90nm G5 Supplies Improving?

[tortoise]

Ah c'mon babies. 2.6 looks pretty good over here, outside of Stevo's Reallity Distortion Field. Heck,its better than AMD's top @ 2.4 ghz.

Heh. The difference is that the AMD64 systems are available NOW, and it isn't like AMD is sitting still with their new product lines. The short-term product pipeline for AMD looks very good. While we use OSX for workstations, the heavy lifting is done by Opteron systems (sorry, PPC can't touch them for memory performance and scalability), and I am quite pleased with their broad and agressive AMD64 roadmap.

Beside which, clock for clock, AMD64 is a bit faster than PPC on average (excepting the DSP stuff which the PPC is quite excellent at). Yeah, I know, flame on.

 

[BWhaler]

I'm waiting to give Apple 5+ grand until a) MANDATORY refresh to the displays, and b) a Rev B G5. A 3.0 Ghz would inspire me more, but with new displays, I could live with a dual 2.6 or 2.8.

The displays are the main blocker in my opinion.

 

[tortoise]

Dual cores and Apple's uphill battle

Its been said that the PPC970 or at least its future derivatives were meant to be multi-core. From what I understand about multi-core processors, the end result is essentially double the speed... or better put (for the average lay-person) double the MHz. Its not like dual processor configurations where applications need to be multiprocessor aware but (again) might best be regarded as double the speed.

Nope, you have it incorrect. A dual core processor is exactly like a dual CPU machine. The difference is that they fit both CPUs in a single socket. So you don't get more speed, you just effectively get two processors in the space and packaging that would have normally only contained one. Which isn't a bad thing at all.

AMD is doing something very cool with their upcoming dual core AMD64 chips: they will be socket compatible with the single core chips. This means that you can take any single processor system and effectively upgrade it to dual processor by swapping out the CPU. Or upgrade a dual processor to a quad processor.

I don't know that the socket compatibility thing would work with PPC though, at least not as the PPC is currently designed. The Opteron memory and I/O subsystems can easily scale to a large number of cores, but the PPC systems aren't so hot at two cores. In the much larger picture, this is really the Achille's heel of the PPC chips actually, and the reason they will have a hard time competing with the AMD64 systems for a lot of real-world tasks. The AMD64 systems are a subsantially more balanced architecture. And that Apple really doesn't exploit the capabilities of the HyperTransport fabric doesn't help.

 

[omnivector]

Nope, you have it incorrect. A dual core processor is exactly like a dual CPU machine. The difference is that they fit both CPUs in a single socket. So you don't get more speed, you just effectively get two processors in the space and packaging that would have normally only contained one. Which isn't a bad thing at all.

AMD is doing something very cool with their upcoming dual core AMD64 chips: they will be socket compatible with the single core chips. This means that you can take any single processor system and effectively upgrade it to dual processor by swapping out the CPU. Or upgrade a dual processor to a quad processor.

I don't know that the socket compatibility thing would work with PPC though, at least not as the PPC is currently designed. The Opteron memory and I/O subsystems can easily scale to a large number of cores, but the PPC systems aren't so hot at two cores. In the much larger picture, this is really the Achille's heel of the PPC chips actually, and the reason they will have a hard time competing with the AMD64 systems for a lot of real-world tasks. The AMD64 systems are a subsantially more balanced architecture. And that Apple really doesn't exploit the capabilities of the HyperTransport fabric doesn't help.

that's not entirely true. dual-core machines will be faster than dual-cpu, because there's far less latency to get from one chip to another since they are both on the same die.

 

[ltgator333]

I don't know that the socket compatibility thing would work with PPC though, at least not as the PPC is currently designed. The Opteron memory and I/O subsystems can easily scale to a large number of cores, but the PPC systems aren't so hot at two cores. In the much larger picture, this is really the Achille's heel of the PPC chips actually, and the reason they will have a hard time competing with the AMD64 systems for a lot of real-world tasks. The AMD64 systems are a subsantially more balanced architecture. And that Apple really doesn't exploit the capabilities of the HyperTransport fabric doesn't help.

