I personally feel that we will see the 2.3 Ghz chip in the summer release of the PMs. 2.3 is a nice place to be especailly with Intel going to climb past 3 Ghz by the time the 970 is officially announced in Macs. Probably this time next year, we will probably be in the high 3 Ghz range or maybe 4. We will defantely pass Intel not in this revision, but maybe in the next. These chips are going to scale.
Plus, like KidRed was saying, people will see the PM attractive at any price if Panther is going to be wicked fast and that it has the 970 in it. Apple just needs to thread Panther like crazy and we will be doing stuff that we never imagined before all at once.
Yep, 970 + Panther + great price = hmmmmmarket share
I doubt that a 970 @ 1ghz would CRUSH a DP1.42 G4.
I guess that depends on your definition of "crush". A 50-100% performance improvement at 42% lower clock rate is pretty crushing, IMO. And the 970 will debut at a higher clock rate than the 7455 is at (the 7457 may even the playing field but betting on two PPC chips being delivered in the same 6 month period isn't something I'm prepared to do!). As I said elsewhere, in some cases we are going to see a performance differential like the SPECmarks demonstrate -- those aren't purely synthetic benchmarks, they just aren't optimized for the G4 (SIMD, SMP, etc). Lots of code isn't optimized for the G4, and on that code I fully expect the G4 to get "crushed".
I personally feel that we will see the 2.3 Ghz chip in the summer release of the PMs. 2.3 is a nice place to be especailly with Intel going to climb past 3 Ghz by the time the 970 is officially announced in Macs. Probably this time next year, we will probably be in the high 3 Ghz range or maybe 4. We will defantely pass Intel not in this revision, but maybe in the next. These chips are going to scale.
I remember when such reckless things were being said when the first PowerMacs came out many long years ago.
Yep, 970 + Panther + great price = hmmmmmarket share
If Apple does just like what you said and that they offer some kind of emulation s/w designed by Apple along with it, Steve might just get his 5% or more.
I remember when such reckless things were being said when the first PowerMacs came out many long years ago.
Yeah, but Moto was fabbing. If things hold true, Apple will have the power of IBM behind them (an IBM which has just been contracted to build graphics chipsets for NVIDIA -- could signal things to come in terms of Apple offerings, but hopefully not in portables!).
Once the 970 hits .09 process, it could very easily scale up quickly. Thing is, the 970 prolly won't be around too long itself, with the 980 coming up in a matter of a few years.
Yeah, but Moto was fabbing. If things hold true, Apple will have the power of IBM behind them (an IBM which has just been contracted to build graphics chipsets for NVIDIA -- could signal things to come in terms of Apple offerings, but hopefully not in portables!).
Once the 970 hits .09 process, it could very easily scale up quickly. Thing is, the 970 prolly won't be around too long itself, with the 980 coming up in a matter of a few years.
Cross your fingers...
The NVIDIA thing looks very interesting. By the time Apple comes out of this year, maybe all the parts of their computer will be made by IBM. Now wouldn't that be an oxymoron as Apple and IBM were enemies in the 70s and 80s. How times change.
I doubt that a 970 @ 1ghz would CRUSH a DP1.42 G4. Benchmarks and real world are two different things, as far as I can see, keeping in mind that I don't know about the technical bits, just what I see.
IBM PPC 970 @ 1.8 GHz SPECfp2000 1051
Mot PPC G4+ @ 1.0 GHz Specfp2000 187
That's NOT a typo.
2*(1.42*187)=531
That implies a single 1.8GHz 970 has near double the floating point of a dual 1.42 G4. A dual 2.3 GHz 970 would have, um, MORE.
There are some fairly prominent tasks where the ppc970 will annihilate the G4+. For programs that actually use floating point, Specfp is a pretty good benchmark - it's a wide pile of actual science/engineering code.
It's true that this is 'benchmarking'. But don't doubt that there _are_ niches where any kind of 970 will destroy a G4. Period. (assuming they don't come out at 0.2 GHz or something silly.)
Another thread degenerates into price/performance.
