[Simon]

According to this article from CeBit the IBM PPC 970 is here and it is running at 1.8GHz. They expect to go to 2.3-2.5GHz with the next revision while staying at .13microns. IBM was quoted as saying Apple could use the chip.

I suppose this is pretty good news.

The article: http://www.macguardians.de/fullstory.php?p=1830

The interesting part in (original) German:

Erstaunlicherweise mussten wir feststellen, dass die schon im Vorfeld angekündigte Vorstellung des PPC 970 leider abgesagt wurde. Auf Nachfrage erfuhren wir, dass der erwartete Prozessor (2,3 bis 2,5 GHz) den Ausblick auf die übernächste Prozessor-Generation darstellt und nicht mit einem baldigen Erscheinen zu rechnen ist. Aktuell sind PPC 970 Prozessoren (nach wie vor in 0,13 Mikron gefertigt, das gilt auch für den kommenden 2,3/2,5 GHz Prozessor) mit einer Leistung von bis zu 1,8 GHz, die laut IBM auch von Apple in Zukunft verbaut werden könnten. Aber Genaueres weiß man wie immer nicht.

And here is my translation (it's probably far from perfect, but it's for sure better than the computer translation I got):

To our suprise we had to realize that the presentation of the PPC 970 which had been announced in advance has been canceled. Upon request we were told that the expected processor (2.3 to 2.5GHz) resembled an outlook onto the after next generation and its appearance shouldn't be expect soon. The current processor is the PPC 970 (still 0.13 micron fabrication, which will also be true for the coming 2.3/2.5 GHz processor) with up to 1.8 GHz, which according to IBM, could be used by Apple as well in the future. But as usual one doesn't know anything more precisely.

[Lateralus]

Except, that it already IS at 2.5GHz. And IBM is not only manufacturing them at those speds now, but they will be using them in their own systems momentarily.

[Centris650]

Yeah, I was reading this over at www.macrumors.com and I'm pretty excited. The questions I have are...

1) When? There are rumors of a summer/fall release. Though I'm not holding my breath!

2) Cost? Knowing how Apple has priced it's Pro line in the past I'd be worried of a costly pMac. However with Apple cutting prices like it has who knows?

I've delayed my mac purchase for now and try to squeeze a few more months out of my aging iBook and wait and see what happens. At the very least I might get a discounted DP 1.42 when the 970 comes out.

[Agasthya]

Originally posted by Centris650:
Yeah, I was reading this over at www.macrumors.com and I'm pretty excited. The questions I have are...

1) When? There are rumors of a summer/fall release. Though I'm not holding my breath!

2) Cost? Knowing how Apple has priced it's Pro line in the past I'd be worried of a costly pMac. However with Apple cutting prices like it has who knows?

I've delayed my mac purchase for now and try to squeeze a few more months out of my aging iBook and wait and see what happens. At the very least I might get a discounted DP 1.42 when the 970 comes out.

1) Rumor is that they will be demoed at WWDC in May
2) Who knows, probably more expensive than the current Macs. Apple is always looking for an excuse to charge more money

[Thain Esh Kelch]

Originally posted by PowerMacMan:
Except, that it already IS at 2.5GHz. And IBM is not only manufacturing them at those speds now, but they will be using them in their own systems momentarily.

YOU obviously has absolutely NO idea of what you are talking about.

[clarkgoble]

Who knows, probably more expensive than the current Macs. Apple is always looking for an excuse to charge more money

I'm not convinced of that. I think Apple recognizes that it needs to increase marketshare. They may have a premium on the high end systems, but I'm betting that most 970 systems will be priced competitively with Dell.

[:XI:]

Originally posted by agasthya:
Apple is always looking for an excuse to charge more money

Show me any company that isn't always looking for an excuse to charge more money. I thought the who reason of being a business was to make money.

