I agree completely. I'm not letting myself get set up for disappointment yet again. What I've got is OK for at least another year, so I'm not desperate personally. For Apple's sake, I hope PPC970-based machines come out sooner than 2H 2003, though. I do think there's a lot more going on behind the scenes than the "big boys" are letting on. Moki and others have been hinting that something big is afoot; I'm looking forward to seeing the other shoe drop WRT the rest of the 970 family - while tempering my hopes with a stiff dose of reality.</strong><hr></blockquote>
Again, we will get the 970 in the fall of next year. In Jan/Feb we will get a new G4 with new mobo possibly the one mentioned in this thread.
<strong> <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" /> lets talk a little bit about the new IBM 970 chip and the mythical motorola G5.
I have bad news for all of you who were/are so excited by the announced IBM chip.
first off all it isnt even coming out until late 2003 at the earliest.
its initial clock speed will be 1.8 ghz.
its specint rating is around 1000.
</strong><hr></blockquote>
The benchmarks for the 970 were estimated and in consideration that no compiler has been even close to optimized for the 970 at this point. SpecInt and SpecFP aren't pure measures of the processor: They also measure the compiler. Since there is basically a 100% speed increase just by moving from gcc 2.95 to 3.2, I would assume that the realistic SPEC numbers for the 970 are closer to 2000.
<strong>If there is any consolation, the 512KB L2 cache and up to 4MB L3 should improve performance despite the pretty weak 167MHz system bus.
If you look at the die for the 7457 you can see how they reworked the internals. the rearranges stuff on die to fit the bigger L2, improved L3 circuitry... possibly they modified the bus?
Also i think we can safely put away the rumors of any 7460, 7470, or maybe 7500 series coming out of Motorola. They named the successor to the 7457 the 7457-RM. And after considering carefully looks like the 7 part stands for the process name (HIP7). But it would be the first to use than nomenclature.</strong><hr></blockquote>
The 7457 cannot use 4MB of L3, only 2MB. It can use 4MB of private memory on the L3 bus, but that would only be of use to embedded applications.
The benchmarks for the 970 were estimated and in consideration that no compiler has been even close to optimized for the 970 at this point. SpecInt and SpecFP aren't pure measures of the processor: They also measure the compiler. Since there is basically a 100% speed increase just by moving from gcc 2.95 to 3.2, I would assume that the realistic SPEC numbers for the 970 are closer to 2000.</strong><hr></blockquote>
IBM have a very nice set of compilers (xlc/xlf) which are able to optimise for the POWER4, and I'm sure they have been altered to optimise for the 970 (it wouldn't take that much work), this will be what they used for assessing the SPEC scores.
The change from gcc 2.95 to 3.2 for SPEC scores is not very great, maybe 10-20% on a good day, and gcc cannot even compile all the SPECfp subprograms, as it cannot do FORTRAN90.
Again, we will get the 970 in the fall of next year. In Jan/Feb we will get a new G4 with new mobo possibly the one mentioned in this thread.</strong><hr></blockquote>
KidRed,
What's your take on the 75XX series of processors from Motorola? Is the 75xx the G4++? Somebody in another thread posted some links to pdf files from Mot. and they reference the 75XX part in errata reports regarding altivec.
What's your take on the 75XX series of processors from Motorola? Is the 75xx the G4++? Somebody in another thread posted some links to pdf files from Mot. and they reference the 75XX part in errata reports regarding altivec.</strong><hr></blockquote>
Well, hehe, I don't know jack about chips and most techy stuff. I only know what I've been privied to and that is a new G4 with new mobo, FSB that is wicked hot. That may be the reason bhind the vents on the new tower. I mentioned this somewhere else and someone replied that it must be a G4+(+) because of the new mobo. So I have no idea what family it's in but if this is Moto's next chip, I assume it's the one slated for Jan/Feb.
Who honestly cares? Why buy anything until the 970 comes out? From what I hear, it will be a beast. My dual gig should do me for another year.