Yeah Sun's UltraSPARC IV chip is a dual core- basicly two UltraSPARC III cores pasted together and unless your code is optimised to take advantage of both it wont. Totally agree, it doesn't necessarily double or increase performance at all.
As for the dual core PPC chips, PPC 970 is as stripped down single core version of IBM's POWER 4 which is a dual core that IBM has been implementing in their PPC based servers and workstations since 2000. POWER 5 is also a dual core + Symmectric Multi-Threading, so it's effectively got at least 4 logical cores, and actually I think more. Basicly, IBM will have no trouble making a dual core chip, but I doubt the chips they will be supplying Apple will be dual cores, but whatever chip they derive from POWER 5 I would think will probably have the SMT capability (works a lot like Intel's Hyper-Threading but better). Also, POWER 5 has an on-die memory controller and I beleive uses a full implementation of hyper-transport for the FSB, so anything AMD may have on us now is just something that IBM's playing with right now or have been doing for a while already.

I actually think that is the one thing that really sucks about having IBM make Apple's chips- IBM will always have that POWER x chip that it only makes for IBM, and their PPC chips will always just be a stripped down version of their last POWER chip.. so actually the tech your seeing now is somewhat they're yesterday's news and half baked ideas for tomarrow.. 970 uses POWER 4's core which is yesterday's chip and has a kinda-hypertransport bus, which is a half baked idea for tomarrow's chip POWER 5....

 

[JFreak]

they should drop single-cpu powermac altogether. duals for all models, please.

 

[thatwendigo]

I don't think wanting a 3 GHz G5 is asking too much... I personally think they should be aiming for higher than 3 GHz for one year... just my $0.02.

Because, you know, Intel jumped a whole gigahertz last year.

You never know...maybe 2.6GHz will replace the entry level 1.6GHz, and we'll have 3GHz systems 'paper launched' at WWDC, with actual availability in September, just like the original G5 launch.

I could be dead wrong, but let's put it this way... I think Apple is being tight, tight, tight about anything they release right now. Remember that warning to the Seed program last week? If they don't even want programmers talking about what's in the OS update, what do you think they would do to anyone who leaked that there were machines over 2.0ghz?

Also... Would Steve let there even be a remote possibility that the 3.0s would be preannounced? This is, at best, an appetite whetter. It's to get attention from people like us that slaver over the rumors.

In fairness to Apple, I don't think Apple itself has spent 12 months preparing for Rev. B...more like they've waited this long for IBM to solidify their supplies of faster PPC970 and PPC970FX chips...which is what the article was about.

Exactly.

3GHz or not, a 600MHz increase over the previous top-end (or put it another way, a 30% jump) is quite a welcome improvement, and will push the G5s back into serious competition with newer Opteron, Athlon 64 and Pentium 4 systems.

It's more than 200mhz and dropping the line, that's for sure.

WTF? No 3GHz within a year like steve promised? AAAAAAH!!! THIS IS ANARCHY! Not to mention bad business, like, oh "100 million songs within a year and 30 million of that will come from the iTunes/Pepsi giveaway!" NOT. "3GHz within a year!" NOT. C'mon apple, get with the program already.

Psssst... It hasn't been a year yet.

Here is the only way I see steve making good on his promise to deliver a 3 ghz power mac. Here is the lineup:

dual 2.2 ghz
dual 2.6 ghz
single 3.0 ghz

There will also be a slash in cinema displays. I have spoken ...

Except that a dual 2.6 would slaughter a single 3.0. The standard gain from SMP-aware applications is around 50%, so those dual 2.6s will be more like single 3.9ghz machines.