[KidRed]

Originally posted by clarkgoble:
Who knows, probably more expensive than the current Macs. Apple is always looking for an excuse to charge more money

I'm not convinced of that. I think Apple recognizes that it needs to increase marketshare. They may have a premium on the high end systems, but I'm betting that most 970 systems will be priced competitively with Dell.

I agree except for the high end. Apple has lowered it's prices across the board and recently acknowledged more agressive pricing to gain market share. I don't see how anyone can longer say that Apple charges too much after the display cuts. Go to the Apple store and compare the prices to 1 year ago. When the 970's come out I expect the same prices if not slightly lower.

[SMacTech]

Originally posted by :XI::
Show me any company that isn't always looking for an excuse to charge more money. I thought the who reason of being a business was to make money.

Gees, now I know what I have been doing wrong all these years.

[SalBaker]

Originally posted by KidRed:
I agree except for the high end. Apple has lowered it's prices across the board and recently acknowledged more agressive pricing to gain market share. I don't see how anyone can longer say that Apple charges too much after the display cuts. Go to the Apple store and compare the prices to 1 year ago. When the 970's come out I expect the same prices if not slightly lower.

I will bet a nickel that when Apple launches the 970 PowerMacs the high-end (screaming-fast machine) will be expensive, and we will all be saying (again), "Don't buy the high-end, it is too expensive. What was Apple thinking! Buy the mid or low-end 970, they're a better deal!"

[slider]

I bet we don't see the prices budge from their current positions. Wouldn't it be something though. It's funny, but true, you can't run OSX on an emachine, so it almost doesn't matter unless you're indifferent to the OS. I can't help but think too that apple maintains a premium price for the look of high prices, what word am I looking for here. Like the price on RR.

[Thain Esh Kelch]

Originally posted by SalBaker:
I will bet a nickel that when Apple launches the 970 PowerMacs the high-end (screaming-fast machine) will be expensive, and we will all be saying (again), "Don't buy the high-end, it is too expensive. What was Apple thinking! Buy the mid or low-end 970, they're a better deal!"

Thats because most people dont get the idea.

The current bus has been starving the processors since we got the dual 800 Mhz machine!

IF the 970 gets a 900 Mhz bus, then you can certainly bet that the top-end machine will be worth the money compared to low-end!

[Scotttheking]

That translation is very poor.

What actually happened is IBM pulled the demo, and there's no word on why.
The processor is being released at 1.8, whenever it is release. It is not here now.

[Norty]

agasthya,
Where in the world did you hear about a 970 Powermac being demonstrated in May? A person at Apple told me that by the end of Sept. there's a good chance there'd be a new system, but that's all I've heard with any certainty.

And to the rest,
Who cares how much a screaming fast new 970 tower will cost? It'll be worth every penny. We Graphic Art folks need it!

[Centris650]

Originally posted by Norty:

And to the rest,
Who cares how much a screaming fast new 970 tower will cost? It'll be worth every penny. We Graphic Art folks need it!

True but all can't afford it.

[businezguy]

There are a lot of people holding out for these machines. If the processors are available (and it looks like they will be), Apple is going to get the machines out as quickly as possible.

It is looking somewhat possible that we will see a release sometime this year rather then early next year.

If and when this release happens, this will definately help Apple with market share, especially as these processors trickle down into the lower end machines. I know I have my credit card ready!

[Dace]

Originally posted by :XI::
Show me a company that isn't always looking for an excuse to charge more money. I thought the whole reason of being in business was to make a profit.

Well, there are 2 ways I can think of in order to make money...

1 Sell a few 'cause you charged high
2 Sell a lot 'cause you charged low

Look at AMD, they're selling more of their processors than Intel's high priced Pentium. Since Apple has a low market share, I think they would benefit by going with the second option.

I know in economy class you learn that if the demand is high, the price goes up...but I think that's only true if you're a monopoly....like Microsoft's Windows

[MindFad]

Originally posted by Norty:
We Graphic Art folks need it!

Word!