OK:
First, Apple will offer the fastest configuration they can engineer at a non-absurd price. That's what they've been doing at the top of the line since at least the Mac SE. We're coming out of a period where their options were subject to unusual constraints, but it's through no fault of Apple's (at least, unless you expect them to have foreseen all the snags that the PPC and Motorola hit) that their top end hasn't held the speed crown for a while.
Second, although Apple will charge what the market will bear, it helps to remember that Apple (and especially Steve) takes an almost childlike pleasure in bringing previously inaccessible technology to affordable price points. That's what the first Macintosh did, in fact. AirSluf pointed out that people were shelling out $10K and up for the equivalent of the Cinema HD display before Apple rolled it out at the comparative bargain price of $4K, and before they turned around and slashed the price to $2K. The SuperDrive is another example, and so is AirPort. So, just because Apple has access to a near-workstation-grade chip doesn't mean they'll charge workstation-grade prices. You can bet they'll make a point of offering Sun performance at PC prices. Maybe not Matsu prices, but close enough.
I'm confident that the PowerMac G4 will no longer be offered more or less the moment that the 970 becomes available. I have a feeling that Steve is counting the days until he can take that albatross off his neck, and he'll do what he has to to see that it comes off. Also, I doubt the 970 will be any more expensive than the custom-made, absolute-bleeding-edge 7455s currently going into PowerMacs. If anything, it will cost less, since it appears that IBM is having no trouble at all producing them.
Fred Anderson has explicitly and repeatedly affirmed that Apple is trading margins for market share, so that argument is settled. Apple is going that way. They might not be doing it as aggressively as some would like, but market share won't mean anything if they can't afford to keep their nest egg, and their half-billion-dollar R&D budget, and their talent. (Before you disagree, look at their SEC filings: Apple has been basically breaking even operationally for the last several years. Their profits, when they have come, have come from Anderson's carefully engineered and well-maintained "cash machine." They don't have a lot of room to maneuver.) They've also been setting up and testing ways to get their message out to consumers more effectively (via Apple reps in CompUSA, and of course the Apple Stores).
Steve has said that Apple will close the megahertz gap. Of course, he said it when we were still stuck with the G4, but he obviously could see a lot farther forward than we can.
I'll believe that MacBidouille somehow got information from someone testing a 2.3GHz 970-based Mac. If IBM's making them, why not try them out? It's every bit in Apple's interest to unleash the absolute fastest Mac they can put together. It was in Apple's interest when the G3 was crushing Pentiums, and it certainly is now that the Pentium is crushing G4s.
As an aside, I believe it was Matsu who claimed that the Wintel world is accustomed to steady, reliable increases. Well, uh, as of the last couple of years, maybe. You don't have to think that far back to remember when Macs got steady, reliable increases and the x86 was faring poorly enough that any number of pundits were predicting its death. Those turned out to be abberations, and I'm confident that the Great 500MHz Plateau will be remembered as an abberation in the PowerPC's history as well.
Don't you know that everyone here knows how to run Apple better than Apple?
I've been called much worse.
My point was that all this whining about how Apple needs to cut their prices to boost market share is naive. Apple is a business, whose first obligation is making money for their shareholders. I hope folks will see that these are the same basic principles that drive any business. There have been a lot of businesses that went belly-up despite growing sales and market share. You cannot grow beyond what your products can profitably sustain.
One important difference, for Apple, between IBM and Motorola is their internal use of the CPUs.
For IBMs 970 to be useful to IBM they have to be competetive with Intels Xeon (at least for the blade servers, IBM might also have other plans for the 970 outside computer CPUs)
For Motorola the G4 is a good processor in embedded applications and they have no interest in the G4 as a computer CPU.
With such divergent goals for the G4 it is not stange if there has been a lot of armtwisting between Apple and Motorola, and that is allways a bad situation.
Regarding the PPC scaling. When the 9600/350 came out the Peniums were at 266 or 300 MHz IIRC. Then the G3 came that dropped a bit in clock speed but gained in performance So from the first PPC in 1994 to the G3/450 in the spring of 1999 the PPC was on par or a bit ahead of the Pentiums.