<strong>Basically, in a more coherent/shorter version, if IBM's long term self-interest disappears, what is there to keep them interested in Apple's needs?
What was there to keep Moto interested? It's worth thinking about since IBM has a history of becoming periodically dis-interested.</strong><hr></blockquote>
Simple: IBM also has an interest in making a good Linux server chip. IBM also is comitted to making fast chips (they aren't planning on moving en masse to the embedded market). IBM has fabs for high end chips and will continue to have fabs for higherend chips in the future.
IBM became disinterested when Apple decided to go with altivec and IBM thought that a more straightforward PPC chip was the right way to go. Now that IBM has realized that vector processing is useful (dolphin chip for Nintendo), they are willing to make a chip that can be used in a high end Mac.
Besides, if the only other choice is Moto, then ANY other choice is better.
<strong>Yeah my dual gig works well, but, a 1.4 Mhz dual w/DDR may entice me and get me through the debugging phase of the 970</strong><hr></blockquote>
wow, a whopping 1.4 mhz! damn, whered you dig up that kind of speed?
It's the same system rumours said we were going to get two-years ago. Good thing I didn't wait for it to come out. <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" />
Are the dates given for the release of the new MOT and IBM chips - Q1, Q2, etc. - referring to the calender year of 2003 or the fiscal year? The corporate fiscal year begins in October, if memory serves.
I said this elsewhere but that SNDF2002RECAP_Q1228.pdf that *WAS* on MOTs site seems to have 'gone missing'. Dare I say *confirmed* </strong><hr></blockquote>
And the pdf files that had reference to the 75xx family of parts have also vanished.
<strong>IBM became disinterested when Apple decided to go with altivec and IBM thought that a more straightforward PPC chip was the right way to go. Now that IBM has realized that vector processing is useful (dolphin chip for Nintendo), they are willing to make a chip that can be used in a high end Mac.</strong><hr></blockquote>
I'd venture to say that the Gekko experience is not what motivated IBM to adopt SIMD... it was something of a hack and only passingly useful. I suspect instead that Apple took some pains to demonstrate their success with the G4, and the results spoke for themselves. The leading reason that IBM was down on SIMD was that they figured nobody would use it and thus making generic code faster would be much more effective. Fortunately AltiVec is easy enough to use (although not trivial), and Apple has pushed it enough, so that there is a clear win to having it.
I'd venture to say that the Gekko experience is not what motivated IBM to adopt SIMD... it was something of a hack and only passingly useful. I suspect instead that Apple took some pains to demonstrate their success with the G4, and the results spoke for themselves. The leading reason that IBM was down on SIMD was that they figured nobody would use it and thus making generic code faster would be much more effective. Fortunately AltiVec is easy enough to use (although not trivial), and Apple has pushed it enough, so that there is a clear win to having it.</strong><hr></blockquote>
It also helps that Linux for PPC is getting more and more optimized for Altivec.
<strong>Are the dates given for the release of the new MOT and IBM chips - Q1, Q2, etc. - referring to the calender year of 2003 or the fiscal year? The corporate fiscal year begins in October, if memory serves.</strong><hr></blockquote>
From the MacNN article ... the obvious reason why Moto don't feature too heavily in the Apple roadmap. AltiVec my arse, nothing's going to make this chip fast enough ...
"The next processor Apple will likely use will be the Motorola MPC7457 and because it moves to a .13micron process architecturally the chip could see 1.8Hz at the top end."
Comments
<strong>
I agree completely. I'm not letting myself get set up for disappointment yet again. What I've got is OK for at least another year, so I'm not desperate personally. For Apple's sake, I hope PPC970-based machines come out sooner than 2H 2003, though. I do think there's a lot more going on behind the scenes than the "big boys" are letting on. Moki and others have been hinting that something big is afoot; I'm looking forward to seeing the other shoe drop WRT the rest of the 970 family - while tempering my hopes with a stiff dose of reality.</strong><hr></blockquote>
Again, we will get the 970 in the fall of next year. In Jan/Feb we will get a new G4 with new mobo possibly the one mentioned in this thread.
