G5 Trinity @ WWDC - Page 5

Thanks for setting the record straight. Now if you could only shed some light onto what we may realistically see at WWDC that would be great.

Sorry, it was late and I just hate it when everything dissolves into a PC vs. Mac price war shouting match.

As far as the "helpful and informative post" goes how's this?


Wrong and wrong. PCIe graphics (on the Mac anyway) are not about "being one louder" or incremental game performance. PCIe will be of HUGE benefit to every OS X user because Apple is in the unique position of integrating the technology directly into both apps and the operating system they run on.

Again, PCIe isn't just 16x, it's 16x in both directions.

AGP = 8x from CPU to GPU, 1x from GPU back to CPU
PCIe = 16x from CPU to GPU, 16x from GPU back to CPU.

That's right, a PCIe native GPU can send rendered data from its memory back to the CPU 16 times faster than AGP! Think about what that means for Quartz Extreme, Open GL and GPU bound applications like Shake and Motion. This is precisely where Apple's 100% control over hardware and software and the subsequent nimbleness to bring solutions to market quickly is so important. I fully expect PCIe to be a new foundation for all types of radical improvements to the Mac experience, and we'll get our first real taste of it at WWDC with PCIe equipped Power Macs and a massively improved graphics engine in Tiger.

Kasper

Are you saying croquer.free.fr are the "sick, lying morons", or somebody else? What about tman's source? He's certainly seems credible. Or is he a sick, lying moron too?

AI hasn't had a meaningful article on the Power Mac roadmap in months. Why don't you tell us what you know instead of just insulting people.

http://www.techtv.com/screensavers/s...649748,00.html

AGP is 266MBps x 8 = 2.1GBps

PCIe 16x equals 200-250MBps x 8 = 2GBps each way for 4 GBps theoretical.

The bidirectionality doesn't really mean anything because that's not the bottleneck in GPUs.


The AGP bus does not allow bus snooping. Since it's sending a lot of data through the system the spec was created to minimize traffic on the FSB. AGP does allow programmers to send PCI Read Requests if they need to snoop the bus but there's no coherency between data coming back on the AGP versus PCI bus.

Enter PCI Espress. Developers can now assign values to their Read Request to either snoop the bus or not snoop the bus. The GPU has internalized the GART chip(removing it from the North Bridge) there generally isn't much data that would need to reside in the CPU L1 or L2 cache(due to size) but if so there is you can flag certain pieces. All handled within the driver.

I think the biggest feature of PCI express isn't the bandwidth but it's intelligence. Bus contention is eased now that read completions that are not dependent on a posted write can be moved ahead. With AGP you can have traffic bottlenecked behind slow posted writes slowing the whole bus down. That's now gone.

I want PCI Express as quickly as possible but I think it's good to temper people explanations.

Yes PCI Express is faster but throughput isn't the Achilles heel of AGP it's the managament of read requests and posted writes. It's removing as much data from the FSB but still maintaining access to the CPU's fast caches when you really need them while keeping the whole show in sync.

Do not expect bandwidth alone to yield a noticeable increase. Nvidia is only purporting a %20 increase in speed and this is dependent on the driver managing write combine buffers efficiently as well as read request/posted writes.

If the WWDC Powermacs don't have PCIe don't freak. You'll still have a wicked machine and by the time you do get PCIe in your computer most of the drivers will be tuned to the efficiencies of PCIe.

So Ensign I guess we are in agreement more than disagreement.

Seems so. Thanks for the extra info.

I just can't get over how much emphasis the Apple Video guys put on the use of a Radeon 9800 Pro in their Motion demo at NAB. With ATi already announcing that all their next gen cards will be PCIe only, it just seems Apple is poised for early adoption.

It's a source, not a publisher.


What if what he knows is that there's someone around spreading crap? It certainly wouldn't be the first time....

I still take exception to Kasper's tone. It was unnecessarily rude and dismissive, especially in light of no specifics of his own. If he "knows" someone is spreading crap, then let's hear the details.

