Reports from the trade show suggested that IBM did announce the chips at 2.5ghz (see macrumors for more info) .. This of course will ruin any surprise apple has for us, cause we know what the next PM's will sport.... unless... IBM was sandbagging a bit and didnt say everything at apples request..
Reports from the trade show suggested that IBM did announce the chips at 2.5ghz (see macrumors for more info) .. This of course will ruin any surprise apple has for us, cause we know what the next PM's will sport.... unless... IBM was sandbagging a bit and didnt say everything at apples request..
That would be nice... but somehow I doubt it. One thing that worries me is the fact that ibm released the 970 in jan of '03... g5's didn't come out (officially) until august. Do you think apple would do this again? I really hope not. I can't stand being on XP any more... I sold my mac in december thinking the g5's would come out in Jan...
Alms for the poor? Anyone have a g4 400 laying around
So hopefully Apple is ontop of their game and are going to implement the new 970fx (faster speeds) in the PowerMacs. Its sad that the XServe proc is better then the PowerMac.
Question: is there the same amount of transistors on the 970fx as the 970? 58 million some?
I think it is a misunderstanding. They refer to the processor type, not the frequency.
which part of this quote from the article, am I not understanding?
Quote:
IBM's 2.5GHz PowerPC 970FX ? used by Apple in its Xserve G5s ? will dissipate about 50 watts of power during typical use, down from 66 watts consumed by its PowerPC 970 predecessor at a lower clock speed, said Norman Rohrer, senior technical staff member at IBM. IBM's chip is also expected to power a new version of Apple's Power Mac.
Reports from the trade show suggested that IBM did announce the chips at 2.5ghz (see macrumors for more info) .. This of course will ruin any surprise apple has for us, cause we know what the next PM's will sport.... unless... IBM was sandbagging a bit and didnt say everything at apples request..
I wouldn't be surprised if IBM has 2.6-2.8 Ghz 970's coming off the line with an exclusive contract with Apple for those parts, thus not naming them as generally shipping in the announcement because they are only shipping to Apple. IBM has shipped faster processors than they said they were going to in their press releases in the past, no reason to think they wont this time either. But then again there is also no guarantee that they will ship anything faster either. We will just have to wait till Apple decides to ship PM's sometime in the next month, hopefully.
I wouldn't be surprised if IBM has 2.6-2.8 Ghz 970's coming off the line with an exclusive contract with Apple for those parts, thus not naming them as generally shipping in the announcement because they are only shipping to Apple. IBM has shipped faster processors than they said they were going to in their press releases in the past, no reason to think they wont this time either. But then again there is also no guarantee that they will ship anything faster either. We will just have to wait till Apple decides to ship PM's sometime in the next month, hopefully.
i'm with you on this!
my only reservation about this is that the high end config's won't ship til' much later than the low end ones like the first batch of G5's.... i want one now!!! hehehe
my only reservation about this is that the high end config's won't ship til' much later than the low end ones like the first batch of G5's.... i want one now!!! hehehe
Don't forget Steve's announcement last summer, they are working against the clock on that promise, and I don't think he made it lightly. He knows a lot more than we do on this subject. As other's have stated the timing of the next update, and the size of the speed boost should be in accordance with making that leap to over 3.0 Ghz by the end of the summer. I would go out on a limb and guess that we will see 2.6's shipping by the end of next month with 3.2's shipping by September. That may be optimistic, but I think that it is within the realm of realistic possibility.
What is puzzling to me is that the article says about 50W at 2.5 GHz. An earlier report stated just under 25W at 2.0 GHz. One of these is wrong. If the widely circulated earlier report is correct, the 970FX should be about 32W at 2.5 GHz.
What is puzzling to me is that the article says about 50W at 2.5 GHz. An earlier report stated just under 25W at 2.0 GHz. One of these is wrong. If the widely circulated earlier report is correct, the 970FX should be about 32W at 2.5 GHz.
WHAT?
dissipation rarely (read: never) scales linearly.
it is perfectly within the realms of possibility that the 2.5's are generally around 50w, while the 2.0 are around 25w. also, keep in mind that there have been different figures quoted...average versus max dissipation.
it is perfectly within the realms of possibility that the 2.5's are generally around 50w, while the 2.0 are around 25w. also, keep in mind that there have been different figures quoted...average versus max dissipation.
