple know details about their Conroe development. If Apple turned IBM down then it's likely that they don't believe IBM will pull this off or it's going to be too expensive to be viable.

 

[dougie_boy]

this is a REALLY fuucking old artical people keep on dragging up. i swear ive seen it 3 times. its datted feb for fuuck sake.

let. it. go.

 

[JonathanDeane]

http://www.ft.com/cms/s/c41f2d94-973c-11da-82b7-0000779e2340.html

Just an interesting article I thought some of you might like to read.

Oh IBM blowing their own horn yet again.

You hit the nail on the head... other then Cell and some of the big iron stuff they do IBM realy isnt a player in the home chip market... If its not X86 its crap !!! lol before the people hunt me down and burn me at the stake, I mean that X86 is good enough for home use and is the most important in that respect.

 

[joset]

Strange, if this technology works as IBM says, that Apple didn't stay with PowerPC. Apple claimed among its reasons to switch to Intel that the PowerPC architecture didn't scale as well as Intel's. I'm sure IBM showed Apple this on their roadmap, just like Intel probably let Apple know details about their Conroe development. If Apple turned IBM down then it's likely that they don't believe IBM will pull this off or it's going to be too expensive to be viable.

I'd be much more careful, here:

a. Although AMD's been working in close relationship with IBM, they're entirely different matters.

b. IBM's PowerPC is/was one thing (which was used by Apple & manufactured by Motorola and, lastly, by Freescale...); the IBM POWER series are to PowerPC what Itanium is to x86...

c. IBM's a very different company, from both Intel & AMD; they hold (I think) the crown of the [computing] patent record, for the last few years; PowerPC, POWER series, Cell (which is an Hitachi patent, I believe...). KiloCore, etc, are mostly IBM technological achievements.

d. IBM is - no-more - playing within the the PowerPC vs x86 league; actually, their closest competitors in the [super] computing arena are... themselves! (they're using PowerPC, Intel & AMD processors for it... it's cheaper!); Sun's way behind...

And the list goes on...

POWER6 is a MCM; but, if IBM's to achieve between 4 to 5GHz, that will be by a single core alone... pretty amazing for a non-NetBurst uArch...


Cheers!

 

[exit2dos]

a. Although AMD's been working in close relationship with IBM, they're entirely different matters.

Not sure what you mean here - at least in the context of my original post. :?

b. IBM's PowerPC is/was one thing (which was used by Apple & manufactured by Motorola and, lastly, by Freescale...); the IBM POWER series are to PowerPC what Itanium is to x86...

I agree with your analogy. I don't know how I got PowerPC out of Power6. (I can read really fast, comprehension is another matter. )

c. IBM's a very different company, from both Intel & AMD; they hold (I think) the crown of the [computing] patent record, for the last few years; PowerPC, POWER series, Cell (which is an Hitachi patent, I believe...). KiloCore, etc, are mostly IBM technological achievements.

This has been IBM's reputaion for years. Which leads me to an off-topic question: Do you know if IBM holds any patents on today's PCs? Going clear back to the PC-clone days some have argued that IBM put its pc design as "public domain" and that's why they tried ps/2 and microchannel in order to regain rights to the platform - while others argue that IBM holds BIOS and some chipset rights and still makes money on all x86 machines. Haven't thought of this in years, but your post reminded me of it, and how I never did learn a definitive answer.

d. IBM is - no-more - playing within the the PowerPC vs x86 league; actually, their closest competitors in the [super] computing arena are... themselves! (they're using PowerPC, Intel & AMD processors for it... it's cheaper!); Sun's way behind...

And the list goes on...

POWER6 is a MCM; but, if IBM's to achieve between 4 to 5GHz, that will be by a single core alone... pretty amazing for a non-NetBurst uArch...

Reading this post and others of yours in other threads, you obviously know far more of the upper-end than I. However, if you want to be taken seriously, you really should put an Inquirer link to back up your comments. :lol:

Thanks for your thoughts. Off to bed I go!

Cheers!

Bottoms up!

 

[chewbenator]

Cool, too bad I don't own a server :?