You're right and you're wrong. You're right the information is spurious, but wrong to think there is a person knowingly spreading misinformation.

No, apparently, the french site got the specs off a japanese blog and failed to realize that the blogger was just guessing the specs.

Perhaps the person who emailed you also could not read japanese.

Re paragraph 1:
No. I'm saying that there is an individual, who lives in canada, who is apparently french canadian, and has been submitting false reports for about 8 months now. Many of which have been published by semi major publications, including AppleInsider. The last one published by AppleInsider came about 4 months ago, and I will soon be labeling those reports accordingly.

I also am unprivileged to any knowledge or 'tman' or who his sources are, so no disrespect there.

Re paragraph 2:
This is where I am biased, though I wish I didn't feel that way. I feel that AI has been the most informative, and due to the circumstances, the most accurate (in a sense), of any publication following the Power Mac G5 rev 2 news. The machines were first seeded in Dec. I know this as fact. Problems arose, though its unclear if they were destine for Feb or Jan. We original heard Jan around the time of the seeding. Then there was a major issue with the supply of chips Apple planned to rev these units with. THIS WAS CONFIRMED BY APPLE. They have no 90nm in adequate quantities over 2GHz, and still do not. They were promised them. They expected them, and it did not happen. We reported on this prior and after the fact. This is believed to be the culprit behind the current hold up, and we've stated this.

Having said all that, I believe our information was and continues to be based on facts and the reality of the current situation.

And for the record, I was not intending to be insulting. I was intending to protect my members from this moron's drivvel.

I have a lot of reports on the PowerMac situation. FWIW the date I've heard for the past 3 months has been June 28th (WWDC). I consistently mentioned in articles to holding a date, but not publishing it due to lack of confirmation. After publishing multiple articles noting further delays (and what some believed to be appleinsider excuses), I was hesitant to throw such a blow on the main pages. I didn't want to fuel a fire in March, when I told readers there was reason to believe no Power Mac G5 updates would surface till almost july.

I've been gathering details and making attempts to confirm them. I have information I believe to be accurate, but I'm not ready to publish it in an article. And FWIW, I would love to just be able to tell you guys the status and details right here, but I'd be pretty unhappy when the information is lifted and published elsewhere the next morning. I want to make sure the information is solid, and then present it to all of you. That is always our goal. I would hope that this is what you all would like, as well?

But, in actuality, I just did tell you what I'm pretty certain of: the 975/3.2GHz rumors which are fabricated.

Hope this helps

Thanks Kasper. See, that wasn't so hard.

Please see tman's posts on pages 1 and 2. I'd be interested in your opinion.

Well, you seem to know tman better than I. Has he been accurate about predictions of this magnitude before? How do we know that the source he spoke to at the conference was not informed by the same material that I feel is misinforming all of us?

I dunno what my opinion is. My opinion is that these rumors have to come to me exclusively from a liar. And frankly, I'm sick of him, and I don't like what he does. And the success he attempts to destroy. I've spoken to many people, and it appears that the source of these rumors are one in the same.

I really can't stand this guy and what he does, though I have no clue as to what drives him.

Kasper

OMFG. Kasper, you're killing me. Really.

It's his first prediction of any kind and he freely admits that he has no way of verifying any of it. His conference source claimed to have first hand knowledge ("in the belly of the beast") and not merely repeating info from another source, French Canadian or otherwise.

I really don't understand all this anti-975 sentiment. Unlike the Power 4/970 development cycle, IBM knew from day one the Power 5 would spawn a low power, low cost version. This could result in the 975 coming to market much sooner after its big brother's debut then the 970 ever could have after the Power 4. In fact, this seems to be exactly what tman's source is saying.

I don't know anything either, just trying to be optimistic.

Well, Kasper, the conspiracy theorist in me suggests the possibility of an Apple-sponsored anti-rumor FUD-monger.

</levity>

Theoretically, it may be so that PPC975 was Apple's exclusive initiative. This might explain why no official information on it is available. However, I doubt that IBM can conceal the whole development of such a CPU only because Jobs told them so.