The same chip, run at the same supply voltage should scale fairly linearly with clock rate from what I understand. The majority of power dissipated comes from energy stored in the junction capacitances, which is proportional to switching frequency and the square of the supply voltage. I doubt they would change supply voltage to achieve a higher clock rate. There is also leakage current but I don't think this would change much with clock rate.
Your comment about maximum verses average or typical power may be valid, or it could simply be an error.
The delay in the new G5 product announcement may not be due to the CPU or support chips. Reports indicate that IBM is getting good yields at 3GHZ with the majority of chips capable of 2.8GHz.
While there may be enough 3.0GHz CPUs and support chips to support production, Apple may be be having trouble getting enough high speed memory to introduce new products. There is a report that Apple only has 25,000 memory devices in stock; they need 100,000 to announce the new G5 product.
The delay in the new G5 product announcement may not be due to the CPU or support chips. Reports indicate that IBM is getting good yields at 3GHZ with the majority of chips capable of 2.8GHz.
While there may be enough 3.0GHz CPUs and support chips to support production, Apple may be be having trouble getting enough high speed memory to introduce new products. There is a report that Apple only has 25,000 memory devices in stock; they need 100,000 to announce the new G5 product.
Intriguing. This is the first I've heard "reports" indicating 3.0 Ghz are even being produced.
The delay in the new G5 product announcement may not be due to the CPU or support chips. Reports indicate that IBM is getting good yields at 3GHZ with the majority of chips capable of 2.8GHz.
While there may be enough 3.0GHz CPUs and support chips to support production, Apple may be be having trouble getting enough high speed memory to introduce new products. There is a report that Apple only has 25,000 memory devices in stock; they need 100,000 to announce the new G5 product.
The same chip, run at the same supply voltage should scale fairly linearly with clock rate from what I understand. The majority of power dissipated comes from energy stored in the junction capacitances, which is proportional to switching frequency and the square of the supply voltage. I doubt they would change supply voltage to achieve a higher clock rate. There is also leakage current but I don't think this would change much with clock rate.
Your comment about maximum verses average or typical power may be valid, or it could simply be an error.
actually, it is quite common for dissipation to increase more than linearly, especially as you approach the limits of that given chip. for instance, from Intel data on the early P4 (1.4-2.0GHz) the dissipation was fairly constant (normalized for percent increase in frequency) until you reached the top end. the increase from 1.6 to 1.8 GHz (a 12.5% increase in frequency) showed an 8.72% increase in dissipation. however, the jump from 1.8 to 2.0GHz (an 11.1% increase in speed) showed an increase in dissipation of 13.92%.
if you really want, i can run a bunch more numbers for different chips/processes/etc. the main point is that leakage has always had a noticable effect as you approach the working limits of a given processor/process. I can't say I have any speific knowledge of the effects of IBM's SSOI on this, but I would guess that if they had overcome this common problem, there would've been a lot more press about it.
which part of this quote from the article, am I not understanding?
Quote:
[i]IBM's 2.5GHz PowerPC 970FX ? used by Apple in its Xserve G5s ? will dissipate about 50 watts of power during typical use, down from 66 watts consumed by its PowerPC 970 predecessor at a lower clock speed, said Norman Rohrer, senior technical staff member at IBM. IBM's chip is also expected to power a new version of Apple's Power Mac.[i]
it should read:
IBM's 2.5GHz, PowerPC 970FX ? used by Apple in its Xserve G5s, ? will dissipate about 50 watts of power during typical use, down from 66 watts consumed by its PowerPC 970 predecessor at a lower clock speed, said Norman Rohrer, senior technical staff member at IBM. IBM's chip is also expected to power a new version of Apple's Power Mac.
