<strong>Ars Technica has a brief <a href="http://arstechnica.com/wankerdesk/3q02/powerpc.html"; target="_blank">look</a> at the 970 up. It?s pretty good, especially the realistic conclusion.</strong><hr></blockquote>
I think a more realistic conclusion could be made when the chip is actually available to the real world, but that's just me.
I think a more realistic conclusion could be made when the chip is actually available to the real world, but that's just me.</strong><hr></blockquote>
You would not think so from the way the 970 is being discussed here. From reading this thread it seems the 970 had already been proclaimed as the greatest CPU ever.
- SPEC tests for PPC 970 and P4 2.8 are roughly similar at this point
- SPEC is very biased towards intel. . . the x86 FPU is crap, and that's a fact. Do your own research.
- I highly doubt the SPEC for the 970 were 64bit programs. Are there any yet for PPC? Um, I don't think so.
- The compiler (gcc) for x86 is disgustingly more optimized than the one for PPC.
- I don't think VMX was measured in these tests.
and lastly:
- The SPEC benchmarks for the 970 are pure conjecture as far as I know.
Three likely things we'll see in a year:
- P4 gains some clock speed, loses some instr/cycle. Net effect: 20% speed increase.
- A Compiler is optimized for 970, PPC in general, along with VMX.
- Thusly, Benchmarks double if not triple for what we'd expect them to be for current 970. So let's be conservative and say 100% speed increase, just due to the compile and proper testing.
I think, when it debuts, the PPC970 will be able to go toe-to-toe with anything from AMD or Intel, if not beat them outright. This is IBM we're talking about here, not Moto. IBM draws a lot of water in the computer world. I think enough to get a good compiler out within a year.
Another interesting tidbit I noticed is that, unlike prior PPC's, the 970 has a microcode interpreter in its core. That doesn't mean anything in terms of speed, I wouldn't think, but I was suprised to see it. . . It implies to me that IBM has no plans to make this an embedded processor.
[quote]
You would not think so from the way the 970 is being discussed here. From reading this thread it seems the 970 had already been proclaimed as the greatest CPU ever.
<hr></blockquote>
A 1.8Ghz 970 cracked more RC5 keys than a Power4, which, if you ask me, is the greatest CPU ever.
[quote] You would not think so from the way the 970 is being discussed here. From reading this thread it seems the 970 had already been proclaimed as the greatest CPU ever. <hr></blockquote>
Perhaps you are not familiar with the sh!t-ridden whore of a company we deal with called MOTOROLA? After dealing with them for YEARS and watching our asses sink into irrelevance, I think a little irrational exuberance is called for.
Plus, all the signs say that the chip will at least be "very competitive" if not, "best chip in the w0rld!".
"Motorola is the problem for Apple, not the solution. The sooner Apple makes a complete committment to IBM the better. IBM will lead Apple to the next generation of power-pc cpu's, not Motorola.
Who is the leader in R & D in the computing industry? IBM
Who is the major backer of UNIX in the corporate setting?
IBM
UNIX underlies OSX.
If there is a G5, it won't be Motorola, it should be IBM."
I got that quote from somewhere.
Hmmm.
"Perhaps you are not familiar with the sh!t-ridden whore of a company we deal with called MOTOROLA? After dealing with them for YEARS and watching our asses sink into irrelevance, I think a little irrational exuberance is called for."
I wouldn't have put it quite like that...
Power 4 doesn't have altivec but the GPUL does.
That's what you call icing on the cake by the looks of things.
I think 2004 Mac Janworld is worst case scenario.
It sounds to me like Apple told Moto to take their G5 and shove it.
Meanwhile, Apple uses the same G4s for now...does dual...minor architecture improvements.
I hear they used 7555 revised to get to 1.25?
Apparently opting not to use the latest G4 revision.
I think the reasons for this are:
Apple are stalling? The next G4 has got to last quite a while.
We may see a new case in Jan 2003. Completely new. The .13 G4s. Up to 1.6 gig with...part of the other reason for the G4's 'delayed' improvement.
ie a proper next gen DDR architecture to compliment the 2nd Gen' Apollos running up to 1.6 gig. (Y'know, the one Moki hinted got 'delayed'...)
If the GPUL doesn't arrive until 2004 Jan'...we may get .13 G4s in Jan' and .9 G4s in New York next year. Or .13 G4 Apollo 2's in Jan' and dual version strategy in New York next year.
Either way...I've got a sneaky feeling Apple will drop the 'all dual' strat' for Jan' for tactical reasons if the GPUL can't ship or announce at New York 2003.
