CONFIRMED IBM Power PC 970 - Page 9

[quote]Originally posted by COS:
<strong>
That of course leads us to the question about Apple's high-end machines... It wouldn't seem unreasonable that Apple would decide to make its high-end machines (ultra high-end machines) equipped with a genuine Power4 chip.

This would make Apple's consumer models running in parallel speed (maybe slightly higher) with that of x86's highest-end systems while putting Apple's high-end leaps and bounds ahead.

It seems like a natural fit if you ask me.</strong><hr></blockquote>

Or maybe there's something like a 990 that we don't know about yet.... after all, those in the know have made it fairly clear that the 970 is not an only child.

[quote]Originally posted by Junkyard Dawg:
<strong>

Well, consider this then,

1. The PPC970 has not one, but TWO Altivec units. Altivec performance will be off the charts. Furthermore, the PPC970's altivec performance won't be choked by a slow bus....according to those who know about this sort of thing, the current G4 altivec quickly chokes because the MPX bus cannot feed it fast enough. So not only will a major bottleneck be removed, but an additional SIMD unit will be added. For altivec-optimized applications, the 970 will shame Pentium 4s, no doubt about it.

2. The typical trade-off for MHz is to lower the MIPS/MHz, partially by adding extra pipelines. Yet, the 970 clocks significantly faster than the G4, and STILL turns out more MIPS. Also, the 970 is going to debut at 1.8 GHz, probably on a 130 nm process. This leaves lots of room for scaling...IBM could migrate the 970 to 90 nm fairly quickly, and probably scale to 3 GHz within a year or so of the 970's introduction. In contrast, the G4 debuted at 400 MHz and is near the end of its life.
</strong><hr></blockquote>

1. i thought the VMX unit on the 970 is divided into two seperate working units which work on the same instructions but split them up before processing (data handling, instruction processing or s.th.) however - i'm no expert on this but i think this is the way the G4/7400 managed altivec-processing too.... so these TWO altivec units are nothing more than ONE altivec unit...

2. the G4 will be available at speeds up to 1.8Ghz on a .13µ process in the first half of 2003...

Oh, my mistake. I thought I read somewhere about the PPC 970 having two independent altivec units.

Anyways, the G4 might scale to 1.8 GHz on a 0.13 process, but the PPC 970 is going to debut at 1.8 GHz. It's got plenty of headroom, while the G4 is going to be nearly maxed out at 1.8 GHz (if it really does scale that high. I think it's more likely that Moto will choke on it).

900MHz internal bus <img src="graemlins/surprised.gif" border="0" alt="[Surprised]" />

[quote]Originally posted by DaveGee:
<strong>TBoxman, I'm glad someone took the time to follow my very simple clue. <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" /> </strong><hr></blockquote>


Eh, BITE ME!

<img src="graemlins/smokin.gif" border="0" alt="[Chilling]" />

The 970+ made in 90 nm process will scale up to 2.5 GHz. It is scheduled to be released in 2004. I can't say if that's early or late in 2004, but IBM have hinted that they'll move to 90 nm rather quickly. I guess that 970 was designed to be made in 90 nm al along. After 2.5 GHz the future is uncertain, there is a proposed continuation of the 970-core, but what it will look like I can't say.

As rumors have stated for a long time, IBM is working on a new core in their 7xx-line of processors; a successor to the 750-series. This processor is codenamed Mojave and will debut in 2004 @ 1.4 GHz, and will have a variant designed with multiprocessor in mind. Perhaps multicore, perhaps SMP.. i don't know. I also cannot say if this is the one with an integrated SIMD or if it's in Polaris. But.. from what I can tell.. Polaris is a continuation of 750-series and Mojave is a rather new core so my bet is on Mojave, not Polaris. Mojave has connections to the next generation 4xx-cores @ 2.4 DMIPS/MHz.

As both 440GP and 970 are Book E-compliant my guess it that Mojave also will be. 440GP will be replaced by 440GX (130 nm) next year and Aurora (90 nm) after that.

[ 10-20-2002: Message edited by: Henriok ]</p>

[quote]Originally posted by Merlion:
<strong>


Eh, BITE ME!