Well we are only holding Steve to his word. He could eliminate such criticism if he was more careful with his wording. To make statements, as he did, about hitting 3GHz within a year, and then, presumably, failing to do so, isn't any fault of ours. How can u hold it against people for comparing what steve said and the apparent reality of the situation? Damn that reality distortion field. *shakes fist as sky*

Wrong, bucko. You're shouting before the deadline is even past. Steve hasn't failed until the end of the Keynote at WWDC, and even then, the wording makes it possible that he could claim that "summer" means August.

How about you actually do what you say you're doing and wait for the "apparent reality of the situation," hmmmm?

Basicly, IBM will have no trouble making a dual core chip, but I doubt the chips they will be supplying Apple will be dual cores, but whatever chip they derive from POWER 5 I would think will probably have the SMT capability (works a lot like Intel's Hyper-Threading but better). Also, POWER 5 has an on-die memory controller and I beleive uses a full implementation of hyper-transport for the FSB, so anything AMD may have on us now is just something that IBM's playing with right now or have been doing for a while already.

Read about it here.

Single-chip Power5s present as four cores (two logical, two virtual, according to the article), with 120 registers for integer and FP, 8 execution units, and a shared 1.92MB L2 cache (3 640k caches on independent buses). The chips operate and a 128bit-128bit dual single-directional bus (functionally a 256-bit 2Ghz FSB) with a half-clock interconnect to other MCMs (1ghz 128-bit), and an on-chip DDR memory controller with a 6GB/s I/O bus. The 32MB L3 cache and memory bus are separated, allowing for a theoretical memory bandwidth of 20GB/s, with the L3 on a 1GHz on-die bus.

From the article:
In any case, a typical 64-way SMP POWER5 system composed out of 32 chips will support 1TB RAM at start.

..

Well, if you just repeat the system density of p655, and have two of such systems in every 4 U of rack height, there comes a 2 TFLOPs cluster with 2 TB RAM with a peak memory bandwidth of more than 2 TB / second, all in one rack, with a bit of room to spare! Just remember a proper fast interconnect with distributed shared-memory capability...

I think there would be no problem at all making the 975 a dual-core chip with a memory controller on-die. In other words... VROOOOM!

 

[JFreak]

only problem is that ibm might want to keep powerX series a bit more advanced than the 97x and it's just likely they drop some features.

 

[Fiveos22]

My 55inch HDTV from Mitshubushi was less then a 23inch HD monitor.

No one doubts how amazing those Apple displays are but given the other flat panal options out there I have seen - for half the price you can get 90% as good a display.

However nice it is to compare the mitsu TV with the monitor, it is quite incorrect. Resolution-wise the 23 inch HD monitor smokes that 55 inch TV. Original TV (vga) is 640x480, Your HDTV ($1799 currently) has a max resolution of 480p (for reference HDTV 720p is 1280x720), Apple 23 inch Cinema has a max resolution of 1920x1200. The difference is noticable and the price difference is to be expected (this is purely in regards to resolution, not other things such as USB ports, etc.).

Your TV has a contrast ratio of 50:34 whereas the Cinema display is 300:1. All of these features need to be taken into account with the price. For an lcd screen of that resolution and size the price is at a premium (think about how many pixels there are and how few are allowed to be dead for an lcd to be sold).

The crux of my argument is, a TV (HD or otherwise) is no replacement for a computer monitor (my $390 Dell has a higher resolution than your TV)...and therefore the two cannot be compared in this instance.

Apple's high end panel is one of the most economical for its quality, they shouldn't have to cut prices to sell them.

 

[foniks2020]

2 x 2 already faster than 3 x 1 intel and amd

All I know is that the 2.0 duals are already faster than the available 3.0 intel and amd chips.... anything more is just kerosene on the fire. Benchmarks will tell you that in optimal conditions that the intel 3.x systems compete with dual 2.0s but in real world situations where a user has email(Outlook), Browser(IE) and anything else other than Photoshop, etc... then those x86 chips don't do nearly as well.... how many of you have a dedicated machine for your work and a secondary for daily tasks? A few... how much productivity is lost switching between them?