<strong> <img src="graemlins/smokin.gif" border="0" alt="[Chilling]" /> lets talk a little bit about the new IBM 970 chip and the mythical motorola G5.
I have bad news for all of you who were/are so excited by the announced IBM chip.
first off all it isnt even coming out until late 2003 at the earliest.
its initial clock speed will be 1.8 ghz.
its specint rating is around 1000.
</strong><hr></blockquote>
The benchmarks for the 970 were estimated and in consideration that no compiler has been even close to optimized for the 970 at this point. SpecInt and SpecFP aren't pure measures of the processor: They also measure the compiler. Since there is basically a 100% speed increase just by moving from gcc 2.95 to 3.2, I would assume that the realistic SPEC numbers for the 970 are closer to 2000.
<strong>If there is any consolation, the 512KB L2 cache and up to 4MB L3 should improve performance despite the pretty weak 167MHz system bus.
If you look at the die for the 7457 you can see how they reworked the internals. the rearranges stuff on die to fit the bigger L2, improved L3 circuitry... possibly they modified the bus?
Also i think we can safely put away the rumors of any 7460, 7470, or maybe 7500 series coming out of Motorola. They named the successor to the 7457 the 7457-RM. And after considering carefully looks like the 7 part stands for the process name (HIP7). But it would be the first to use than nomenclature.</strong><hr></blockquote>
The 7457 cannot use 4MB of L3, only 2MB. It can use 4MB of private memory on the L3 bus, but that would only be of use to embedded applications.
michael
<strong>
The benchmarks for the 970 were estimated and in consideration that no compiler has been even close to optimized for the 970 at this point. SpecInt and SpecFP aren't pure measures of the processor: They also measure the compiler. Since there is basically a 100% speed increase just by moving from gcc 2.95 to 3.2, I would assume that the realistic SPEC numbers for the 970 are closer to 2000.</strong><hr></blockquote>
IBM have a very nice set of compilers (xlc/xlf) which are able to optimise for the POWER4, and I'm sure they have been altered to optimise for the 970 (it wouldn't take that much work), this will be what they used for assessing the SPEC scores.
The change from gcc 2.95 to 3.2 for SPEC scores is not very great, maybe 10-20% on a good day, and gcc cannot even compile all the SPECfp subprograms, as it cannot do FORTRAN90.
Forget any possibility of doubling these scores.
[edit for spelling]
michael
[ 10-23-2002: Message edited by: mmicist ]</p>
<strong>
Again, we will get the 970 in the fall of next year. In Jan/Feb we will get a new G4 with new mobo possibly the one mentioned in this thread.</strong><hr></blockquote>
KidRed,
What's your take on the 75XX series of processors from Motorola? Is the 75xx the G4++? Somebody in another thread posted some links to pdf files from Mot. and they reference the 75XX part in errata reports regarding altivec.
<strong>
KidRed,
What's your take on the 75XX series of processors from Motorola? Is the 75xx the G4++? Somebody in another thread posted some links to pdf files from Mot. and they reference the 75XX part in errata reports regarding altivec.</strong><hr></blockquote>
Well, hehe, I don't know jack about chips and most techy stuff. I only know what I've been privied to and that is a new G4 with new mobo, FSB that is wicked hot. That may be the reason bhind the vents on the new tower. I mentioned this somewhere else and someone replied that it must be a G4+(+) because of the new mobo. So I have no idea what family it's in but if this is Moto's next chip, I assume it's the one slated for Jan/Feb.
Who honestly cares? Why buy anything until the 970 comes out? From what I hear, it will be a beast. My dual gig should do me for another year.
<strong>Basically, in a more coherent/shorter version, if IBM's long term self-interest disappears, what is there to keep them interested in Apple's needs?