<img src="graemlins/smokin.gif" border="0" alt="[Chilling]" /> </strong><hr></blockquote>

Looks to me like I already did! <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" />

[quote]Originally posted by Henriok:
<strong>

......As rumors have stated for a long time, IBM is working on a new core in their 7xx-line of processors; a successor to the 750-series. This processor is codenamed Mojave and will debut in 2004 @ 1.4 GHz, and will have a variant designed with multiprocessor in mind. ....
[ 10-20-2002: Message edited by: Henriok ]</strong><hr></blockquote>

1.4 GHz G3 in 2004??

[quote]Originally posted by Henriok:
<strong> The 970+ made in 90 nm process will scale up to 2.5 GHz. It is scheduled to be released in 2004. I can't say if that's early or late in 2004, but IBM have hinted that they'll move to 90 nm rather quickly. I guess that 970 was designed to be made in 90 nm al along. After 2.5 GHz the future is uncertain, there is a proposed continuation of the 970-core, but what it will look like I can't say.</strong><hr></blockquote>

I would be exceptionally surprised to discover there was any truth in the PPC970 topping out at 2.5 GHz on a 90 nm process. 1.8 Ghz on a 0.13µm process has been described as "conservative".

I wouldn't be surprised to learn there is more being worked upon or that IBM plans a rapid switch to a 0.09µm process. Once you get down to that level a lot of options open up.

The 970 is not Book E.

[quote]Originally posted by wmf:
<strong>The 970 is not Book E.</strong><hr></blockquote>

The 970 is BookE. Or maybe IBM lied in the conference.

BookE refers to the method of having a 64-bit PowerPC switching between 64 and 32-bit processing modes.

Barto

[quote]Originally posted by Barto:
<strong>

The 970 is BookE. Or maybe IBM lied in the conference.

BookE refers to the method of having a 64-bit PowerPC switching between 64 and 32-bit processing modes.

Barto</strong><hr></blockquote>

BookE refers to an awful lot more than just that. IBM never said the 970 was bookE, that was an unwarranted inference originally drawn by David Wang.

Making the 970 bookE would, as you say, have entirely changed the way in which 32/64 bit switches are accomplished, and hence made the processor unuseable in IBMs AIX systems.

PPC970 is *not* bookE.

michael

[quote]Originally posted by Junkyard Dawg:
<strong>Anyways, the G4 might scale to 1.8 GHz on a 0.13 process, but the PPC 970 is going to debut at 1.8 GHz. It's got plenty of headroom, while the G4 is going to be nearly maxed out at 1.8 GHz (if it really does scale that high. I think it's more likely that Moto will choke on it).</strong><hr></blockquote>

I certainly have to agree there. I have no reason to believe Mot could get the G4 much past 1.4-1.5, in fact. They've had all this time with Apollo (which they had said would scale easily), and all they've given us is 1.2. If they ever manage to get past 1.5, then I'll truly be shocked. In any case, if IBM gets a conservative 1.8GHz out of the great pipeline depth of the 970, then it's folly to believe the G4 could ever get there with its seven stages. Of course, I'd readily assent to the resident engineers if their opinions contradict mine, but I don't think they would.

The 970 might top out at 2.5 GHz, but that doesn't mean that there won't be a follow-up series... for example, if the 970 uses POWER4 techology perhaps the follow-up will use POWER5 technology. No matter how you slice it, IBM isn't getting out of the processor business anytime soon. In fact I'd say that they're just getting warmed up.

I also don't think that IBM's G3 has reached the end of the line. While 1.4 GHz might not be impressive for a desktop processor these days, the G3 is primarily used for embedded devices and 1.4 GHz is impressive there. It'll probably be so power efficient that it'll generate power!

The quote from the IBM filing was interesting -- I wonder who they consider "their competition" is? Just Motorola? Intel and AMD are already shipping on 0.13, I believe, and Intel at least should reach 0.09 fairly soon.

There are a lot of people pessimistic about the time line of the 970. I'm more optimistic -- I think IBM hitting full production in time for 2H '03 means we'll see a September introduction (and ship) of an Apple PowerMac using the 970. The portable & consumer versions may have to wait until the 0.09 micron version... maybe. A 1.2 GHz GPUL looks like it'll be cool enough for the iMac and possibly the Ti Book.