Just my thoughts.
It's what's going to happen in the mean time that will provoke much thought. I tend to concur with Kidred on this one...
<strong>[qb]MacOS X also has to be modified to be 64-bit compliant, but I expect that's already been done. That just leaves getting the hardward working reliably.</strong>
The PPC 970 can be run in 32 bit mode, so only a minimum of changes needs to made to get OS 10 to run it. Apple can then take their sweet time transitioning to 64 bit Mac OS.[/QB]<hr></blockquote>
Would it be a sane proposition tp say that if Apple have an Intel compatible build of OS X that is up to date with the commercial release, then they would probably also have a 64bit version maybe running on Power4 hardware from IBM or Proto-gpul boxes also keeping step. I'm guessing that 10.3 will ship with a 32bit and 64bit installer in the box.
This bus seems suspiciously like a 450 MHz 32 bit Hypertransport bus, albiet modified, and Apple joined the HT consortium in early 2H 01, so maybe the design work started then.
It also suspiciously looks like the GX bus already used on the POWER4 but will some modifications (like DDR).
You would not think so from the way the 970 is being discussed here. From reading this thread it seems the 970 had already been proclaimed as the greatest CPU ever.</strong><hr></blockquote>
Well I don't know about that but one thing is for sure... While IBM told us the story of the GPUL this Tuesday they forgot to put 'to be continued' at the end of their talk. <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" />
Oh, and by the comments being made by a few of the fellow AI members I'm not the only one who knows this.
What I want to know is how does the 970 stack up against the AppleInsider designed G5 we almost shipped a year ago? For some reason there seems to be less excitement about this. Damn you reality, damn you to hell! I just knew moki would ruin this place when he started talking sense. Nirvana always seems to be a year away. Well, I guess at least we now know what it looks like, and that we have a real future there. I can't help but wonder, however, what the hell we are going to talk about until then? <img src="confused.gif" border="0">
I think, when it debuts, the PPC970 will be able to go toe-to-toe with anything from AMD or Intel, if not beat them outright. This is IBM we're talking about here, not Moto. IBM draws a lot of water in the computer world. I think enough to get a good compiler out within a year.</strong>
Never forget what Intel is. They are the dominant microprocessor manufacturer in the market. They will get to 90 nm first. And I highly doubt that the PPC 970 will outright outperform Intel chips by the time it is released. The only negative Intel has right now, which quickly can be mitigated, is that IA-32 and IA-64 are totally different and incompatible architectures. Someone messed up at Intel when they designed the IA-64 ISA. Otherwise, they have all the advantages: fab, marketing, and the chips to get them where they need to go.
The odd thing about the PPC 970 and the Power4 is that their clock rates are not that great for 14-stage pipeline processors. One would hope IBM will be able to squeeze 200 to 500 more MHz out of them. If their was just a little more MHz, then there is reason to be really optimistic, but as it stand now, it'll make Apple's marketing job just a little bit easier.
It also suspiciously looks like the GX bus already used on the POWER4 but will some modifications (like DDR).</strong>
I think you're more correct than me. They did have to put in quite a few features into the GX bus to support processor bus type features though.[/QB]<hr></blockquote>
This could also be a hybrid HT/GX bus. How about the idea that IBM is making the controller chip (memory controller, PCI controller, ethernet) and Apple will incorporate it into its motherboard? Apple has done this before when a radically new processor came out such as the G3. Apple used the Motorola made MPC106 memory/PCI controller in the beige G3 as well as the B&W. I could safely assume that IBM has a controller project going on in tandem with the 970 project to be released together. This would help Apple out a lot and it would let them release machines as soon as the processor is available in quantity for them. And while they sell these machines they can work on their own proprietary controllers in house in the mean time.
I wish I could but I can't... but here's a hint. Someone who posted on this very (edit) err last page seems to be on to something... He didn't write a ton of words but what he did say could indeed come true.
Comments
<strong>the CPU and controller communicate with 2 32bit buses; thats at least 120 pins for one CPU.</strong><hr></blockquote>
2x32=120? Is it using differential signaling?
[ 10-17-2002: Message edited by: FotNS ]</p>
<strong>Ars Technica has a brief <a href="http://arstechnica.com/wankerdesk/3q02/powerpc.html"; target="_blank">look</a> at the 970 up. It?s pretty good, especially the realistic conclusion.</strong><hr></blockquote>
I think a more realistic conclusion could be made when the chip is actually available to the real world, but that's just me.