Then we had the G4 clock down, the AMD 1 Ghz Athlon starting a real race between Intel and AMD in January 2000, while the G4 would be stuck at 500 Mhz until 2001
So 1994-1999 the race was made by 601 and then the 604 and lastly the G3. All 3 CPUs made well and it was a good period only marred by the nonnative OS and applications during the nubus PPC year. 1999-2003 have not had 3 diffrent CPU generations but only one the G4 and the kindest I can say is: at least it has been consistent the 2x clock speed gap it lost in 1999 and 2000, it has keept 2001 and 2002 and 2003 without any signs of failing. If anything it seems to err on the safe side and sometimes archiving almost a 3X clock gap
Matsu Prices? Amorph, come on now. I've always said that I'm perfectly happy to pay a "little" more for macs of similar performance, and a little more again for macs of far greater performance. In the desktop realm this question will be settled with 970's (hopefully) so we are left to ruminate on price.
I say, if 970's provide the kind of performance that they suggest, then the CURRENT PM prices are acceptable, they are high end prices already, matching PPC970 performance to said prices would solve a lot of problems, and hopefully let the rest of the consumer line get respectable (without impinging on the PM line)
When, however, I take to ranting about the lack of a real consumer/edu/business machine, I not up the buisness of saying a PM ought to be priced like an eMachines, though the G4 PM's are iffy. I'm saying that they need to make a consumer model with far broader appeal. Clearly that is a headless machine of some sort with a basic degree of expansion/upgradeability.
Right now, Apple is missing such a machine. The iMac doesn't qualify for well document reasons.
The quoted speeds do make some sort of sense, taking rounding into account.
If Apple have a memory controller than can cope with a 450 MHz maximum bus speed (900MHz effective due to DDR transmission), and the PPC970 can run it's bus at an integer fraction of the core clock, then we get chips at 1/3, 1/4, and 1/5 giving clock speeds of 1.35GHz, 1.8GHz, and 2.25GHz, which might be rounded to give the 1.4, 1.8, 2.3 mentioned by MacBidouille. Any other speeds would require a slower bus, reducing bandwidth, maybe Apple didn't want to do that.
When, however, I take to ranting about the lack of a real consumer/edu/business machine, I not up the buisness of saying a PM ought to be priced like an eMachines, though the G4 PM's are iffy. I'm saying that they need to make a consumer model with far broader appeal. Clearly that is a headless machine of some sort with a basic degree of expansion/upgradeability.
I think we mostly agree on this, except that I don't think expansion/upgradeability is nearly as much of an issue as it used to be. An AGP card, an optional modem, an optional wireless node, a spare drive bay (beyond the main drive and DVD) and maybe a single PCI slot (but probably not). DIMM slots and ZIF mounted processors are much par-for-the-course. Most expansion needs are now either built-in or handled just fine by USB / FireWire. The box in this market space should be as small as possible because that would broaden its appeal far more than a couple of slots. And a fast G4 is plenty (a la the current high end iMac). The price should be at eMac levels or below, and this ought to be doable because the components are almost the same minus the display.
The quoted speeds do make some sort of sense, taking rounding into account.
If Apple have a memory controller than can cope with a 450 MHz maximum bus speed (900MHz effective due to DDR transmission), and the PPC970 can run it's bus at an integer fraction of the core clock, then we get chips at 1/3, 1/4, and 1/5 giving clock speeds of 1.35GHz, 1.8GHz, and 2.25GHz, which might be rounded to give the 1.4, 1.8, 2.3 mentioned by MacBidouille. Any other speeds would require a slower bus, reducing bandwidth, maybe Apple didn't want to do that.
michael
Instead of doing that I expect the memory controller to be run at half the processor speed. So what if bandwidth is a little lower on slower machines? It makes sense from a marketing and technical viewpoint.
Instead of doing that I expect the memory controller to be run at half the processor speed. So what if bandwidth is a little lower on slower machines? It makes sense from a marketing and technical viewpoint.
Yes, but what if Apple can't get the companion chip above the 900MHz rate?
Comments
Originally posted by kim kap sol
I do though breath down on their necks. Some of them find it errotic and I have to stop before they get wild ideas.
Whatever. THe PPC 970 is going to be great!! Does that meet your approval now??