The only reason I tossed those components together for an example was because of the uninformed post that rudely said you cannot build a dual-Opteron system with a DVD-R drive, 160GB hard drive, 512MB RAM, video card, case and power supply could not be made for $1000 - $1300, and that anyone who said otherwise was a PC troll living in la-la land.

Remember - I went on in my post to say that you should never compare home built PCs to Macs. That being said, I'll try to answer your questions.


For an OS I would use Linux (or BSD). For applications, here are a few, but there are a lot more free apps that I would download: GNOME, Gimp, OpenOfffice, Apache web server, MySQL, Mozilla, WINE, LAME, Xmms, DivX Video, CDR-Toaster, Kino, Cinelerra, Blinder, Bluefish, Gnumeric, Rubrica 2, GNOME-PIM, MagicPoint


Keyboard and mouse - forgot them, add on another $50.00
Bluetooth, Firewire 800: and Modem: No.
Gigabit ethernet, fans, and 400 Watt power supply: Yes.


*sigh* Who cares? This was not a Mac vs PC comparison. But it does write DVD+R at 8X -- Lite-On DVD-RW/+RW Drive, Model LDW-811S/ 851S, Retail $79.00 (Write Speed: 8X DVD+R, 4X DVD-R, 4X DVD+RW, 2X DVD-RW, 40X CD-R, 24X CD-RW Read Speed: 40X CD-ROM, 12X DVD-ROM).

And for only $19 more you can get a 12X writer
Artec DVD+/-RW Drive, Model VOM-12E48X, Retail $97.99
(Write Speed: 12X DVD+R, 4X DVD+RW, 8X DVD-R, 4X DVD-RW, 40X CD-R, 24X CD-RWRead Speed: 40X CD-ROM, 12X DVD-ROM)


The Opteron 242 is clocked at 1.6GHz


Besideds the point, but thats why I always prefer to build PC rather then buy whole systems -- you know you are putting in quality parts and not some of the cheap crap you get stuck with when you buy a pre-built system.


Both systems? There is no other system. I was responding to another post and showing that you can build a Dual Opteron for $1000 - $1300. Of course it is the card that comes with the single processor Powermac...

Since I only build PCs for gaming I have no desire to build a dual processor computer. This systems was thrown together as a quick example, I did not spend a lot of time trying to get the lowest prices (I used mostly retail parts), Nor did I try to make it as powerful as I could have. I just saw someone make a ludicrous and insulting statement and felt compelled to respond. It was in no way meant to be a comparison between Macs and Opterons.

Now back to our regularly scheduled topic.

Does anyone have any solid information on the PPC975? I've been searching the web and all I've found are rumors. Is it on one of the published roadmaps?

So if they can barely get any 2 ghz parts now at the end of May,exactly what can we expect in a month? Another major drop in market share? Looks like apple poisons everything they touch!

Easy, easy... It simply looks like Jobs will soon have to end it all in public, presumably at WWDC. Sources close to Apple HQ claim he is going to do harakiri.

The AGP bus is a bottleneck when it comes to upstream comms. Increasing that from 1x to 16x speed will have a significant effect on the role of GPUs in the future and although you won't see it in first gen cards you can be certain you will in later gens.

Where are these sources close to Apple? I am assuming that this means Apple will not hit 3GHz as promised at WWDC.

And please let's not argue over the later revised time frame.

AFAIK there is no official word from IBM of any POWER5 derivative, so everything we "know" about 975 could be fabricated. On the other hand did Steve say at WWDC'03 that "the next generation G5" would run at 3 GHz "a year from now" and i really don't think that 970FX constitutes a next generation.

I find Kasper's comments quite interesting. And it is obvious, more or less, that Croquer is copying rumors here and there. As most representative case, I would mention the last updates in the Powerbook line. AppleInsider posted in the News Comments the exact model numbers, just to see them the next day in Croquer presented as a genuine source information .