<strong>
I think a more realistic conclusion could be made when the chip is actually available to the real world, but that's just me.</strong><hr></blockquote>
You would not think so from the way the 970 is being discussed here. From reading this thread it seems the 970 had already been proclaimed as the greatest CPU ever.
<strong>Ars Technica has a brief <a href="http://arstechnica.com/wankerdesk/3q02/powerpc.html"; target="_blank">look</a> at the 970 up. It?s pretty good, especially the realistic conclusion.
[ 10-17-2002: Message edited by: FotNS ]</strong><hr></blockquote>
I disagree with Hannibal on the conclusion.
Things we know:
- SPEC tests for PPC 970 and P4 2.8 are roughly similar at this point
- SPEC is very biased towards intel. . . the x86 FPU is crap, and that's a fact. Do your own research.
- I highly doubt the SPEC for the 970 were 64bit programs. Are there any yet for PPC? Um, I don't think so.
- The compiler (gcc) for x86 is disgustingly more optimized than the one for PPC.
- I don't think VMX was measured in these tests.
and lastly:
- The SPEC benchmarks for the 970 are pure conjecture as far as I know.
Three likely things we'll see in a year:
- P4 gains some clock speed, loses some instr/cycle. Net effect: 20% speed increase.
- A Compiler is optimized for 970, PPC in general, along with VMX.
- Thusly, Benchmarks double if not triple for what we'd expect them to be for current 970. So let's be conservative and say 100% speed increase, just due to the compile and proper testing.
I think, when it debuts, the PPC970 will be able to go toe-to-toe with anything from AMD or Intel, if not beat them outright. This is IBM we're talking about here, not Moto. IBM draws a lot of water in the computer world. I think enough to get a good compiler out within a year.
Another interesting tidbit I noticed is that, unlike prior PPC's, the 970 has a microcode interpreter in its core. That doesn't mean anything in terms of speed, I wouldn't think, but I was suprised to see it. . . It implies to me that IBM has no plans to make this an embedded processor.
[quote]
You would not think so from the way the 970 is being discussed here. From reading this thread it seems the 970 had already been proclaimed as the greatest CPU ever.
<hr></blockquote>
A 1.8Ghz 970 cracked more RC5 keys than a Power4, which, if you ask me, is the greatest CPU ever.
[ 10-17-2002: Message edited by: Splinemodel ]</p>
Perhaps you are not familiar with the sh!t-ridden whore of a company we deal with called MOTOROLA? After dealing with them for YEARS and watching our asses sink into irrelevance, I think a little irrational exuberance is called for.
Plus, all the signs say that the chip will at least be "very competitive" if not, "best chip in the w0rld!".
That is much better than where we sit today.
<strong>
A 1.8Ghz 970 cracked more RC5 keys than a Power4, which, if you ask me, is the greatest CPU ever.</strong><hr></blockquote>
You forget RC5 is heavily altivec optimised and the POWER4 doesn't have altivec, or VRF.
Who is the leader in R & D in the computing industry? IBM
Who is the major backer of UNIX in the corporate setting?
IBM
UNIX underlies OSX.
If there is a G5, it won't be Motorola, it should be IBM."
I got that quote from somewhere.
Hmmm.
"Perhaps you are not familiar with the sh!t-ridden whore of a company we deal with called MOTOROLA? After dealing with them for YEARS and watching our asses sink into irrelevance, I think a little irrational exuberance is called for."
I wouldn't have put it quite like that...
Power 4 doesn't have altivec but the GPUL does.
That's what you call icing on the cake by the looks of things.
I think 2004 Mac Janworld is worst case scenario.
It sounds to me like Apple told Moto to take their G5 and shove it.
Meanwhile, Apple uses the same G4s for now...does dual...minor architecture improvements.
I hear they used 7555 revised to get to 1.25?
Apparently opting not to use the latest G4 revision.
I think the reasons for this are:
Apple are stalling? The next G4 has got to last quite a while.
We may see a new case in Jan 2003. Completely new. The .13 G4s. Up to 1.6 gig with...part of the other reason for the G4's 'delayed' improvement.
ie a proper next gen DDR architecture to compliment the 2nd Gen' Apollos running up to 1.6 gig. (Y'know, the one Moki hinted got 'delayed'...)
If the GPUL doesn't arrive until 2004 Jan'...we may get .13 G4s in Jan' and .9 G4s in New York next year. Or .13 G4 Apollo 2's in Jan' and dual version strategy in New York next year.