Originally posted by Mac OS X Addict
I personally feel that we will see the 2.3 Ghz chip in the summer release of the PMs. 2.3 is a nice place to be especailly with Intel going to climb past 3 Ghz by the time the 970 is officially announced in Macs. Probably this time next year, we will probably be in the high 3 Ghz range or maybe 4. We will defantely pass Intel not in this revision, but maybe in the next. These chips are going to scale.
Plus, like KidRed was saying, people will see the PM attractive at any price if Panther is going to be wicked fast and that it has the 970 in it. Apple just needs to thread Panther like crazy and we will be doing stuff that we never imagined before all at once.
Yep, 970 + Panther + great price = hmmmmmarket share
Originally posted by Matsu
I doubt that a 970 @ 1ghz would CRUSH a DP1.42 G4.
I guess that depends on your definition of "crush". A 50-100% performance improvement at 42% lower clock rate is pretty crushing, IMO. And the 970 will debut at a higher clock rate than the 7455 is at (the 7457 may even the playing field but betting on two PPC chips being delivered in the same 6 month period isn't something I'm prepared to do!). As I said elsewhere, in some cases we are going to see a performance differential like the SPECmarks demonstrate -- those aren't purely synthetic benchmarks, they just aren't optimized for the G4 (SIMD, SMP, etc). Lots of code isn't optimized for the G4, and on that code I fully expect the G4 to get "crushed".
Originally posted by Mac OS X Addict
I personally feel that we will see the 2.3 Ghz chip in the summer release of the PMs. 2.3 is a nice place to be especailly with Intel going to climb past 3 Ghz by the time the 970 is officially announced in Macs. Probably this time next year, we will probably be in the high 3 Ghz range or maybe 4. We will defantely pass Intel not in this revision, but maybe in the next. These chips are going to scale.
I remember when such reckless things were being said when the first PowerMacs came out many long years ago.
Originally posted by BobtheTomato
I remember when such reckless things were being said when the first PowerMacs came out many long years ago.
Yeah, care to elaborate? Just curious. When the first PMs came out, geez when was that like 7 years ago?
Originally posted by KidRed
Yep, 970 + Panther + great price = hmmmmmarket share
If Apple does just like what you said and that they offer some kind of emulation s/w designed by Apple along with it, Steve might just get his 5% or more.
Originally posted by BobtheTomato
I remember when such reckless things were being said when the first PowerMacs came out many long years ago.
Yeah, but Moto was fabbing. If things hold true, Apple will have the power of IBM behind them (an IBM which has just been contracted to build graphics chipsets for NVIDIA -- could signal things to come in terms of Apple offerings, but hopefully not in portables!).
Once the 970 hits .09 process, it could very easily scale up quickly. Thing is, the 970 prolly won't be around too long itself, with the 980 coming up in a matter of a few years.
Cross your fingers...
Originally posted by fred_lj
Yeah, but Moto was fabbing. If things hold true, Apple will have the power of IBM behind them (an IBM which has just been contracted to build graphics chipsets for NVIDIA -- could signal things to come in terms of Apple offerings, but hopefully not in portables!).
Once the 970 hits .09 process, it could very easily scale up quickly. Thing is, the 970 prolly won't be around too long itself, with the 980 coming up in a matter of a few years.
Cross your fingers...
The NVIDIA thing looks very interesting. By the time Apple comes out of this year, maybe all the parts of their computer will be made by IBM. Now wouldn't that be an oxymoron as Apple and IBM were enemies in the 70s and 80s. How times change.
Originally posted by Mac OS X Addict
Yeah, care to elaborate? Just curious. When the first PMs came out, geez when was that like 7 years ago?
The first PowerMacs were introduced in 1994. Do the math.
Originally posted by Mr. Me
The first PowerMacs were introduced in 1994. Do the math.
Thanks.
Originally posted by Matsu
I doubt that a 970 @ 1ghz would CRUSH a DP1.42 G4. Benchmarks and real world are two different things, as far as I can see, keeping in mind that I don't know about the technical bits, just what I see.