But Kasper's comments are bad news too:




After that, I don't believe we will see anytime soon this so heavily discussed and anticipated 3 GHz G5. I am curious to see how Apple will cover this historical goof (didn't yet happen, but now I am confident it will).

The closest we have to ***CONFEEERMED!!!!11 is this http://www-901.ibm.com/servers/eserv...lreport_1.html

Awright, it's in chinese, but if you babelfish it, it'll become ... eh, sort of readable. But it's the number 975 hosted on IBM's site, at least. :-)

See this quote:

Let's presume Kasper is right (though, of course, everybody can be misled by deliberately false information or even fooled into believing it by providing 'sources' with the same misinformation). While all is not lost yet, this, IMHO, clearly indicates huge problems with CPU yields anyway. The 3GHz summer deadline is about to arrive no later than the 30th of September anyway, however you count, regardless of your religion, etc. I should say, 4 months are not such a great deal of time, considering the whole process from CPU manufacturing to computer assembly. I don't want to raise FUD, but Mr. Jobs personally is quickly approaching a difficult (or tense, at the very least) moment of his life. If everything turns out as it should, these problems will very soon get forgotten. Otherwise, he will have a hard time explaining why Apple didn't hold his promise. And, worse yet, he won't be able to blame IBM because IBM did not announce 3GHz CPUs this year, not to mention this summer.

Poor old Stevie...

ooooh.. nice! But I guess that this is just the contribution of the internal news department and they don't have any more information than we do.

Yeah, except for the fact that it's now officially non-vaporware!

Zapchud is right, seeing "PPC 975" on an IBM hosted site is huge all by itself.

Thanks Telomar I did a google search for AGP and Upstream and got linked to an article that explains what Ensign Pulver was talking about. My apologies for my ignorance. With all things being considered having much faster upstream is a feature worth having. In fact I was preoccupied with graphics cards and forgot that cards like ATI All in Wonder have video features and then you have the bandwidth constraints of HD video. I'm humbled yet excited...hopefully Powermacs can strike first. Plus reading on how PCI Express is excellent with Isochronous transactions I can only guess what benefits will become available to high end audio and video cards. Those using them are bound to be happy.

Back OT I am really hoping that there is a a 975 and in a way I believe that there "has" to be. I figure the die shrink of the 970 was good for a few hundred more megahertz at least but I don't think Apple can hit 3Ghz with these chips. I always assumed they would segue into the POWER5 derivative. It just seams a bit early but I'm hoping that something is working right here.

... According to IBM indicated that, the Power5 time arteries will be
situated between 1.6GHz to 2GHz, next year will release 90nm system
regulation according to the Power4+ model Power5+, when the arteries
might arrive 3GHz. IBM estimated 1.6GHz Power5 may achieve 1.3GHz
Power4 two time of potency. Moreover, IBM as the foundation constructs
the larger-scale multi-processor system take Power5, PowerPC 975 also
directly continues to use the Power5 core. But, reverse with initially
the Power4 strategy, Power5 as soon as will start by the lower step
server primarily, in the future only then gradually to extend to the
high step server. Behind this represents strategy transition, worth
noting ...

http://www.970eval.com/

A nice little tidbit here...

Well, it's sad that Kaspar's sources lead him to believe the 975 rumor is bogus. If that's the case, Apple is in a world of hurt. The 970fx bump was NEEDED at the beginning of the year/early spring. If all Apple can pull off is that speed bump at WWDC, I'll be sadly disappointed. But if Apple has half a clue - and sometimes we wonder - development of the next gen PMac continued despite the 970fx supply issue.

Kaspar did say Apple has, "no 90nm in adequate quantities over 2GHz". But the 975 rumors are that it is 130nm based. My hope is that the 90nm supply issue is only hurting the iMac now - the PMac bump window is gone.