 

[voxel]

Do you know if IBM holds any patents on today's PCs?

Doubt it. I think... Award/Phoenix created their BIOS using "clean-room" engineering techniques (no original docs or people who worked on the BIOS previously) so only a few are paying IBM dinero effectivo for their BIOS.

http://en.wikipedia.org/wiki/Clean_room_design

As for Apple ditching PowerPC - I think it was due to IBM/Motorola/FreeScale NOT being able to provide a low watt PowerPC CPU for the MacBooks.

http://www.insanely-great.com/news.php?id=6167

Anyhow all this ill-informed "IBM sucks because PowerPC/Cell suck" is retarded fanboyism. PowerPC != Power6. Cell != Power6.

http://www.top500.org/lists/2006/06

Power6 competes with the Itanic and for the most part is better supported and faster. The IBM of old - OS/2, PS/2, Microchannel - was proprietary and crudgy. The new IBM is open and plays well with others.

 

[m25]

Their chip based on Netburst2 will perform slightly worse than a K8 and require a dedicated power line :lol: :lol: :lol:

...forgot about your car's radiator.

 

[martyjs]

Looks good in thery but what wil AMD and Intel have by then, it don't count until it's out. 8)

 

[joset]

a. Although AMD's been working in close relationship with IBM, they're entirely different matters.

Not sure what you mean here - at least in the context of my original post. :?

My point was to emphasize IBM's role in both Apple & AMD (and Intel!), although in different proportions & partnership characteristics and in computing, in general; so, since Apple is not (directly) working with IBM, I only mentioned AMD when you only referred to Apple. Sorry for that and... context adjusted!

(...)

This has been IBM's reputaion for years. Which leads me to an off-topic question: Do you know if IBM holds any patents on today's PCs? Going clear back to the PC-clone days some have argued that IBM put its pc design as "public domain" and that's why they tried ps/2 and microchannel in order to regain rights to the platform - while others argue that IBM holds BIOS and some chipset rights and still makes money on all x86 machines. Haven't thought of this in years, but your post reminded me of it, and how I never did learn a definitive answer.

That's really a good question. Actually, I haven't got any data which can corroborate or refute any of your above points. Nevertheless, I'm pretty sure IBM is/has been in more 'places' than one can imagine...

Reading this post and others of yours in other threads, you obviously know far more of the upper-end than I. However, if you want to be taken seriously, you really should put an Inquirer link to back up your comments. :lol:

Thanks for your thoughts. Off to bed I go!

Cheers!

Bottoms up!

Well, thanks for reading my posts.
That's a funny remark; I'll try to follow your suggestion, whenever The Inq.'ll allow me to quote it or link to it.


<echo off>

 

[benjamin]

b. IBM's PowerPC is/was one thing (which was used by Apple & manufactured by Motorola and, lastly, by Freescale...); the IBM POWER series are to PowerPC what Itanium is to x86...

The PowerPC 970 is a stripped-down POWER4 with an Altivec compatible unit bolted on.

A powerful lineage

One of the original design goals of the Apple-IBM-Motorola partnership that developed the PowerPC architecture back in 1991 was to define a 64-bit architecture that was a superset of the 32-bit architecture, in order to provide application binary compatibility for 32-bit applications. The PowerPC architecture that was born of this partnership is -- and always was -- a 64-bit architecture derived from the IBM POWER architecture. From the very beginning, PowerPC was designed to support switching between the 64-bit mode and the 32-bit mode. As a relative of the IBM POWER4 and POWER5 processors, the PowerPC 970 family may be a new generation of PowerPC processors, but it inherits a history of over ten years of 64-bit computing at IBM.

Source

I was under the impression that Itanium was taken from the ground up. If not, of course, add this post to the rest of the useless froth on THG forums.

 

[mpjesse]

I agree with Joset.

I do believe this Power6 is the beginning of the end for IBM's PC desktop processor division.