I would like to take the time to say I TOLD YOU SO! To many people on this forum. In ExtremeTech's AGP 8X review, there is a sentance.

"AGP 8X brings with it two noteworthy features: a 2X speed increase to 2.2GB/sec and the ability to have more than one AGP graphics device in a system."

Ha! Everyone who told me that you could only have one AGP graphics card in a computer, you know who you are.

-----------

I can't find it right now, but there is a web site with 2 pictures of the MPF conference. One shows a photograph (on a screen) comparing of the Power4 and PowerPC 970. The other shows the techical details of the PowerPC 970.

Anyway, on the tech slide, it says built on Power4 and BookE technology.

Barto

Programmer,
For me the most interesting statement was, "We have orders that would fully load the facility well into midyear of 2003." IBM is talking about their backlog for this plant. In order to fulfill their orders by midyear then it will need to be at full scale production very soon.

I'm more optimistic on the delivery of 970 systems than others. I'm betting that the reports of 2H '03 regard 970 general availability and are similar to Mot's web site not listing a 1.25 gHz 7455 or even that it exists. Apple has probabaly entered into a purchase agreement with Mot for ALL 1.25 chips hence the reason you can't order them. The same can be said for Apple's agreement with IBM.

So, if Apple has dibs on most- if not all- 970's then it stands to reason that IBM is unable to announce availability of the chip to other OEMs until Apple's appetite is satiated. I'm also betting that Apple's demand will lesson just as IBM says it will- by midyear '03. Which is about the time that all the orders for the new Macs will have been satisfied. This means that Apple will announce a 970 machine in January with low-end shipping systems by March and the high-end by April.

AGP 8X is a fairly pointless spec, becaus the 2 features it introduces (2-slot capability and faster speed) will only start to be used at the time PCI Express arrives.

And PCI Express is a much more capable spec anyway.

Barto

I'm sorry.. 970 is not Book E according so my sources. That was a statement made from those attending MPF, but nothing in 970's documentation mentions Book E.. it does mention that it's fully PPC AS version 2.0 compliant for 64- and 32-bit application binary compatibility with PowerPC Books I and II though.

And.. I do hope that 970+ scale past 2.5 GHz, but that's something I don't know anything about. My IBM source say 2.5 GHz but since they haven't even begun producing 970 in 130 nm yet.. if 1.8 GHz is conservative then 2.5 Ghz might be as well. I guess a 40% increase in performance when moving from one fab to a smaller one is reasonable, so if 1.8 GHz for the 970 really is 2.2 GHz.. then we have 3.1 for 970+.

I'm not good at predicting the prospects of performance increase in procesosrs.. perhaps someone else can shed some light in this area. I'm just telling you what I'm told, and that is that IBM is planning to bring 970+ to 2.5 GHz in 2004.

I hope the PPC970 is like the P4: starts out slower, but quickly overtakes the competition.

Err. Maybe it won't start out slower. SPEC scores aren't 100% accurate and Altivec may tip the scales.

Either way, this is a HUGE improvement over the G4. Apple is lucky IBM saved their butts. I don't know how much longer Mac users can deal with "speed envy."

Hopefully we won't have to wait much longer. My iBook is fast enough for general use, but the G4 I use for design is long in tooth. I could put a dual processor PPC970 machine to good use.

[quote]Originally posted by Henriok:
<strong>IBM is planning to bring 970+ to 2.5 GHz in 2004.</strong><hr></blockquote>

I was confused as to whether you meant it was just being brought to that speed or limited at that speed. I can't see that being its frequency cap but certainly that might be a number they aim for in scaling. They do seem to have a penchant for multiples of 0.2 GHz though. Not sure if there is a technical reason for that one or it is just coincidence.

[quote]Originally posted by Telomar:
<strong>

I was confused as to whether you meant it was just being brought to that speed or limited at that speed. I can't see that being its frequency cap but certainly that might be a number they aim for in scaling. They do seem to have a penchant for multiples of 0.2 GHz though. Not sure if there is a technical reason for that one or it is just coincidence.</strong><hr></blockquote>

I too, think it's just an introductory speed. IBM will release faster chips as they refine their manufacturing process like any chip maker.

