Kickaha and Amorph couldn't moderate themselves out of a paper bag. Abdicate responsibility and succumb to idiocy. Two years of letting a member make personal attacks against others, then stepping aside when someone won't put up with it. Not only that but go ahead and shut down my posting priviledges but not the one making the attacks. Not even the common decency to abide by their warning (afer three days of absorbing personal attacks with no mods in sight), just shut my posting down and then say it might happen later if a certian line is crossed. Bullshit flag is flying, I won't abide by lying and coddling of liars who go off-site, create accounts differing in a single letter from my handle with the express purpose to decieve and then claim here that I did it. Everyone be warned, kim kap sol is a lying, deceitful poster.
Now I guess they should have banned me rather than just shut off posting priviledges, because kickaha and Amorph definitely aren't going to like being called to task when they thought they had it all ignored *cough* *cough* I mean under control. Just a couple o' tools.
Don't worry, as soon as my work resetting my posts is done I'll disappear forever.
This PDF witch seems to be a presentation of some Holger Holthoff at IBM pSeries Technical Support held May 19th 2003:
Some interessting tibits:
Codenames:
GP = Power4 (1.3 GHz)
GQ = Power4+ (1.7 GHz)
GR = Power5 (1.8+ GHz)
GS = Power5+ (2.5+ GHz)
GPUL = 970 (1.8 GHz)
GRUL = 97x (2.5+ GHz)
So, there we have it. The Power5 derivative is GRUL, but it leaves GPUL2 which might well be 970+.
Excellent! That pdf also says on p2 (top) that the GR-UL (aka Power 5 mini-me) is currently designated "ppc97x". Looks like no 980 for a long time then...
With the Power4 using thicker circuits for reliability I imagine the Power5 will do the same. Therefore, we should be able to expect the GR-UL to start in the 2.5+ GHz range as most have been guessing. This document seems to confirm this likelyhood.
This PDF witch seems to be a presentation of some Holger Holthoff at IBM pSeries Technical Support held May 19th 2003:
Some interessting tibits:
Codenames:
GP = Power4 (1.3 GHz)
GQ = Power4+ (1.7 GHz)
GR = Power5 (1.8+ GHz)
GS = Power5+ (2.5+ GHz)
GPUL = 970 (1.8 GHz)
GRUL = 97x (2.5+ GHz)
So, there we have it. The Power5 derivative is GRUL, but it leaves GPUL2 which might well be 970+.
No, that would have to be the GQUL if you're going to hold to the naming convention. No such thing as a GPUL2.
Look, you can try to second guess TGB and you'd have a good chance of succeeding. But there are those who say that TGB or NMR are just a vehicle some folk use to post rumors from sensitive sources that could be compromised if the info were to appear on their regular rumor site. A way of distancing reports from sources known to be at least moderately close to the publishers of said website.
And there are others who say it's just a dumping ground for info from less than reliable sources. Except TGB's track record is too good for that.
No, that would have to be the GQUL if you're going to hold to the naming convention. No such thing as a GPUL2.
[snip]
In any event, I'm going with TGB's assessment.
I agree that there is an inconsitency in the pdf file. On one page it lists the Power4 on the CMOS 9S2 process as the "GR" (p1 at bottom) but on the next page (p2, top) it lists the "GR" as being Power5 along with the "GS". I haven't heard of IBM working on a low-k (which the "GR" is - p2, top) version of the Power4 so I'd bet that the error is on the first page. I've only heard of IBM working on the Power5 to suceed the current Power4 ("GQ").
Note that in the top slide on p2 it also mentions the GR-UL (the 97x). This says that the naming convention given by TGB of "P5UL" is inconsistent with IBM's current nomenclature and hence a little suspect.
I could be splitting hairs, clutching at straws, reading too much into the pdf, etc, but I think my statements above are reasonable.
No, that would have to be the GQUL if you're going to hold to the naming convention. No such thing as a GPUL2.
Not nessessarily. The 970+ is a second incarnation of the 970, aka GPUL, i's not a derivative of Power4+ aka GQ. Calling it GQUL would be wrong since GPUL already brached off of the Gx-tree.
And.. If this PDF is correct (and why shoudn't it? Even if there are some minor inconsistences in it) it's more likely that this is correct than NMR's artichle. It's is a document that comes directly from IBM after all, which NMR doesn't. It's probably composed from several existing presentations, made at different points in time. There might have been a "GR" Power4 before they renamed it Power5. Who knows? I don't.
The great finding of this document, as I see it, is the name "97x" for GRUL. We've always thought of it as 980 but that was a designation made up by someone (The Reg?) just because we didn't have anything better to call it. No we have, and it isn't called 980, it's called 97x.