They are simply saying the 970fx is being used in the Xserve. They are just writing it wrong. (they are from europe
actually, it is quite common for dissipation to increase more than linearly, especially as you approach the limits of that given chip. for instance, from Intel data on the early P4 (1.4-2.0GHz) the dissipation was fairly constant (normalized for percent increase in frequency) until you reached the top end. the increase from 1.6 to 1.8 GHz (a 12.5% increase in frequency) showed an 8.72% increase in dissipation. however, the jump from 1.8 to 2.0GHz (an 11.1% increase in speed) showed an increase in dissipation of 13.92%.
if you really want, i can run a bunch more numbers for different chips/processes/etc. the main point is that leakage has always had a noticable effect as you approach the working limits of a given processor/process. I can't say I have any speific knowledge of the effects of IBM's SSOI on this, but I would guess that if they had overcome this common problem, there would've been a lot more press about it.
I can easily accept these smaller deviation from linear you mention, but the 50W at 2.5 GHz seems like an extreme jump if 25W at 2.0 GHz is correct. Regarding leakage, I would expect it to increase with higher supply voltage and not much with greater clock rates at the same voltage. But I'm just looking at things theoretically, and rather superficially at that. Real life is always more complicated.
I remember there was discussion about the low leakage IBM was able to achieve in contrast to what Intel is doing at 90 nm. It was a few days back on the Register. I'm almost asleep, otherwise I'd look it up. I've enjoyed the exchange.
Intriguing. This is the first I've heard "reports" indicating 3.0 Ghz are even being produced.
Quote:
Originally posted by whoami
hmmmm...
where did you stumble upon this info?
I think he talks about information provided by Croquer. They are the first to report these numbers about frequency and memory stock, and they say that Apple would wait until WWDC in order to announce the new Power Macs.
Heck, why not wait until WWDC? Are people really cringing this much for new Power Macs? If they can deliver good enough yields @ 3GHz AT WWDC (i.e., ready to order and arrive at people's homes shortly), then let them do it. Not to mention, if new PowerBooks appear therewith......
Comments
Originally posted by sls
30 W Opteron EE...
Just thought I'd add that it's 30 W @ 1.4GHz. But still, kudos to AMD considering this is still on a 0.13µm process.
http://www.macworld.co.uk/news/main_...fm?NewsID=7942
Originally posted by oldmacfan
I found this article interesting, it states that Apple is using the 2.5ghz G5 in the X-Serve
http://www.macworld.co.uk/news/main_...fm?NewsID=7942
I think it is a misunderstanding. They refer to the processor type, not the frequency.
Originally posted by gsxrboy
Reports from the trade show suggested that IBM did announce the chips at 2.5ghz (see macrumors for more info) .. This of course will ruin any surprise apple has for us, cause we know what the next PM's will sport.... unless... IBM was sandbagging a bit and didnt say everything at apples request..
That would be nice... but somehow I doubt it. One thing that worries me is the fact that ibm released the 970 in jan of '03... g5's didn't come out (officially) until august. Do you think apple would do this again? I really hope not. I can't stand being on XP any more... I sold my mac in december thinking the g5's would come out in Jan...
Alms for the poor? Anyone have a g4 400 laying around
So hopefully Apple is ontop of their game and are going to implement the new 970fx (faster speeds) in the PowerMacs. Its sad that the XServe proc is better then the PowerMac.
Question: is there the same amount of transistors on the 970fx as the 970? 58 million some?
Originally posted by PB
I think it is a misunderstanding. They refer to the processor type, not the frequency.
which part of this quote from the article, am I not understanding?
IBM's 2.5GHz PowerPC 970FX ? used by Apple in its Xserve G5s ? will dissipate about 50 watts of power during typical use, down from 66 watts consumed by its PowerPC 970 predecessor at a lower clock speed, said Norman Rohrer, senior technical staff member at IBM. IBM's chip is also expected to power a new version of Apple's Power Mac.
Originally posted by gsxrboy
Reports from the trade show suggested that IBM did announce the chips at 2.5ghz (see macrumors for more info) .. This of course will ruin any surprise apple has for us, cause we know what the next PM's will sport.... unless... IBM was sandbagging a bit and didnt say everything at apples request..