IBM PPC 970 @ 1.8 GHz SPECfp2000 1051
Mot PPC G4+ @ 1.0 GHz Specfp2000 187
That's NOT a typo.
2*(1.42*187)=531
That implies a single 1.8GHz 970 has near double the floating point of a dual 1.42 G4. A dual 2.3 GHz 970 would have, um, MORE.
There are some fairly prominent tasks where the ppc970 will annihilate the G4+. For programs that actually use floating point, Specfp is a pretty good benchmark - it's a wide pile of actual science/engineering code.
It's true that this is 'benchmarking'. But don't doubt that there _are_ niches where any kind of 970 will destroy a G4. Period. (assuming they don't come out at 0.2 GHz or something silly.)
OK:
First, Apple will offer the fastest configuration they can engineer at a non-absurd price. That's what they've been doing at the top of the line since at least the Mac SE. We're coming out of a period where their options were subject to unusual constraints, but it's through no fault of Apple's (at least, unless you expect them to have foreseen all the snags that the PPC and Motorola hit) that their top end hasn't held the speed crown for a while.
Second, although Apple will charge what the market will bear, it helps to remember that Apple (and especially Steve) takes an almost childlike pleasure in bringing previously inaccessible technology to affordable price points. That's what the first Macintosh did, in fact. AirSluf pointed out that people were shelling out $10K and up for the equivalent of the Cinema HD display before Apple rolled it out at the comparative bargain price of $4K, and before they turned around and slashed the price to $2K. The SuperDrive is another example, and so is AirPort. So, just because Apple has access to a near-workstation-grade chip doesn't mean they'll charge workstation-grade prices. You can bet they'll make a point of offering Sun performance at PC prices. Maybe not Matsu prices, but close enough.
I'm confident that the PowerMac G4 will no longer be offered more or less the moment that the 970 becomes available. I have a feeling that Steve is counting the days until he can take that albatross off his neck, and he'll do what he has to to see that it comes off. Also, I doubt the 970 will be any more expensive than the custom-made, absolute-bleeding-edge 7455s currently going into PowerMacs. If anything, it will cost less, since it appears that IBM is having no trouble at all producing them.
Fred Anderson has explicitly and repeatedly affirmed that Apple is trading margins for market share, so that argument is settled. Apple is going that way. They might not be doing it as aggressively as some would like, but market share won't mean anything if they can't afford to keep their nest egg, and their half-billion-dollar R&D budget, and their talent. (Before you disagree, look at their SEC filings: Apple has been basically breaking even operationally for the last several years. Their profits, when they have come, have come from Anderson's carefully engineered and well-maintained "cash machine." They don't have a lot of room to maneuver.) They've also been setting up and testing ways to get their message out to consumers more effectively (via Apple reps in CompUSA, and of course the Apple Stores).
Steve has said that Apple will close the megahertz gap. Of course, he said it when we were still stuck with the G4, but he obviously could see a lot farther forward than we can.
I'll believe that MacBidouille somehow got information from someone testing a 2.3GHz 970-based Mac. If IBM's making them, why not try them out? It's every bit in Apple's interest to unleash the absolute fastest Mac they can put together. It was in Apple's interest when the G3 was crushing Pentiums, and it certainly is now that the Pentium is crushing G4s.
As an aside, I believe it was Matsu who claimed that the Wintel world is accustomed to steady, reliable increases. Well, uh, as of the last couple of years, maybe. You don't have to think that far back to remember when Macs got steady, reliable increases and the x86 was faring poorly enough that any number of pundits were predicting its death. Those turned out to be abberations, and I'm confident that the Great 500MHz Plateau will be remembered as an abberation in the PowerPC's history as well.
Originally posted by Nonsuch
Prepare to be branded an apologist.
Don't you know that everyone here knows how to run Apple better than Apple?
I've been called much worse.
My point was that all this whining about how Apple needs to cut their prices to boost market share is naive. Apple is a business, whose first obligation is making money for their shareholders. I hope folks will see that these are the same basic principles that drive any business. There have been a lot of businesses that went belly-up despite growing sales and market share. You cannot grow beyond what your products can profitably sustain.