Here's something to chew on: The Smell of Fear There's so much confusion about the 980, I don't know what to believe. Is he right about the 980 being 4x faster per clock and shipping by the end of next year? Lots of other interesting stuff in the AppleLust editorial.
Here's something to chew on: The Smell of Fear There's so much confusion about the 980, I don't know what to believe. Is he right about the 980 being 4x faster per clock and shipping by the end of next year? Lots of other interesting stuff in the AppleLust editorial.
Here's something to chew on: The Smell of Fear There's so much confusion about the 980, I don't know what to believe. Is he right about the 980 being 4x faster per clock and shipping by the end of next year? Lots of other interesting stuff in the AppleLust editorial.
Well the Power 5 is supposed to be 4x faster per clock thatn the Power 4. So perhaps he is extrapolating that the 980 (if it is derived form the Power 5) will be 4x faster per clock than the 970 (power 4 derived).
Here's something to chew on: The Smell of Fear There's so much confusion about the 980, I don't know what to believe. Is he right about the 980 being 4x faster per clock and shipping by the end of next year? Lots of other interesting stuff in the AppleLust editorial.
I read and enjoyed the article several days ago. However, we have yet to get G5's into our hands. The notion, however, that there is confusion about the 980 is just silly because the processor has not been announced.
While there are alot of truths and reasonable clarification in that TSoF editorial, there are also plenty of lies and exaggerations. Don't listen to all he's saying...
And there's no way the Power5 lite is 4 times as the 970 per clock in application performance, per core. No way! It might be 4x faster at certain operations (FastPath is a technology that is said to do so at certain operations, but not all, by far.) per clock, but overall, for regular code: Forget it.
The Power 5 is supposed to be 4 times faster than the original Power 4, not per clock.
Carson does well when he sticks to the point, but there are some real whoppers in that article. The Opteron does not "emulate" a 64 bit processor. It is a 64 bit processor. The "PPC 980" is an invention of rumors sites and tabloids that is not mentioned in any IBM document that anyone has been able to find. All IBM will say is that there is indeed a roadmap, and they are indeed making prototypes of the "next generation" chip, whatever that is. He also forgets to mention that the Itanium 2 addressed many of the performance problems that dogged the original Itanium - its albatross is the paucity of native applications.
Finally, he is himself too focused on platform bragging rights to acknowledge all the technical discussions on the general silliness of cross-platform benchmarking - there's some debate over whether it's even possible to compare the sort of things VeriTest was asked to compare in any meaningful way.
Thanks, I knew the good folks here would have some good insights to share. This helps me keep on the straight and narrow.
The important thing is that Apple has a very bright future ahead and that's all that really matters. I think there will be much gnashing of teeth in the PC world in the months and years ahead as Apple muscles in on some PC turf.
My pet theory is that the Power5 is actually a dual core 980. This would make it possible to ship the 980 before the POWER5.
That'd indeed be very nice. And it'd be very cost-effective for both Apple and IBM, since they'd only have to develop a single version of the chip, and optimize that. It all depends on how high-end and costly IBM plan it to be (low-end at IBM is often as high as, or higher than Apples own high-end definition) , but they've indicated that it will be much more scalable, from low to high end.
I don't find it unlikely that the elusive 980 will be a single-core version of the Power 5
According to a particularly juicy (if characteristically succinct) post by M. Isobe over on Ars, it just might have VMX on board. He dug up a reference to VMX from an IBM presentation apparently on the POWER5.
Disclaimer: The presentation's a PowerPoint file, and so I haven't seen it myself.
Not nessessarily. The 970+ is a second incarnation of the 970, aka GPUL, i's not a derivative of Power4+ aka GQ. Calling it GQUL would be wrong since GPUL already brached off of the Gx-tree.
And.. If this PDF is correct (and why shoudn't it? Even if there are some minor inconsistences in it) it's more likely that this is correct than NMR's artichle. It's is a document that comes directly from IBM after all, which NMR doesn't. It's probably composed from several existing presentations, made at different points in time. There might have been a "GR" Power4 before they renamed it Power5. Who knows? I don't.
The great finding of this document, as I see it, is the name "97x" for GRUL. We've always thought of it as 980 but that was a designation made up by someone (The Reg?) just because we didn't have anything better to call it. No we have, and it isn't called 980, it's called 97x.
The following snippet is from the presentation mentioning VMX and Power5 in almost the same breath (see Amorph's post above of info from M.Isobe on Ars):
-POWER5 GS Processor - 2.5GHz
-10GF per CPU (ie 10 giga flops...)
-single core (so they can get full bandwidth per core)
What does this say? That the Power5 will have a single core version (did we doubt this ) and that it will be fast (did we doubt this either ).
If the Power5 has VMX in it from the start then there will be much less work to make the GR-UL or GS-UL versions than there was for the Power4 based GP-UL so we should be able to expect the -UL version soon after, at the same time or even before the full Power5.