"Normal" (effective) bus frequencies: 533MHz, 666MHz, 800MHz.

The PowerPC 970 bus (ApplePI?) has 1 and 1/8th the Hz, the extra 1/8th being address data.

So, PowerPC 970 (effective) bus frequencies: 600MHz, 750MHz, 900MHz.

The PowerPC 970 appears to be fixed at 4x the (real) clock speed. DDR Bus is 1/2 the CPU clock.

1.2GHz, 1.5GHz, 1.8GHz.

Future clock speeds will be 2^n of the original 3.

2.4GHz, 3.0GHz, 3.6GHz
4.8GHz, 6.0GHz, 7.2GHz
9.6GHz, 12.0GHz, 14.4GHz

Barto

[ 10-21-2002: Message edited by: Barto ]</p>

[quote]Originally posted by Barto:
<strong>"Normal" (effective) bus frequencies: 533MHz, 666MHz, 800MHz.

The PowerPC 970 bus (ApplePI?) has 1 and 1/8th the Hz, the extra 1/8th being address data.

So, PowerPC 970 (effective) bus frequencies: 600MHz, 750MHz, 900MHz.

The PowerPC 970 appears to be fixed at 4x the (real) clock speed. DDR Bus is 1/2 the CPU clock.

1.2GHz, 1.5GHz, 1.8GHz.</strong><hr></blockquote>

Some interesting reasoning except it is known there is a 1.4 GHz part, which wouldn't fit in. I'd be inclined to guess and just say the speeds jump by 50 MHz each time for the bus or 200 MHz for the processor.

The data of the proposed processors I've seen states 1.4-1.8 GHz for 970 and 1.8-2.5 GHz for 970+. I suspect these are the specs of the initial batch of processors. I have no idea how 970 will scale in its 130 nm fab nor how 970+ will do.

The original G4 (7400) scaled miserably, but other PPC offerings have done quite well I guess. If 970 scales the same way the 970 will go quite a bit past 2 GHz. These are completely new processors on a new fab in a new foundary, so there are _a_lot_ of room for error in every prediction we make. IBM predicts 1.8 GHz initially, and I have a feeling that IBMs predictions are on the safe end of things.. as are the SPEC-marks.

[ 10-21-2002: Message edited by: Henriok ]</p>

I saw this comment on Real World Tech:

"Unless Apple can also obtain a low cost support chip from IBM, the PowerPC 970 processor would likely force the Apple Macintosh product lines to become even more upscale, and Apple would likely retain the use of the PowerPC G4 processors for the lower end iMac and eMac product lines."

Thats not a good thing right off the bat. I would rather they used an IBM support chip and have a cheaper system all the while developing they're own support chip to deliver at the opportune time.

[quote]Originally posted by Outsider:
<strong>I saw this comment on Real World Tech:

"Unless Apple can also obtain a low cost support chip from IBM, the PowerPC 970 processor would likely force the Apple Macintosh product lines to become even more upscale, and Apple would likely retain the use of the PowerPC G4 processors for the lower end iMac and eMac product lines."

Thats not a good thing right off the bat. I would rather they used an IBM support chip and have a cheaper system all the while developing they're own support chip to deliver at the opportune time.</strong><hr></blockquote>

Okay this entire post is based on either insider info and or forum speculation (or both).

It's been said that the 970 was NOT designed for Apple but instead that Apple signed on later in the process. IBM had/has *other* plans for the 970 already. It's also been said that the 970 didn't begin it's life with VMX (altivec etc) but instead was added to the design at the request of Apple.

I dunno how much of the following is true but if we take it at face value that it is... Then I have to ask the following... Would IBM add AltiVec to the 970 just so Apple could sell a small number of boxes with it?

Remember iMacs and iBooks make up the lions share of Apple sales (in # of raw boxes sold) and even with the latest push to putting TWO PPCs in the PRO desktops the iMacs and iBooks still account for a much larger number (counting CPUs).