Still, I think they should keep the prices basically the same come 970 time. It will only do Apple good, IMO. [/armchair CEO]

[nickdman]

IBM can make good chips BUT remeber they must avoid going into the light of Cyrix.



For those youngins here Cyrix chips were the only chips that you had to under clock in order for them not to melt

[Athens]

Cyrix chips became the VIA C3 CPU and are the coolist running CPU's on the market and very low power. To bad there performance is pretty crappy too.

When the 970 hits the market for Macs im still going to be waiting a year for the price to come down AND I CANT WAIT TO GET ONE!

[CubeBoy]

Originally posted by Dace:
Well, there are 2 ways I can think of in order to make money...

1 Sell a few 'cause you charged high
2 Sell a lot 'cause you charged low

Look at AMD, they're selling more of their processors than Intel's high priced Pentium. Since Apple has a low market share, I think they would benefit by going with the second option.

I know in economy class you learn that if the demand is high, the price goes up...but I think that's only true if you're a monopoly....like Microsoft's Windows

Taken from pricewatch.com

Pentium 4 3.06: $552 | Athlon 3000+: $561
Pentium 4 2.80: $353 | Athlon 2800+: $364
Pentium 4 2.66: $233 | Athlon 2700+: $254
Pentium 4 2.53: $182 | Athlon 2600+: $224
Pentium 4 2.26: $169 | Athlon 2400+: $124

AMD been losing money and marketshare for the last one and a half years, the Athlon's just doesn't have that price/performance advantage anymore.

[DrBoar]

Apparenlty (SPEC200 scores) a 970 at 700 MHz is faster than a G4/1.42 so as long as Apple use CPUs faster than say 1 GHz the current limit of 1.8 or 2.5 really does not matter that much compared to the G4. If Apple would replace the G4 SP1 DP1.25 DP1.42 with 970 at the same clock speed it would boost the CPU performance by 2 to 4 times. That is a very big leap to actually increase the speed to SP1.4 DP 1.4 DP 1.8 or even higher just make the leap "very very big" or "very very very big". I am not claming that Apple can not make computers that will make us dissapointed, I am sure they have the competence and track record for that .

My point is is that compared to a G4/1.42 any resonable 970 is faster be it 1.4 or 1.8 or 2.5.

Then we have the *fact* "The current bus has been starving the processors since we got the dual 800 Mhz machine!".

I this it true dual 1.2 GHz upgrades in 100 MHz bus G4 AGP should be much slower than the current dual 1.25 with 167 MHz bus. The dual 1.42 should also be much less than 77% faster than a dual 800.

The notion that if only the G4 had a fast bus they would spank the pentiums seem to be appealing so some. Looking through the tests of xl8yourmac.com and barefeats.com does not give that notion any support

I am sure that a G4 can be starved, a 800 MHz G4 on a 50 MHz bus in a 7500 sure is starved, but I have yet to see any signs of starvation on the 100 Mhz bus let alone on the 133-167 MHz ones. As far as I can tell the main reason for the slowness of the G4 is that the G4 is a slow CPU

[Scotttheking]

Originally posted by Scotttheking:
That translation is poor.

What actually happened is IBM pulled the demo, and there's no word on why.
The processor is being released at 1.8, whenever it is release. It is not here now.

ok, big edit. Either I read something different from what was there, or the translation posted above is not the same as what was there.
Also, I saw that posted on ars by someone with a different name.


Sorry about that simon.

[Lateralus]

Originally posted by Thain Esh Kelch:
YOU obviously has absolutely NO idea of what you are talking about.

Apparently, neither do you. I would suggest that you investigate more before sharing your 'insights'.

Number one: The 2.5GHz 970's are tapping out very well, but are not yet in high volume production. The main reason is because there is no current demand for them...

Number two: They will go into high volume production when IBM begins using them in their own PPC 970 powered Blade servers.