Comments
Now I guess they should have banned me rather than just shut off posting priviledges, because kickaha and Amorph definitely aren't going to like being called to task when they thought they had it all ignored *cough* *cough* I mean under control. Just a couple o' tools.
Don't worry, as soon as my work resetting my posts is done I'll disappear forever.
This PDF witch seems to be a presentation of some Holger Holthoff at IBM pSeries Technical Support held May 19th 2003:
Some interessting tibits:
Codenames:
GP = Power4 (1.3 GHz)
GQ = Power4+ (1.7 GHz)
GR = Power5 (1.8+ GHz)
GS = Power5+ (2.5+ GHz)
GPUL = 970 (1.8 GHz)
GRUL = 97x (2.5+ GHz)
So, there we have it. The Power5 derivative is GRUL, but it leaves GPUL2 which might well be 970+.
Originally posted by Henriok
He'sactually completely right!
This PDF witch seems to be a presentation of some Holger Holthoff at IBM pSeries Technical Support held May 19th 2003:
Some interessting tibits:
Codenames:
GP = Power4 (1.3 GHz)
GQ = Power4+ (1.7 GHz)
GR = Power5 (1.8+ GHz)
GS = Power5+ (2.5+ GHz)
GPUL = 970 (1.8 GHz)
GRUL = 97x (2.5+ GHz)
So, there we have it. The Power5 derivative is GRUL, but it leaves GPUL2 which might well be 970+.
Excellent! That pdf also says on p2 (top) that the GR-UL (aka Power 5 mini-me) is currently designated "ppc97x". Looks like no 980 for a long time then...
With the Power4 using thicker circuits for reliability I imagine the Power5 will do the same. Therefore, we should be able to expect the GR-UL to start in the 2.5+ GHz range as most have been guessing. This document seems to confirm this likelyhood.
MM
Originally posted by Henriok
He'sactually completely right!
This PDF witch seems to be a presentation of some Holger Holthoff at IBM pSeries Technical Support held May 19th 2003:
Some interessting tibits:
Codenames:
GP = Power4 (1.3 GHz)
GQ = Power4+ (1.7 GHz)
GR = Power5 (1.8+ GHz)
GS = Power5+ (2.5+ GHz)
GPUL = 970 (1.8 GHz)
GRUL = 97x (2.5+ GHz)
So, there we have it. The Power5 derivative is GRUL, but it leaves GPUL2 which might well be 970+.
No, that would have to be the GQUL if you're going to hold to the naming convention. No such thing as a GPUL2.
Look, you can try to second guess TGB and you'd have a good chance of succeeding. But there are those who say that TGB or NMR are just a vehicle some folk use to post rumors from sensitive sources that could be compromised if the info were to appear on their regular rumor site. A way of distancing reports from sources known to be at least moderately close to the publishers of said website.
And there are others who say it's just a dumping ground for info from less than reliable sources. Except TGB's track record is too good for that.
In any event, I'm going with TGB's assessment.
Originally posted by Tomb of the Unknown
No, that would have to be the GQUL if you're going to hold to the naming convention. No such thing as a GPUL2.
[snip]
In any event, I'm going with TGB's assessment.
I agree that there is an inconsitency in the pdf file. On one page it lists the Power4 on the CMOS 9S2 process as the "GR" (p1 at bottom) but on the next page (p2, top) it lists the "GR" as being Power5 along with the "GS". I haven't heard of IBM working on a low-k (which the "GR" is - p2, top) version of the Power4 so I'd bet that the error is on the first page. I've only heard of IBM working on the Power5 to suceed the current Power4 ("GQ").
Note that in the top slide on p2 it also mentions the GR-UL (the 97x). This says that the naming convention given by TGB of "P5UL" is inconsistent with IBM's current nomenclature and hence a little suspect.
I could be splitting hairs, clutching at straws, reading too much into the pdf, etc, but I think my statements above are reasonable.
MM
Originally posted by Tomb of the Unknown
No, that would have to be the GQUL if you're going to hold to the naming convention. No such thing as a GPUL2.
Not nessessarily. The 970+ is a second incarnation of the 970, aka GPUL, i's not a derivative of Power4+ aka GQ. Calling it GQUL would be wrong since GPUL already brached off of the Gx-tree.
And.. If this PDF is correct (and why shoudn't it? Even if there are some minor inconsistences in it) it's more likely that this is correct than NMR's artichle. It's is a document that comes directly from IBM after all, which NMR doesn't. It's probably composed from several existing presentations, made at different points in time. There might have been a "GR" Power4 before they renamed it Power5. Who knows? I don't.
The great finding of this document, as I see it, is the name "97x" for GRUL. We've always thought of it as 980 but that was a designation made up by someone (The Reg?) just because we didn't have anything better to call it. No we have, and it isn't called 980, it's called 97x.