What I'm saying is I have my doubts that the 970 will only be used only an 'even more upscale box'.

Taking this comment alone....

"Unless Apple can also obtain a low cost support chip from IBM..."

If I'm not mistaken IBM has more than a fair bit of experience designing and producing custom ASICs and who better to do the support chip in the first place then the folks or built the cpu? Finally if IBM also has plans for the 970 then wouldn't that mean that they too would need a support chip? All I'm saying is I think most of this has been addressed quite some time ago. It's not like IBM just went to Apple last month and said 'Hey check this out pretty neat eh?'.

Remember as we are all talking about the 970 Apple and IBM or MOT or insert your favorite x86 cpu maker is on their way to planning the next next next gen CPUs.

Dave

[ 10-21-2002: Message edited by: DaveGee ]</p>

[quote]Originally posted by DaveGee:
<strong>
Finally if IBM also has plans for the 970 then wouldn't that mean that they too would need a support chip?
Dave</strong><hr></blockquote>

Makes perfect sense to me. IBM did say this chip is for the desktop market. They didn't say the $10,000 workstation market.

Several has entertained the notion that as the 970 is a high performance CPU it has to be a high cost CPU as well.

That is not how it works. The producer is out there to make as much money as possible not to do that on a single chip. If IBM can produce a lot of the 970 at a resonabley low cost compared to AMD and Motorola CPUs, they might make more money selling a lot of them at a low price. The fact that the 604E was used in servers like the RS600 does not mean that the CPU as such was expensive, in fact it was used in budget computers such as the 7300/200.

Regarding the 90micro version of the G4 by Motorola. The last time Motorola manufactured a leading edge CPU it was the 40 MHz 68030 back in 1990. This was in the day of OS 6 and the multifinder way before OS 7 (remember that the high end 604E in the 8600/9600 and the fast G3s in the B&W was made by IBM).

My guess is that 2003 is the last year there will be any Motorola CPUs in the Macs. :cool:

[quote]Originally posted by Outsider:
<strong>"Unless Apple can also obtain a low cost support chip from IBM, the PowerPC 970 processor would likely force the Apple Macintosh product lines to become even more upscale, and Apple would likely retain the use of the PowerPC G4 processors for the lower end iMac and eMac product lines."
</strong>
Thats not a good thing right off the bat. I would rather they used an IBM support chip and have a cheaper system all the while developing they're own support chip to deliver at the opportune time.<hr></blockquote>

I don't agree with the original quote's reasoning. What is it about the support chip that will force the already upscale PowerMac further upwards? Apple has significantly increased the complexity of its support chips in the past and yet the PowerMacs (and high end Macs before that) have always hovered around the same price points. PC makers use chipsets of similar complexity for the PentiumIV at lower price points. This particular chipset may require shouldering an R&D burden (or sharing it with IBM, at least), or may eat into Apple's margins for a while... but that's why Apple has substantial margins. It used to be that a 64-bit wide 167 MHz bus was "out there", but guess what's under my desk now? This bus is synchronous paired unidirectional 32-bits wide channels @ 450 MHz w/ DDR, which is an interesting mix of simple and complex elements. I'm pretty sure that it was designed specifically to have low-cost fast implementations, just like HyperTransport is. I'm also pretty sure that Apple has had at least a year to design a support chip, and that by the time 2H '03 rolls around they will have it ready for production. The main question, to me, is whether it'll appear across the line immediately or just start in the PowerMac line (at the current price points). I'm guessing it'll lead in the PowerMacs for 6-9 months before leaking into the high end of the consumer machines, unless Motorola really does roll over and play dead.

[quote]Originally posted by DrBoar:
<strong>Several has entertained the notion that as the 970 is a high performance CPU it has to be a high cost CPU as well. </strong><hr></blockquote>

This reminds me of another point. IBM could have designed a workstation chip, but they didn't. They didn't build a chip with ~1000 SPECmarks because that's the fastest thing they could build, they built this particular chip because they decided it was a good balance between cost, power consumption, and performance. What would the point of building such a balanced chip be if its cost was bloated by the system around the chip? IBM's design analysis extended beyond the chip's boundaries to include the system required to support it, especially since they wanted Apple to use it as well. If it was only going to go into expensive workstations they'd have thrown 100+ million transistors at it and really gone for the jugular.