[Eug]

Originally posted by nickdman:
IBM can make good chips BUT remeber they must avoid going into the light of Cyrix.

For those youngins here Cyrix chips were the only chips that you had to under clock in order for them not to melt

You're showing your age. For ultra low wattage x86 chips the Cyrix/VIA chips are now the chips of choice.

In fact you can even get a 533 MHz chip with a complete motherboard and ports, etc., that fits in the palm of your hand. It's fast enough to run Win XP, although it certainly won't be screamingly fast. It's definitely fast enough for Linux-based office apps though.

They make a point of demoing one of those chips HEATSINKLESS AND FANLESS, and it's able to play Quake III without crashing. If you tried this with a current AMD chip you'd probably completely fry it in under 30 seconds.

[Scotttheking]

Originally posted by DrBoar:
Then we have the *fact* "The current bus has been starving the processors since we got the dual 800 Mhz machine!".

I this it true dual 1.2 GHz upgrades in 100 MHz bus G4 AGP should be much slower than the current dual 1.25 with 167 MHz bus. The dual 1.42 should also be much less than 77% faster than a dual 800.

The notion that if only the G4 had a fast bus they would spank the pentiums seem to be appealing so some. Looking through the tests of xl8yourmac.com and barefeats.com does not give that notion any support

I am sure that a G4 can be starved, a 800 MHz G4 on a 50 MHz bus in a 7500 sure is starved, but I have yet to see any signs of starvation on the 100 Mhz bus let alone on the 133-167 MHz ones. As far as I can tell the main reason for the slowness of the G4 is that the G4 is a slow CPU

First, let me say that the G4 is starved for memory bandwidth. But don't just take my word for it. This is a quote from BadAndy on Ars, where he states the same thing

It would be a very potent Altivec machine (assuming IBM's altivec implementation either preserves current latencies, or has register-rename-on-load OOOE issue ... and I expect the latter from IBM; the other execution issue units do this of course), because the bandwidth is so good compared to existing G4+. Among other things this would mean that we'd start to see MP altivec algorithms ... which today we generally don't bother because I CPU can usually more than choke the FSB. (An exception being where going MP gets you access to the two L3 caches... and that's why you do it, for things like large cache-blocked matrix operations, etc.)

What I fear is that Apple will be a long time getting the machines out the door (in which case towers will be REALLY dead, because a lot of people are holding off waiting), or will start out SP only, or will have strangled memory bandwidth (even "full" PC2100 would be distinctly underwhelming), or worst of all will do all of these things together.

Emphasis mine.

[Eug]

Yep, on the PC side with a GHz-class CPU, going from a 100 MHz bus to a 133 MHz bus gives a significant boost in performance.

Also on the PC side (although I've only seen the comparisons with a somewhat faster CPU), going from a 167 MHz bus to a 333 MHz bus also give a significant boost in performance. This is obviously most noticeable in very memory intensive apps, including mainstream apps like the SimCity 4 game apparently.

[Athens]

Originally posted by CubeBoy:
Taken from pricewatch.com

Pentium 4 3.06: $552 | Athlon 3000+: $561
Pentium 4 2.80: $353 | Athlon 2800+: $364
Pentium 4 2.66: $233 | Athlon 2700+: $254
Pentium 4 2.53: $182 | Athlon 2600+: $224
Pentium 4 2.26: $169 | Athlon 2400+: $124

AMD been losing money and marketshare for the last one and a half years, the Athlon's just doesn't have that price/performance advantage anymore.

Um AMD has been gaining market share for over 2 years, there whole sale prices are no comparison for people like me who build PCs for a living. And performance wise Athlon 1800+ is as good or still better then a P4 2.53 depending on MB/Memory combinations used in both systems.

[Cipher13]

Don't hold your breath for these things.

[Eug]

Originally posted by Athens:
And performance wise Athlon 1800+ is as good or still better then a P4 2.53 depending on MB/Memory combinations used in both systems.