Either way...I've got a sneaky feeling Apple will drop the 'all dual' strat' for Jan' for tactical reasons if the GPUL can't ship or announce at New York 2003.
Just my thoughts.
It's what's going to happen in the mean time that will provoke much thought. I tend to concur with Kidred on this one...
Lemon Bon Bon
[ 10-17-2002: Message edited by: Lemon Bon Bon ]</p>
<strong>[qb]MacOS X also has to be modified to be 64-bit compliant, but I expect that's already been done. That just leaves getting the hardward working reliably.</strong>
The PPC 970 can be run in 32 bit mode, so only a minimum of changes needs to made to get OS 10 to run it. Apple can then take their sweet time transitioning to 64 bit Mac OS.[/QB]<hr></blockquote>
Would it be a sane proposition tp say that if Apple have an Intel compatible build of OS X that is up to date with the commercial release, then they would probably also have a 64bit version maybe running on Power4 hardware from IBM or Proto-gpul boxes also keeping step. I'm guessing that 10.3 will ship with a 32bit and 64bit installer in the box.
*and why are we all suddenly unregistered!?*
[ 10-17-2002: Message edited by: robster ]</p>
This bus seems suspiciously like a 450 MHz 32 bit Hypertransport bus, albiet modified, and Apple joined the HT consortium in early 2H 01, so maybe the design work started then.
It also suspiciously looks like the GX bus already used on the POWER4 but will some modifications (like DDR).
<strong>
You would not think so from the way the 970 is being discussed here. From reading this thread it seems the 970 had already been proclaimed as the greatest CPU ever.</strong><hr></blockquote>
Well I don't know about that but one thing is for sure... While IBM told us the story of the GPUL this Tuesday they forgot to put 'to be continued' at the end of their talk. <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" />
Oh, and by the comments being made by a few of the fellow AI members I'm not the only one who knows this.
Dave
[ 10-17-2002: Message edited by: DaveGee ]</p>
It also suspiciously looks like the GX bus already used on the POWER4 but will some modifications (like DDR).</strong>
I think you're more correct than me. They did have to put in quite a few features into the GX bus to support processor bus type features though.
<strong>I can't help but wonder, however, what the hell we are going to talk about until then?
Digital Devices
I think, when it debuts, the PPC970 will be able to go toe-to-toe with anything from AMD or Intel, if not beat them outright. This is IBM we're talking about here, not Moto. IBM draws a lot of water in the computer world. I think enough to get a good compiler out within a year.</strong>
Never forget what Intel is. They are the dominant microprocessor manufacturer in the market. They will get to 90 nm first. And I highly doubt that the PPC 970 will outright outperform Intel chips by the time it is released. The only negative Intel has right now, which quickly can be mitigated, is that IA-32 and IA-64 are totally different and incompatible architectures. Someone messed up at Intel when they designed the IA-64 ISA. Otherwise, they have all the advantages: fab, marketing, and the chips to get them where they need to go.
The odd thing about the PPC 970 and the Power4 is that their clock rates are not that great for 14-stage pipeline processors. One would hope IBM will be able to squeeze 200 to 500 more MHz out of them. If their was just a little more MHz, then there is reason to be really optimistic, but as it stand now, it'll make Apple's marketing job just a little bit easier.
<strong>[qb]Originally posted by Outsider:
It also suspiciously looks like the GX bus already used on the POWER4 but will some modifications (like DDR).</strong>
I think you're more correct than me. They did have to put in quite a few features into the GX bus to support processor bus type features though.[/QB]<hr></blockquote>
This could also be a hybrid HT/GX bus. How about the idea that IBM is making the controller chip (memory controller, PCI controller, ethernet) and Apple will incorporate it into its motherboard? Apple has done this before when a radically new processor came out such as the G3. Apple used the Motorola made MPC106 memory/PCI controller in the beige G3 as well as the B&W. I could safely assume that IBM has a controller project going on in tandem with the 970 project to be released together. This would help Apple out a lot and it would let them release machines as soon as the processor is available in quantity for them. And while they sell these machines they can work on their own proprietary controllers in house in the mean time.
<strong>DaveGee, pray tell, pray tell....</strong><hr></blockquote>
I wish I could but I can't... but here's a hint. Someone who posted on this very (edit) err last page seems to be on to something... He didn't write a ton of words but what he did say could indeed come true.
edit.. damn I'm on the next page...
Dave
[ 10-17-2002: Message edited by: DaveGee ]</p>