[quote]Originally posted by Programmer:
<strong>

I don't agree with the original quote's reasoning. What is it about the support chip that will force the already upscale PowerMac further upwards? Apple has significantly increased the complexity of its support chips in the past and yet the PowerMacs (and high end Macs before that) have always hovered around the same price points. PC makers use chipsets of similar complexity for the PentiumIV at lower price points. This particular chipset may require shouldering an R&D burden (or sharing it with IBM, at least), or may eat into Apple's margins for a while... but that's why Apple has substantial margins. It used to be that a 64-bit wide 167 MHz bus was "out there", but guess what's under my desk now? This bus is synchronous paired unidirectional 32-bits wide channels @ 450 MHz w/ DDR, which is an interesting mix of simple and complex elements. I'm pretty sure that it was designed specifically to have low-cost fast implementations, just like HyperTransport is. I'm also pretty sure that Apple has had at least a year to design a support chip, and that by the time 2H '03 rolls around they will have it ready for production. The main question, to me, is whether it'll appear across the line immediately or just start in the PowerMac line (at the current price points). I'm guessing it'll lead in the PowerMacs for 6-9 months before leaking into the high end of the consumer machines, unless Motorola really does roll over and play dead.</strong><hr></blockquote>

I agree that Apple has had enough time to at least have significant time to have a preliminary support chip but it would have had to be with much help with IBM. In fact i would not be the least surprised if this chip would not only be partially designed by IBM, but also fabricated and sold under the IBM name. IBM would also have exclusive rights to ship it with their own systems.

IBM may become a chipset maker and supplier for Apple. That's depends what wants Apple.

Concerning the prize of the Mobo, i don't think that the new mobo will cost a fortune. When i see the specs of the latest products of the X86 world and see their prize , i came to the conclusion that the performance prize ratio is better and better.

For example an asus A7V7X can support DDR memory (from pc 2100 to PC 3200) serial ATA, ethernet, USB 2, firewire, internal audio 6 channel, ata 133 and raid and AGP 8 X for 185 euros (perhaps a little less in $).

I doubt that an Apple mobo will cost more than 300 $.

[quote]Originally posted by Outsider:
<strong>
I agree that Apple has had enough time to at least have significant time to have a preliminary support chip but it would have had to be with much help with IBM. In fact i would not be the least surprised if this chip would not only be partially designed by IBM, but also fabricated and sold under the IBM name. IBM would also have exclusive rights to ship it with their own systems.</strong><hr></blockquote>

Intel supplies chipsets to go with their processors, no reason that IBM couldn't do the same. More likely, however, is that IBM & Apple would collaborate -- IBM could provide the bus interface "modules" which Apple would integrate onto their core chipset in place of the MPX interface. Then again, don't under estimate Apple's chipset design abilities, as far as we know they have a pretty solid team onsite.

"I dunno how much of the following is true but if we take it at face value that it is... Then I have to ask the following... Would IBM add AltiVec to the 970 just so Apple could sell a small number of boxes with it?"

If Motorola really are letting IBM to the Apple market, it would make sense that the 970...could...in time, be pervasive through Apple's whole line. Especially with the subsequent .9 die shrink. A million of anthing a quarter is no small amount of money.

Apple and Moto relations seem strained to me. Knowing Steve Jobs from what I've managed to garner...I don't think he could tolerate their level of 'competance'. Moto' may grace the consumer lines 2nd half this year/next year before the door hits them on the way out. But a .13 or .9 G4 would keep the consumer line happy for the next year-ish.

Seems to make sense that an IBM or Apple chipset or joint mobo venture would accompany the co-operation on the 970 front.

If Apple offered the business of Powermacs, Powerbooks and potentially iMac and iBook sales in due course in front of their nose as a carrot...then I'm sure IBM could include Altivec for a major customer. Everything I've read about the way IBM designs chips ala Power4 suggests they'd come round to the idea of VMX eventually. It's a tech' they could use in the embedded markets as Moto' does. So, if VMX doesn't fit for IBM re: 970, it's benefits may extend to other chips for IBM.