Of course an Athlon 1800+ would get smoked by a P4 2.53 in most apps, with the same amount of memory.

I'm not sure what you're trying to say.

[D'Espice]

Originally posted by Simon:
According to this article from CeBit the IBM PPC 970 is here and it is running at 1.8GHz. They expect to go to 2.3-2.5GHz with the next revision while staying at .13microns. IBM was quoted as saying Apple could use the chip.

I suppose this is pretty good news.

The article: http://www.macguardians.de/fullstory.php?p=1830

The interesting part in (original) German:



And here is my translation (it's probably far from perfect, but it's for sure better than the computer translation I got):

To our suprise we had to realize that the presentation of the PPC 970 which had been announced in advance has been canceled. Upon request we were told that the expected processor (2.3 to 2.5GHz) resembled an outlook onto the after next generation and its appearance shouldn't be expect soon. The current processor is the PPC 970 (still 0.13 micron fabrication, which will also be true for the coming 2.3/2.5 GHz processor) with up to 1.8 GHz, which according to IBM, could be used by Apple as well in the future. But as usual one doesn't know anything more precisely.

Simon is right, that translation is far from perfect, however I think he got the gist. The point is: The PPC970 is not going to be shown on Cebit and it'll not be available soon. The first available PPC970 will be not be faster than 1.8 GHz.

Scott is right by saying that the MPC74xx is memory bandwidth limited. Given that it cannot take full advantage (actually cannot take any advantage) of DDR-SDRAM due to the chipset, the best boost will be a new chipset and an MPC74xx supporting the DDR Frontsidebus.

PS: I'm going to Hannover this weekend and I'll try to gather as much information about that microprocessor as possible.

[CubeBoy]

Originally posted by Athens:
Um AMD has been gaining market share for over 2 years, there whole sale prices are no comparison for people like me who build PCs for a living. And performance wise Athlon 1800+ is as good or still better then a P4 2.53 depending on MB/Memory combinations used in both systems.

AMD lost 1,303,000,000 dollars so far. It's market share is the lowest it's been in 4 years (11%). Dual Channel DDR provides the best performance on both Pentium 4s and Athlons, Current P4 Single Channel DDR memory offers at least close performance to P4s equipped with PC1066 RDRAM.

Performance wise, a Athlon 1800+ performs like a P4 1.8 ghz, even though it's clocked significantly slower, thats the whole point of AMD's PR ratings although at times it's skewed. In actuality, a 2.53 ghz P4 performs more like a Athlon 2700+ on Content Creation, Most Games, Audio/Video, and Scientific/Engineering programs. It performs like a 2400+ on office programs.

[Scotttheking]

take the AMD discussion somewhere else, it has nothing to do with this thread.

[Athens]

Originally posted by Eug:
Of course an Athlon 1800+ would get smoked by a P4 2.53 in most apps, with the same amount of memory.

I'm not sure what you're trying to say.

You need reading lessions, Thats not what I said, I pitty people that cant read. I said the 1800+ is as good and in some ways still better then a P4 2.53GHz

Learn to READ!

[DrBoar]

From Scotttheking "First, let me say that the G4 is starved for memory bandwidth. But don't just take my word for it. This is a quote from BadAndy on Ars, where he states the same thing"

I am sorry, but action speakes louder than words. Show me, were actual tests show that the AV is starved! that is on an at least 100 Mhz bus???


Look at the dual 500 on either 66 or 100 MHz bus were a 50% boost of bus speed gets you marginal speed bumps
http://www.xlr8yourmac.com/G4CARDS/X..._G4_500MP.html


Is dual 500 to slow then to to the dual 1 Ghz test.
http://www.xlr8yourmac.com/G4CARDS/g...ghz/index.html
Going from a 133 SDR to a 167 DDR ( sort of). That is at least a 26% increase in bus performance with dual 1 GHz G4. Still most application test is within+/-5% of each other

QED

[Catfish_Man]

Originally posted by DrBoar:
From Scotttheking "First, let me say that the G4 is starved for memory bandwidth. But don't just take my word for it. This is a quote from BadAndy on Ars, where he states the same thing"

I am sorry, but action speakes louder than words. Show me, were actual tests show that the AV is starved! that is on an at least 100 Mhz bus???