For IBMs 970 to be useful to IBM they have to be competetive with Intels Xeon (at least for the blade servers, IBM might also have other plans for the 970 outside computer CPUs)
For Motorola the G4 is a good processor in embedded applications and they have no interest in the G4 as a computer CPU.
With such divergent goals for the G4 it is not stange if there has been a lot of armtwisting between Apple and Motorola, and that is allways a bad situation.
Regarding the PPC scaling. When the 9600/350 came out the Peniums were at 266 or 300 MHz IIRC. Then the G3 came that dropped a bit in clock speed but gained in performance
Then we had the G4 clock down, the AMD 1 Ghz Athlon starting a real race between Intel and AMD in January 2000, while the G4 would be stuck at 500 Mhz until 2001
So 1994-1999 the race was made by 601 and then the 604 and lastly the G3. All 3 CPUs made well and it was a good period only marred by the nonnative OS and applications during the nubus PPC year. 1999-2003 have not had 3 diffrent CPU generations but only one the G4 and the kindest I can say is: at least it has been consistent
I say, if 970's provide the kind of performance that they suggest, then the CURRENT PM prices are acceptable, they are high end prices already, matching PPC970 performance to said prices would solve a lot of problems, and hopefully let the rest of the consumer line get respectable (without impinging on the PM line)
When, however, I take to ranting about the lack of a real consumer/edu/business machine, I not up the buisness of saying a PM ought to be priced like an eMachines, though the G4 PM's are iffy. I'm saying that they need to make a consumer model with far broader appeal. Clearly that is a headless machine of some sort with a basic degree of expansion/upgradeability.
Right now, Apple is missing such a machine. The iMac doesn't qualify for well document reasons.
If Apple have a memory controller than can cope with a 450 MHz maximum bus speed (900MHz effective due to DDR transmission), and the PPC970 can run it's bus at an integer fraction of the core clock, then we get chips at 1/3, 1/4, and 1/5 giving clock speeds of 1.35GHz, 1.8GHz, and 2.25GHz, which might be rounded to give the 1.4, 1.8, 2.3 mentioned by MacBidouille. Any other speeds would require a slower bus, reducing bandwidth, maybe Apple didn't want to do that.
michael
Originally posted by Matsu
When, however, I take to ranting about the lack of a real consumer/edu/business machine, I not up the buisness of saying a PM ought to be priced like an eMachines, though the G4 PM's are iffy. I'm saying that they need to make a consumer model with far broader appeal. Clearly that is a headless machine of some sort with a basic degree of expansion/upgradeability.
I think we mostly agree on this, except that I don't think expansion/upgradeability is nearly as much of an issue as it used to be. An AGP card, an optional modem, an optional wireless node, a spare drive bay (beyond the main drive and DVD) and maybe a single PCI slot (but probably not). DIMM slots and ZIF mounted processors are much par-for-the-course. Most expansion needs are now either built-in or handled just fine by USB / FireWire. The box in this market space should be as small as possible because that would broaden its appeal far more than a couple of slots. And a fast G4 is plenty (a la the current high end iMac). The price should be at eMac levels or below, and this ought to be doable because the components are almost the same minus the display.
Originally posted by mmicist
The quoted speeds do make some sort of sense, taking rounding into account.
If Apple have a memory controller than can cope with a 450 MHz maximum bus speed (900MHz effective due to DDR transmission), and the PPC970 can run it's bus at an integer fraction of the core clock, then we get chips at 1/3, 1/4, and 1/5 giving clock speeds of 1.35GHz, 1.8GHz, and 2.25GHz, which might be rounded to give the 1.4, 1.8, 2.3 mentioned by MacBidouille. Any other speeds would require a slower bus, reducing bandwidth, maybe Apple didn't want to do that.
michael
Instead of doing that I expect the memory controller to be run at half the processor speed. So what if bandwidth is a little lower on slower machines? It makes sense from a marketing and technical viewpoint.
Originally posted by Programmer
Instead of doing that I expect the memory controller to be run at half the processor speed. So what if bandwidth is a little lower on slower machines? It makes sense from a marketing and technical viewpoint.
Yes, but what if Apple can't get the companion chip above the 900MHz rate?
michael