In any case dual 2.6 would bring you up to and beyond par with anything out there for many months to come. Plus 10.4 should add some performance gains on the software front.. while windows continues to be "patched" and gets more bloated every day, ruining performance at every turn.

 

[ssamani]

Well we are only holding Steve to his word. He could eliminate such criticism if he was more careful with his wording. To make statements, as he did, about hitting 3GHz within a year, and then, presumably, failing to do so, isn't any fault of ours. How can u hold it against people for comparing what steve said and the apparent reality of the situation? Damn that reality distortion field. *shakes fist as sky*

PS_people like you, who do nothing but make excuses for steve's mistakes, aren't any better then those of us who complain about him not living up to his word

Well I guess this is why Apple rarely comments on future plans. If they don't hit their targets, sometimes not their own fault, they get crucified. From all the news around the 970FX and 90nm processing for IBM and Intel it is clear that the issue is out of Apple's hands and both IBM and Intel have been caught out with the move to 90nm. If you look at the only two predictions Steve has made recently we have:

- iTMS passes its annual sales target after only a few months and Steve upgrades the target to something that is amazingly high. Pepsi screws up its own promotion, missing expected targets by 400% (5m vs 25m) - partly Apple's fault, but largely down to Pepsi not distributing codes fast enough. The news agencies, including reputable ones with no Apple axe to grind like the BBC, report that iTMS missed its "original" target, even though that had been passed 7 or 8 months previously.

- Steve predicts the G5 will hit 3GHz within a year, no-one is sure whether he means from when he said it or from ship date, giving a 3 month window. Five months later the XServe G5 is delayed and it is very clear to everyone that IBM had problems moving to 90nm just like Intel and is hitting delays and everyone complains that Apple isn't keeping its promise, even though it is anywhere up to 7 months before the target date. Hello? Apple publicly blames IBM for the XServe delay and you don't slag off your *key* partner in public unless you are very sure that they are at fault. So what does everyone say? Oh Steve to keep his promise and we are still up to four months away from the anniversary of the G5 shipping.

No wonder they never make future predictions

There are a couple of concepts here that have gone missing, chiefly innocent until proven guilty. Unfortunately we've gone down the road of passing as guilty six months prior to the crime being committed. I for one consider this Thought Crime to have been prosecuted on the basis of a Minority Report.

Now if you want to slag Apple off for something, there's plenty - we have the fact they had to settle over G3 Beige support from OS X, iBook motherboard issues, eMac display issues, PB display spotting, abysmal Finder, inconsistent use of brushed metal, etc., etc., which are all within the remit of their design, manufacture or QA (i.e. I am not an Apple apologist). CPU clock speeds are out of their hands. The other principle being missed is don't shoot the messenger of IBM's fab problems.

 

[ZildjianKX]

they should drop single-cpu powermac altogether. duals for all models, please.

I say drop the DP line and go SP dual core... it's going to be silly seeing a huge G5 powermac holding two processors when you'll be able to get a dual core minitower PC pretty soon.

 

[aussiemac86]

I dont see apple moving many more of the current PM line up with the present PM pricing. As the majority of people to purchase PM's are pro users they would have a fairly good idea of whats coming up soon, and hardly anyone who knows there must be new PM's in the next 2 months is going to pay the same amount as they will be for an inferior machine now.

Apple should release maybe a 2.2 and drop some prices just as a teaser and then relase the big guns at WWDC.

 

[mj_1903]

Plus 10.4 should add some performance gains on the software front.. while windows continues to be "patched" and gets more bloated every day, ruining performance at every turn.

Having seen a 10.4 build (8A68 from memory) running, I can assure you it runs faster, but at this time it crashes faster as well.

Don't bother asking me about new features though, there were none that I could see. The branch I saw was the development branch for the kernel and the Cocoa frameworks.

 

[thatwendigo]

I say drop the DP line and go SP dual core... it's going to be silly seeing a huge G5 powermac holding two processors when you'll be able to get a dual core minitower PC pretty soon.