Originally posted by Rolo
Here's something to chew on: The Smell of Fear There's so much confusion about the 980, I don't know what to believe. Is he right about the 980 being 4x faster per clock and shipping by the end of next year? Lots of other interesting stuff in the AppleLust editorial.
Wishful thinking.
Originally posted by Rolo
Here's something to chew on: The Smell of Fear There's so much confusion about the 980, I don't know what to believe. Is he right about the 980 being 4x faster per clock and shipping by the end of next year? Lots of other interesting stuff in the AppleLust editorial.
Well the Power 5 is supposed to be 4x faster per clock thatn the Power 4. So perhaps he is extrapolating that the 980 (if it is derived form the Power 5) will be 4x faster per clock than the 970 (power 4 derived).
Things like IBMs version of SMT and what not...
Originally posted by Rolo
Here's something to chew on: The Smell of Fear There's so much confusion about the 980, I don't know what to believe. Is he right about the 980 being 4x faster per clock and shipping by the end of next year? Lots of other interesting stuff in the AppleLust editorial.
I read and enjoyed the article several days ago. However, we have yet to get G5's into our hands. The notion, however, that there is confusion about the 980 is just silly because the processor has not been announced.
And there's no way the Power5 lite is 4 times as the 970 per clock in application performance, per core. No way! It might be 4x faster at certain operations (FastPath is a technology that is said to do so at certain operations, but not all, by far.) per clock, but overall, for regular code: Forget it.
The Power 5 is supposed to be 4 times faster than the original Power 4, not per clock.
Finally, he is himself too focused on platform bragging rights to acknowledge all the technical discussions on the general silliness of cross-platform benchmarking - there's some debate over whether it's even possible to compare the sort of things VeriTest was asked to compare in any meaningful way.
The important thing is that Apple has a very bright future ahead and that's all that really matters. I think there will be much gnashing of teeth in the PC world in the months and years ahead as Apple muscles in on some PC turf.
Originally posted by Outsider
My pet theory is that the Power5 is actually a dual core 980. This would make it possible to ship the 980 before the POWER5.
That'd indeed be very nice. And it'd be very cost-effective for both Apple and IBM, since they'd only have to develop a single version of the chip, and optimize that. It all depends on how high-end and costly IBM plan it to be (low-end at IBM is often as high as, or higher than Apples own high-end definition) , but they've indicated that it will be much more scalable, from low to high end.
I don't find it unlikely that the elusive 980 will be a single-core version of the Power 5
Originally posted by Zapchud
I don't find it unlikely that the elusive 980 will be a single-core version of the Power 5
It's actually what the 970 is to the Power4 minus VMX. Simplify with one design and increase cores to fit the required performance levels.
--
Ed
Originally posted by Ed M.
What if Power5 sports a VMX (AltiVec) unit?
According to a particularly juicy (if characteristically succinct) post by M. Isobe over on Ars, it just might have VMX on board. He dug up a reference to VMX from an IBM presentation apparently on the POWER5.
Disclaimer: The presentation's a PowerPoint file, and so I haven't seen it myself.
Originally posted by Henriok
Not nessessarily. The 970+ is a second incarnation of the 970, aka GPUL, i's not a derivative of Power4+ aka GQ. Calling it GQUL would be wrong since GPUL already brached off of the Gx-tree.
And.. If this PDF is correct (and why shoudn't it? Even if there are some minor inconsistences in it) it's more likely that this is correct than NMR's artichle. It's is a document that comes directly from IBM after all, which NMR doesn't. It's probably composed from several existing presentations, made at different points in time. There might have been a "GR" Power4 before they renamed it Power5. Who knows? I don't.
The great finding of this document, as I see it, is the name "97x" for GRUL. We've always thought of it as 980 but that was a designation made up by someone (The Reg?) just because we didn't have anything better to call it. No we have, and it isn't called 980, it's called 97x.
The following snippet is from the presentation mentioning VMX and Power5 in almost the same breath (see Amorph's post above of info from M.Isobe on Ars):
-POWER5 GS Processor - 2.5GHz
-10GF per CPU (ie 10 giga flops...)
-single core (so they can get full bandwidth per core)
What does this say? That the Power5 will have a single core version (did we doubt this
If the Power5 has VMX in it from the start then there will be much less work to make the GR-UL or GS-UL versions than there was for the Power4 based GP-UL so we should be able to expect the -UL version soon after, at the same time or even before the full Power5.
MM
Originally posted by Zapchud
The Power 5 is supposed to be 4 times faster than the original Power 4, not per clock.
IBM stated the Power5 will be 4 times as fast as the Power4 when it was introduced. The statement related to a 2GHz Power5 and a 1GHz Power4.