Poor Apple they seem to be cursed with CPU supply issues. I hope they have a contract with IBM that provides penalties if IBM can't deliver. IBM's problems are costing Apple $$$. I have $ sitting in the bank earmarked for a new PMac, if Apple had bumped in Spring, I'd be typing this from a dual G5 rather than a G3/350. If all they do is release the Spring bump at WWDC, I may hold back till the next gen, because it can't (shouldn't) be very far away - maybe I will buy that CPU upgrade card to squeeze a little more life out of my old box.

*sigh*

- Jasen.

I think you'll find that those who buy Dual Processor Opteron or Xeon machines will not buy them at $1000. This is just silly. They want their hardware to be reliable. No one I've known with Dual proc machines has gone the cheap route, it's too expensive in the long run. If you're going to get a reliable Dual proc system, you're easily spending more than twice that. Remember, you not only want it to be more reliable but you also want the hardware to be well matched against eachother. In the examples shown here they are not. The G5s come like this out of the box, with the exception of the poor GeForce 5200.

Also, to get a decent spec system here in the UK (which I know is more expensive), I'm talking Athlon 3200 XP, Radeon 9800 SATA, DDR400, and the rest of the system to match you're talking about £1300 from building it yourself and without a screen. I admit the Powermacs are no way near on it for price, but a dual proc machine at the price some of you have suggested is ridiculous.

It' still vaporware, but it's official vaporware until we actually see them delivered.


I have always thought the 975 would be a 90 nm device, but I guess it won't need to since its parent, the POWER5, is 130 nm. But 970 was 130 nm, and its parent (POWER4) was originally 180 nm, so there's not much to go on here.

You are not the only one. I did not want to buy a Rev 1 G5, I have a Rev1 G3 PM and it is fraught with problems. It is obvious thatthe 130nm G5 is too hot and I wonder about he long term reliability of these machines. I would ahave been happy to buy a 2.6ghz 90nm machine six months ago which was the right time frame for a speed bump.

Quite frankly will know that nearly all their lost market share over the last four years can be attributed to poor processor performance. nearly everything else has been positive news stability, virus protection etc and software support. There are one or two notable exceptions such as Coral and to a lesser extent Adobe.

We will without doubt get something at the WWDC that we do now know, but what and by how much we just don't. I suspect that we will see 3ghz announced, but shipping later say September with immediate availability of G5 imacs and 2.4 or 2.6 PM's.

Sorry about the above botched post- restated as follows:

I don't know if you have ever attended a business function like the one I was at. I attend far too many. Everyone wore a lapel tag with name and company name. And I'm not predicting anything. I simply, right or wrong, believe the guy who told me this for reasons of my own that I have stated in this forum.

You're wrong -- they can and do conceal processor development projects routinely.



As for their current yield on 90 nm, earlier this month they (IBM) showed a graph of their target yield trend line vs. actual yields and it clearly showed serious issues through the early part of the year but they have improved to the point where they are almost where they wanted to be at this point (i.e. their improvement has outpaced the project rate of improvement, allowing them to gain quite a bit of ground). Considering that Apple obviously intended to ship on the projected yields at the start of the year and their current yields are now better than that, it would seem that Apple should now be in a position to ship the 970FX.

OK. I stand corrected. One question remains, though: does PPC 975 exist even on paper? As far as I can tell, there's only one public file on one official server in the whole world, which mentions PPC 975 only once. That may as well be a typo. Is it?

Here

I just got this nugget from hollywoodindustry.com

"Spurring development of 64-bit applications is an upcoming programming tool that will facilitate 64-bit software development. IBM will release a new compiler that will convert programming language written by software coders into a language the 64-bit processor can understand. Industry experts say the new compiler will allow 50% faster speeds compared to the open-source compiler, called gcc, that's been used since the G5's inception.

All these enhancements allow OS X to begin down the path to pure 64-bit processing, taking advantage of the wide bandwidth where possible and practical, while maintaining compatibility with 32-bit applications that are widely installed and available today."

The link for the full story is Here.

This plus dual 3ghz G5's is enough speed to trounce anything in its path!

I posted this info right here in this forum days ago (5th from end of page 2) and no one noticed till now?

BlueKnight