Or we could go dual-processor, dual-core and have, at minimum, four processors per machine. With SMT, that would be eight processors. Also, "pretty" soon is next year, at the earliest. Jonah is already a year off target, though I expect some serious traction once Intel realigns their engineering teams.

Also endian seems to be pretty on top of things, and they show no AMD plans to go dual core before at least 2006. Intel is slated to ship Jonah in the second half of 2005, or more than a year from now, after it's already slipped once.

Apple should release maybe a 2.2 and drop some prices just as a teaser and then relase the big guns at WWDC.

No dice. It's a logistical problem if they do that, aside from the lowered pricing. Also, if something happens between now and WWDC, the lowered prices would hurt Apple in the long run.

 

[sun-ice]

G5 Update

What about this:

- PowerMac DP G5 2.6 GHz
- PowerMac DP G5 2.4 GHz
- PowerMac DP G5 2.2 GHz

- iMac SP G5 2.2 GHz 20'
- iMac SP G5 2.0 GHz 17'
- iMac SP G5 2.0 GHz 15'

This could be pretty nice ?

This introduce G5 in all desktop (exept eMac) and prepare the G5 LapTop...
With Luck, we could say good bye to 32Bit processors...!!!

 

[reyesmac]

I don't care how fast the top end is as long as its 3ghz+ and the low end is the exact same model as the 2ghz that is currently selling. None of that crippled motherboard stuff. Its been too long since an update. Maybe bringing the price down to what a powermac used to cost before the G5 would be nice too. Apple used to reduce the price of the Powermac for a while before they came out with new models, I don't know what their problem is these days. Its not about volume for them anymore, its all about profit margin.
Its too early in the morning for me to see this as anything but another lack luster update. Its nice that the Mac platform will reach 3ghz, but most of the people out there wont experience that speed until a year or two later when they put that chip on a cheaper mac. Most business I know don't buy the top of the line models.

 

[dragula53]

Wrong, bucko. You're shouting before the deadline is even past. Steve hasn't failed until the end of the Keynote at WWDC, and even then, the wording makes it possible that he could claim that "summer" means August.

He said 12 months, not next summer

 

[.a]

... within 12 months . i just wonder what people would have said, when steve jobs told us "the next revision of the g5 will be within 12 months" ¿?
.a

 

[Skiniftz]

.... I like stable macs...

You must have really hated OS9...

 

[CmdrLaForge]

Ah c'mon babies. 2.6 looks pretty good over here, outside of Stevo's Reallity Distortion Field. Heck,its better than AMD's top @ 2.4 ghz and it seems that Intel is coming back to 2-2.5 ghz range with its forthcoming line of CPU's.

Exactly - most are missing the point. 2.6 is pretty fast and even faster then AMDs top CPUs. And the speed increase between 2.0 and 2.6 is significant.

In terms of solutions - why is a dual 3.0 a solution and dual 2.6 is not ? Its just about some stupid numbers. If they would follow the same numbering scheme that AMD used for a while now - everybody would be happy. Just call the 2.6 a 3.0. Most wouldn't even know the difference - same as AMD !

 

[Skiniftz]

G5 iMac?

 

[sun-ice]

I prefer the actual design of iMac...
And I hope Apple will keep the same white half buble for the G iMac...

(sorry for your "mix"... but...)

white => low perform

aluminium => high perform

 

[gekko513]

I wish people would be a little more sensible and consider the situation as it is rather than getting caught up in predictions that were made last year.

The processor manufacturers have had more trouble making the transition to 90nm than were expected. The prediction was most likely based on a less troublesome transition.

I for one will not get pissed if we don't get 3.0GHz. But even if I expect the worst, I hope for the best.

 

[PPC970FX]

Water

Maybe Vapo will be inn the 3Ghz G5, or they will have PPC975. Dual dual core 3Ghz. 1,5Ghz FSB with 725 Mhz DDR2 RAM.