What was there to keep Moto interested? It's worth thinking about since IBM has a history of becoming periodically dis-interested.</strong><hr></blockquote>
Simple: IBM also has an interest in making a good Linux server chip. IBM also is comitted to making fast chips (they aren't planning on moving en masse to the embedded market). IBM has fabs for high end chips and will continue to have fabs for higherend chips in the future.
IBM became disinterested when Apple decided to go with altivec and IBM thought that a more straightforward PPC chip was the right way to go. Now that IBM has realized that vector processing is useful (dolphin chip for Nintendo), they are willing to make a chip that can be used in a high end Mac.
Besides, if the only other choice is Moto, then ANY other choice is better.
<strong>Yeah my dual gig works well, but, a 1.4 Mhz dual w/DDR may entice me and get me through the debugging phase of the 970</strong><hr></blockquote>
wow, a whopping 1.4 mhz! damn, whered you dig up that kind of speed?
Check this out:
<a href="http://www.architosh.com/news/2002-10/2002c-1023-mcp7457-rm1.phtml" target="_blank">http://www.architosh.com/news/2002-10/2002c-1023-mcp7457-rm1.phtml</a>
--
Ed
<strong>MPC 7457 and 970 info
Check this out:
<a href="http://www.architosh.com/news/2002-10/2002c-1023-mcp7457-rm1.phtml" target="_blank">http://www.architosh.com/news/2002-10/2002c-1023-mcp7457-rm1.phtml</a>
--
Ed</strong><hr></blockquote>
I said this elsewhere but that SNDF2002RECAP_Q1228.pdf that *WAS* on MOTs site seems to have 'gone missing'. Dare I say *confirmed*
<strong>
I said this elsewhere but that SNDF2002RECAP_Q1228.pdf that *WAS* on MOTs site seems to have 'gone missing'. Dare I say *confirmed* </strong><hr></blockquote>
And the pdf files that had reference to the 75xx family of parts have also vanished.
<strong>IBM became disinterested when Apple decided to go with altivec and IBM thought that a more straightforward PPC chip was the right way to go. Now that IBM has realized that vector processing is useful (dolphin chip for Nintendo), they are willing to make a chip that can be used in a high end Mac.</strong><hr></blockquote>
I'd venture to say that the Gekko experience is not what motivated IBM to adopt SIMD... it was something of a hack and only passingly useful. I suspect instead that Apple took some pains to demonstrate their success with the G4, and the results spoke for themselves. The leading reason that IBM was down on SIMD was that they figured nobody would use it and thus making generic code faster would be much more effective. Fortunately AltiVec is easy enough to use (although not trivial), and Apple has pushed it enough, so that there is a clear win to having it.
<strong>
I'd venture to say that the Gekko experience is not what motivated IBM to adopt SIMD... it was something of a hack and only passingly useful. I suspect instead that Apple took some pains to demonstrate their success with the G4, and the results spoke for themselves. The leading reason that IBM was down on SIMD was that they figured nobody would use it and thus making generic code faster would be much more effective. Fortunately AltiVec is easy enough to use (although not trivial), and Apple has pushed it enough, so that there is a clear win to having it.</strong><hr></blockquote>
It also helps that Linux for PPC is getting more and more optimized for Altivec.
<strong>Are the dates given for the release of the new MOT and IBM chips - Q1, Q2, etc. - referring to the calender year of 2003 or the fiscal year? The corporate fiscal year begins in October, if memory serves.</strong><hr></blockquote>
Winter-Jan/Feb-G4, Fall-Aug/Sept-970.
"The next processor Apple will likely use will be the Motorola MPC7457 and because it moves to a .13micron process architecturally the chip could see 1.8Hz at the top end."
<strong>
The 7457 cannot use 4MB of L3, only 2MB. It can use 4MB of private memory on the L3 bus, but that would only be of use to embedded applications.
michael</strong><hr></blockquote>
Ah, I misunderstood. Still double the L2 should alleviate some performance issues, or at least keep them at bay.
Heh. Moto' can keep their 1.8 gig single data ram wonder.
Lemon Bon Bon