I wouldn't be surprised if IBM has 2.6-2.8 Ghz 970's coming off the line with an exclusive contract with Apple for those parts, thus not naming them as generally shipping in the announcement because they are only shipping to Apple. IBM has shipped faster processors than they said they were going to in their press releases in the past, no reason to think they wont this time either. But then again there is also no guarantee that they will ship anything faster either. We will just have to wait till Apple decides to ship PM's sometime in the next month, hopefully.
Originally posted by @homenow
I wouldn't be surprised if IBM has 2.6-2.8 Ghz 970's coming off the line with an exclusive contract with Apple for those parts, thus not naming them as generally shipping in the announcement because they are only shipping to Apple. IBM has shipped faster processors than they said they were going to in their press releases in the past, no reason to think they wont this time either. But then again there is also no guarantee that they will ship anything faster either. We will just have to wait till Apple decides to ship PM's sometime in the next month, hopefully.
i'm with you on this!
my only reservation about this is that the high end config's won't ship til' much later than the low end ones like the first batch of G5's.... i want one now!!! hehehe
Originally posted by whoami
i'm with you on this!
my only reservation about this is that the high end config's won't ship til' much later than the low end ones like the first batch of G5's.... i want one now!!! hehehe
Don't forget Steve's announcement last summer, they are working against the clock on that promise, and I don't think he made it lightly. He knows a lot more than we do on this subject. As other's have stated the timing of the next update, and the size of the speed boost should be in accordance with making that leap to over 3.0 Ghz by the end of the summer. I would go out on a limb and guess that we will see 2.6's shipping by the end of next month with 3.2's shipping by September. That may be optimistic, but I think that it is within the realm of realistic possibility.
Originally posted by oldmacfan
I found this article interesting, it states that Apple is using the 2.5ghz G5 in the X-Serve
http://www.macworld.co.uk/news/main_...fm?NewsID=7942
What is puzzling to me is that the article says about 50W at 2.5 GHz. An earlier report stated just under 25W at 2.0 GHz. One of these is wrong. If the widely circulated earlier report is correct, the 970FX should be about 32W at 2.5 GHz.
Originally posted by snoopy
What is puzzling to me is that the article says about 50W at 2.5 GHz. An earlier report stated just under 25W at 2.0 GHz. One of these is wrong. If the widely circulated earlier report is correct, the 970FX should be about 32W at 2.5 GHz.
WHAT?
dissipation rarely (read: never) scales linearly.
it is perfectly within the realms of possibility that the 2.5's are generally around 50w, while the 2.0 are around 25w. also, keep in mind that there have been different figures quoted...average versus max dissipation.
Originally posted by concentricity
WHAT?
dissipation rarely (read: never) scales linearly.
it is perfectly within the realms of possibility that the 2.5's are generally around 50w, while the 2.0 are around 25w. also, keep in mind that there have been different figures quoted...average versus max dissipation.
The same chip, run at the same supply voltage should scale fairly linearly with clock rate from what I understand. The majority of power dissipated comes from energy stored in the junction capacitances, which is proportional to switching frequency and the square of the supply voltage. I doubt they would change supply voltage to achieve a higher clock rate. There is also leakage current but I don't think this would change much with clock rate.
Your comment about maximum verses average or typical power may be valid, or it could simply be an error.
While there may be enough 3.0GHz CPUs and support chips to support production, Apple may be be having trouble getting enough high speed memory to introduce new products. There is a report that Apple only has 25,000 memory devices in stock; they need 100,000 to announce the new G5 product.
Originally posted by shawk
The delay in the new G5 product announcement may not be due to the CPU or support chips. Reports indicate that IBM is getting good yields at 3GHZ with the majority of chips capable of 2.8GHz.
While there may be enough 3.0GHz CPUs and support chips to support production, Apple may be be having trouble getting enough high speed memory to introduce new products. There is a report that Apple only has 25,000 memory devices in stock; they need 100,000 to announce the new G5 product.
Intriguing. This is the first I've heard "reports" indicating 3.0 Ghz are even being produced.
Originally posted by shawk
The delay in the new G5 product announcement may not be due to the CPU or support chips. Reports indicate that IBM is getting good yields at 3GHZ with the majority of chips capable of 2.8GHz.