No one will buy these. the only customer i forsee purchasing these would be the makers of superclusters like Cray or NEC. but even cray is being pressured to reduce energy costs of their systems. and the only superclusters they sell go to LANL, Sandia, Livermore, etc. with Sun competing again and Intel's new Xeon lineup, I doubt anyone will be seriously interested in these new PowerPC's.

no one will want these power hungry overclocked PowerPC's.

apple effectively ended IBM's PowerPC. and who can blame them?

 

[joset]

The PowerPC 970 is a stripped-down POWER4 with an Altivec compatible unit bolted on.

A powerful lineage

One of the original design goals of the Apple-IBM-Motorola partnership that developed the PowerPC architecture back in 1991 was to define a 64-bit architecture that was a superset of the 32-bit architecture, in order to provide application binary compatibility for 32-bit applications. The PowerPC architecture that was born of this partnership is -- and always was -- a 64-bit architecture derived from the IBM POWER architecture. From the very beginning, PowerPC was designed to support switching between the 64-bit mode and the 32-bit mode. As a relative of the IBM POWER4 and POWER5 processors, the PowerPC 970 family may be a new generation of PowerPC processors, but it inherits a history of over ten years of 64-bit computing at IBM.

Source 1

Although the design, process & most of the microarchitectural philosophy was the same, that's pretty much to compare an Intel P4 DT with an Intel Xeon Server: both are derivatives of a single uArch, NetBurst.
IBM went a [big] step further: The POWER series come in a MCM (2 to 4 chips with off-die L3 cache, in its current implementation).
For a glimpse & compare on the POWER4:

Source 2

Meanwhile, POWER6 has much less to do with the PowerPC (and, as a matter of fact, with all the previous POWER chips), if you care to search.

I was under the impression that Itanium was taken from the ground up. If not, of course, add this post to the rest of the useless froth on THG forums.

Yes, Itanium was made from the ground up. That, doesn't invalidate the comparison (a bit exaggerated, perhaps) between Itanium & the x86; up until very recently, Itanium still emulated x86, for retro-compatibility's sake (not really the same as adding an Altivec/VMX/..., but you get the point).

Anyway, the PowerPC still floats in the embedded space; a POWER5/6 would give IBM a hard time to shove it in there...

My point still is to try to differentiate IBM from the current "AMD vs Intel" issue, as chip/platform manufacturers & putting it forth as, most probably, the biggest technology giant, if you will.


Cheers!

 

[benjamin]

I agree with Joset.

I do believe this Power6 is the beginning of the end for IBM's PC desktop processor division.

POWER is a server chip.

No one will buy these. the only customer i forsee purchasing these would be the makers of superclusters like Cray or NEC.

Or Big Blue themselves?

I doubt anyone will be seriously interested in these new PowerPC's.

That's fairly naïve.

no one will want these power hungry overclocked PowerPC's.

POWER -> PowerPC. The parent architecture is, basically, a superset.

apple effectively ended IBM's PowerPC. and who can blame them?

Mhm. The desktop market is the only market of importance, etc, etc.

As I'm nearly done pissing in the wind, here's a link. Or, see above.

 

[benjamin]

Although the design, process & most of the microarchitectural philosophy was the same, that's pretty much to compare an Intel P4 DT with an Intel Xeon Server: both are derivatives of a single uArch, NetBurst.

IBM went a [big] step further: The POWER series come in a MCM (2 to 4 chips with off-die L3 cache, in its current implementation).
For a glimpse & compare on the POWER4:

Source 2

Indeed, PPC970 was, IIRC, built upon the single core ripped from POWER4, in much the same way AMD are now boasting of their own 'modular' microarchitecture.

Meanwhile, POWER6 has much less to do with the PowerPC (and, as a matter of fact, with all the previous POWER chips), if you care to search.

It's a generation increment... this is to be expected? As far as I remember, there has been no real PowerPC spinoff from POWER5... but you're right - it's been a while since I did care

My point still is to try to differentiate IBM from the current "AMD vs Intel" issue, as chip/platform manufacturers & putting it forth as, most probably, the biggest technology giant, if you will.

I understand and acknowledge your point... who is aiming for which market, ne'er the twain shall meet - my original post was just a nitpick...