Look at the dual 500 on either 66 or 100 MHz bus were a 50% boost of bus speed gets you marginal speed bumps
http://www.xlr8yourmac.com/G4CARDS/X..._G4_500MP.html


Is dual 500 to slow then to to the dual 1 Ghz test.
http://www.xlr8yourmac.com/G4CARDS/g...ghz/index.html
Going from a 133 SDR to a 167 DDR ( sort of). That is at least a 26% increase in bus performance with dual 1 GHz G4. Still most application test is within+/-5% of each other

QED

I'm not sure what you're saying with this: "Is dual 500 to slow then to to the dual 1 Ghz test". If you can clarify it, I might be able to reply to it properly. It should be relatively easy to code an Altivec program that doesn't fit in the cache, then test it to see whether it's bus limited (it should be. A 1GHz G4+ doing 1 VFMADD/cycle requires 384,000,000,000 bits/sec for optimal performance. (128 bits + 128 bits + 128 bits) * 1,000,000,000 cycles per second).

[Scotttheking]

Originally posted by DrBoar:
From Scotttheking "First, let me say that the G4 is starved for memory bandwidth. But don't just take my word for it. This is a quote from BadAndy on Ars, where he states the same thing"

I am sorry, but action speakes louder than words. Show me, were actual tests show that the AV is starved! that is on an at least 100 Mhz bus???


Look at the dual 500 on either 66 or 100 MHz bus were a 50% boost of bus speed gets you marginal speed bumps
http://www.xlr8yourmac.com/G4CARDS/X..._G4_500MP.html


Is dual 500 to slow then to to the dual 1 Ghz test.
http://www.xlr8yourmac.com/G4CARDS/g...ghz/index.html
Going from a 133 SDR to a 167 DDR ( sort of). That is at least a 26% increase in bus performance with dual 1 GHz G4. Still most application test is within+/-5% of each other

QED

I never said every app will get a benefit. Some will not, some will. Not every app is bandwidth limited.

Next, there are no scientific apps in that benchmark list. That is one area where memory bandwidth limits become very apparent.

Now, you can go on telling yourself that the G4 is a poor processor if it makes you feel better. You are wrong, but it is your choice to believe you are right.

Also, I submit, for your reading pleasure, more info then you ever wanted to know.
Future Apple CPU thread
Be sure to check out the linked "Want Bandwidth" thread in the first post, then read all 42 pages of that thread. Spread throughout it is quite a bit of information talking about memory bandwidth, and the lack of it in the G4.

Enjoy.

[DrBoar]

OK I try to phrase is more stringently
A. It appers that the G4 is lagging behind the P4 in many comon application, or do do not agree on this?


B I claim that this lagging is not due to bandwith limitations

I am sure that the G4 can get bandwith limitations on the currently fastest bus aviable while doing some fairly esoteric tasks.
http://www.xlr8yourmac.com/G4CARDS/g...hz/index2.html
Do use the SMP aware application AltiVec Fractial with about a 10% boost by the faster bus, so it does appear that not all AltiVec applications stres the bus

So my point is while the the G4 can be driven into bus saturation, bandwith limitations is not the reason for G4 lagging behind the P4 in many applications

"Is dual 500 to slow then to to the dual 1 Ghz test".
To be more verbose: If you think that the dual 500mHz G4 is too slow to saturate the bus look at the test of the twise as fast dual 1 GHz GHz to see if they saturate the bus.

(It now struck me that the dual 500 on a 66 MHz stresses the bus at least as much as the more deeply piplined 1GHz does a 133 MHz bus)