While there may be enough 3.0GHz CPUs and support chips to support production, Apple may be be having trouble getting enough high speed memory to introduce new products. There is a report that Apple only has 25,000 memory devices in stock; they need 100,000 to announce the new G5 product.
hmmmm...
where did you stumble upon this info?
Originally posted by snoopy
The same chip, run at the same supply voltage should scale fairly linearly with clock rate from what I understand. The majority of power dissipated comes from energy stored in the junction capacitances, which is proportional to switching frequency and the square of the supply voltage. I doubt they would change supply voltage to achieve a higher clock rate. There is also leakage current but I don't think this would change much with clock rate.
Your comment about maximum verses average or typical power may be valid, or it could simply be an error.
actually, it is quite common for dissipation to increase more than linearly, especially as you approach the limits of that given chip. for instance, from Intel data on the early P4 (1.4-2.0GHz) the dissipation was fairly constant (normalized for percent increase in frequency) until you reached the top end. the increase from 1.6 to 1.8 GHz (a 12.5% increase in frequency) showed an 8.72% increase in dissipation. however, the jump from 1.8 to 2.0GHz (an 11.1% increase in speed) showed an increase in dissipation of 13.92%.
if you really want, i can run a bunch more numbers for different chips/processes/etc. the main point is that leakage has always had a noticable effect as you approach the working limits of a given processor/process. I can't say I have any speific knowledge of the effects of IBM's SSOI on this, but I would guess that if they had overcome this common problem, there would've been a lot more press about it.
Originally posted by oldmacfan
which part of this quote from the article, am I not understanding?
[i]IBM's 2.5GHz PowerPC 970FX ? used by Apple in its Xserve G5s ? will dissipate about 50 watts of power during typical use, down from 66 watts consumed by its PowerPC 970 predecessor at a lower clock speed, said Norman Rohrer, senior technical staff member at IBM. IBM's chip is also expected to power a new version of Apple's Power Mac.[i]
it should read:
IBM's 2.5GHz, PowerPC 970FX ? used by Apple in its Xserve G5s, ? will dissipate about 50 watts of power during typical use, down from 66 watts consumed by its PowerPC 970 predecessor at a lower clock speed, said Norman Rohrer, senior technical staff member at IBM. IBM's chip is also expected to power a new version of Apple's Power Mac.
They are simply saying the 970fx is being used in the Xserve. They are just writing it wrong. (they are from europe
Originally posted by concentricity
actually, it is quite common for dissipation to increase more than linearly, especially as you approach the limits of that given chip. for instance, from Intel data on the early P4 (1.4-2.0GHz) the dissipation was fairly constant (normalized for percent increase in frequency) until you reached the top end. the increase from 1.6 to 1.8 GHz (a 12.5% increase in frequency) showed an 8.72% increase in dissipation. however, the jump from 1.8 to 2.0GHz (an 11.1% increase in speed) showed an increase in dissipation of 13.92%.
if you really want, i can run a bunch more numbers for different chips/processes/etc. the main point is that leakage has always had a noticable effect as you approach the working limits of a given processor/process. I can't say I have any speific knowledge of the effects of IBM's SSOI on this, but I would guess that if they had overcome this common problem, there would've been a lot more press about it.
I can easily accept these smaller deviation from linear you mention, but the 50W at 2.5 GHz seems like an extreme jump if 25W at 2.0 GHz is correct. Regarding leakage, I would expect it to increase with higher supply voltage and not much with greater clock rates at the same voltage. But I'm just looking at things theoretically, and rather superficially at that. Real life is always more complicated.
I remember there was discussion about the low leakage IBM was able to achieve in contrast to what Intel is doing at 90 nm. It was a few days back on the Register. I'm almost asleep, otherwise I'd look it up. I've enjoyed the exchange.
Originally posted by LudwigVan
Intriguing. This is the first I've heard "reports" indicating 3.0 Ghz are even being produced.
Originally posted by whoami
hmmmm...
where did you stumble upon this info?
I think he talks about information provided by Croquer. They are the first to report these numbers about frequency and memory stock, and they say that Apple would wait until WWDC in order to announce the new Power Macs.