Whether IBM can follow up on their promises for POWER6 remain to be seen, but there's no way you can say whatever they come up with has no viable use, as long as it's binary compatible with existing software. Likewise, PowerPC, whether it's a 970MP-based blade or a 74xx based Cisco router, has it's place. You don't see ARM chips inside your gaming boxes, but do they exist? Shit, yea, they do. And ARM make rather a good living.

 

[joset]

Meanwhile, POWER6 has much less to do with the PowerPC (and, as a matter of fact, with all the previous POWER chips), if you care to search.

It's a generation increment... this is to be expected? As far as I remember, there has been no real PowerPC spinoff from POWER5... but you're right - it's been a while since I did care

Well, you make it sound something was already left behind...
Maybe you'll care again, if/when IBM comes up with a new "engine" (it'll get old & go to a museum, sooner or later!)
Anyway, I'm more interested in technologies (although not only computing wise...) than in corporations/brands; even the prospect of a [feasible] "new" technique/technology grabs my interest. I really enjoy the mere 'thought' of thinking about it. IBM POWER series included.

(...)

Whether IBM can follow up on their promises for POWER6 remain to be seen, but there's no way you can say whatever they come up with has no viable use, as long as it's binary compatible with existing software. Likewise, PowerPC, whether it's a 970MP-based blade or a 74xx based Cisco router, has it's place. You don't see ARM chips inside your gaming boxes, but do they exist? ****, yea, they do. And ARM make rather a good living.

Kent, UK, uh?!
ARM is almost everywhere, from computing to comms. You can name a bunch of IP & chip manufacturers which will hardly be mentioned in this forum; and. more often than not, they're behind the tasks most of us attribute to Intel/AMD... or IBM.


Cheers!

 

[joset]

I agree with Joset.

I do believe this Power6 is the beginning of the end for IBM's PC desktop processor division.

No one will buy these. the only customer i forsee purchasing these would be the makers of superclusters like Cray or NEC. but even cray is being pressured to reduce energy costs of their systems. and the only superclusters they sell go to LANL, Sandia, Livermore, etc. with Sun competing again and Intel's new Xeon lineup, I doubt anyone will be seriously interested in these new PowerPC's.

no one will want these power hungry overclocked PowerPC's.

apple effectively ended IBM's PowerPC. and who can blame them?

Sorry for the delay (I keep repeating myself : My MB went out of business; I've to use a friend's iMac to keep up-to-date, whenever I can...).

Here's an interesting note on the POWER6 microarchitecture:

Dr Frank Soltis, an IBM chief scientist, said Intel had delivered 4 GHz Pentium 4 chips to several manufacturers but subsequently needed to cancel plans because of problems with current leakage. Soltis said IBM had solved these problems by using a combination of 90nm and 65nm parts in the Power design.

http://en.wikipedia.org/wiki/POWER6

This idea already crossed my mind - the mix up of two process nodes - mainly on what concerned AMD's [tricky] transition to the 65nm node.

Anyway, IBM deliberately left the desktop space not long ago, either giving up Intel's and their own PowerPC processor lines. For the time being, its POWER series will be integrated into ultrahigh-end supercomputing platforms.
But, I go the cautious way: There's Cell, KiloCore, AMD (of course!) and all the potential of IBM's R&D; one way or the other, IBM 'will be there'.

It's my personal conviction (no data to back it up) that, it was really IBM to quit Apple and not the other way around. Aside from the processor's IP rights, the PowerPC 970 was being fabbed by Motorola and, lately, by Freescale. Apple was 'small peanuts' & the embedded space was growing (automotive; communications; military; you-name-it). Certainly, IBM had other plans for their immense technological & profitable aims.
Again, my opinion: They already had (have) a foot in the DT market (AMD); server-wise, they had thrown almost everything into it but Sun; and, they spun off their mobile flag into the 'Made-in-China's' Lenovo.
Other than a year-over-year patent # recordist, technology/IP prime licensor, parts of its sheer dimension, they might come up with more than just 'daring projects', DT-wise.
So, not being [directly] involved in A vs B conspiracies, IBM's everywhere, if you care to look closely. My view, anyway.


Cheers!