The way I see it. Most of the R&D is done via Power 4 and future Power5 chips. It's merely repackaging for a company as big as IBM. Hardly hard work to include VMX and Mobo designs for not only Apple but for your own markets as well.

Yeesh. X86 mobo makers spit out Motherboards left, right and centre. Why is there this question mark over Apple or even IBM? Two substantial companies in their own right.

If we view the IBM/APPLE collaboration as a real long term project, I see no reason why the partnership couldn't go from strength to strength. IBM gives Apple a definite cpu road map and in turn, Apple can offset the R&D. Back scratching.

Unix, PPC, Desktop, Gaming, Multimedia, the Internet, low end server. They have much in common.

And there's no way I see 970 being used in uberworkstations. It's pretty obvious it's a desktop chip...that will be competitive with whatever is in the x86 desktop market this time next year. Especially if you extrapolate the spec scores.

At which, the x86 might 'eek' ahead. But then, this 'platform' neutral bench indicates a G4 is ten times slower than intel's current? If anything, if the G4 is in the 'ball park' with current performance than I think the 970 will have the edge in 'real world' performance. But we'll have to wait and see for that. But hey, a PPC chip that has 'level playing field' spec scores? That's a feat in itself! And any chip that will have four times the performance of a G4 at its debut speed sounds okay to me. Got cash? I'm saving mine.


Lemon Bon Bon

[ 10-21-2002: Message edited by: Lemon Bon Bon ]

PS. A headline/link/source on Macrumors:

"RealWorldTech posts some detailed notes from IBM's presentation last week. Here are some hilights:

- the SIMD unit is AltiVec compatible
- PowerPC 970 has already "taped out", parts exist in labs, undergoing performance eval and debugging
- Second half of 2003 is when volume production is expected
- if Apple adopts the 970, 32-bit applications can run seamlessly, after OS modifications are made
- due to the subsystem support required for the chip, "Unless Apple can also obtain a low cost support chip from IBM, the PowerPC 970 processor would likely force the Apple Macintosh product lines to become even more upscale".

[ 10-21-2002: Message edited by: Lemon Bon Bon ]</p>

My calculations:

800MHz DDR bus (400MHz x 2)
-----------------------
400 x 3 = 1200MHz
400 x 3.5 = 1400MHz
400 x 4 = 1600MHz


900MHz DDR bus (450MHz x 2)
-----------------------
450 x 4 = 1800MHz
450 x 4.5 = 2025MHz
450 x 5 = 2250MHz
450 x 5.5 = 2475MHz

The ones in bold are the only ones we know for sure. They were the only speeds directly quoted by IBM. The other speeds are extrapolated.

IBM quoted 1.4 GHz chips as existing. When technical documentation appears my guess is the bus will look like the GX bus except with a 4:1 clockspeed to bus speed ratio instead of 3:1. Oh and the PPC 970 will be double pumped.

[quote]Originally posted by Krassy:
<strong>
1. i thought the VMX unit on the 970 is divided into two seperate working units which work on the same instructions but split them up before processing (data handling, instruction processing or s.th.) however - i'm no expert on this but i think this is the way the G4/7400 managed altivec-processing too.... so these TWO altivec units are nothing more than ONE altivec unit...
</strong><hr></blockquote>

The two AltiVec units are seperate execution units in the same way as the G4 has 3 integer units, a branch unit, a condition unit, a floating point unit, and a couple of vector units. Each of these handles some of the instructions. I don't know if the 970 divides its vector instructions into math & permute, or some other way... but there are two units which appear as the one "logical" view of the VMX unit.

There's a picture of it on this page <a href="http://www.extremetech.com/article2/0,3973,636954,00.asp" target="_blank">Extremetech</a>

[ 10-21-2002: Message edited by: Bigc ]</p>

[quote]Originally posted by Kecksy:
<strong>I hope the PPC970 is like the P4: starts out slower, but quickly overtakes the competition.</strong><hr></blockquote>

I like to think of it as the PowerPC 601 - it started out a little slower, but then went on to completely kick Intel's ass with the 604, 604e, and G3.