First, how do you know those numbers are the "TYPICAL" (not that I doubt you, just curious).
Also, where do you get the "1.3 GHz 970 30W @ 1.3V"? You have said it twice, but it doesn't make it any more true unless we see some good figures (do they even make a 1.3GHz 970?)
I am confident in the future of the Power PC chips. Even if IBM has shown some disinterest in desktop powerPC chip, he has spent a lot of money, and research in the developpement of the power chips.
The power 4 chips is the result of a 4 years developpement project. This is a great amount of research for a chip. The result is an oustanding core. And this core his higly pipelined, and have the room to further developpement without having to make a new design from scratch.
The PPC 970 is the single core VMX enhanced version of the power 4. When you compare it to the 7457, you understand that he is way above it, despite the same amount of transistors, the same fabbing process, and nearly the same power consumption at equal mhz.
The PPC 9xx will evoluate in parallelar with the power chips, because the PPC 9XX are the light version of the power chips family. Most of the R&D, spend for the power line, is integrated in the PPC 9xx line. Futhermore, with the new products of developpement (see the programmer posts), both improvement will be made at the same time : it means that the next generation of PPC 9xx chip could be released only a couple of months after the power 5.
Some features of the new PPC chip can be extrapolated from various rumors and interviews.
- the chip will have Hyperthreading
- it will be built upon 90 nm process
- it will have 1 MB of L2 cache
- the VMX engine will be enhanced, like the one of the 745X series.
First, how do you know those numbers are the "TYPICAL" (not that I doubt you, just curious).
Also, where do you get the "1.3 GHz 970 30W @ 1.3V"? You have said it twice, but it doesn't make it any more true unless we see some good figures (do they even make a 1.3GHz 970?)
Nothing prevent IBM to downclock a chip. Any PPC 970 can be downclocked. However, decrease the voltage is a much harder job.
I still maintain that there is no real technical reason Apple can't put the current 970 in Powerbooks right now @ 1.0 and 1.2Ghz. The heat and power consumption characteristics would be somewhat less than optimal, but not unfeasible. Power and heat will gradually decrease in magnitude as the efficiency of the production line ramps up and more of the processes switch to 90 nm. So 970 Powerbooks by next spring, guaranteed
I agree with Amorph. We know very little of the details of the 970 to be able to second guess Apple's engineers. You have to imagine that they have been prototyping a G5 PB since they first got evaluation samples. I would venture that not only is it too hot but it may be missing some key features that are needed in a PB. Can a 970 go to sleep? Maybe it needs a constant clock to prevent losing some items. It appears to have some power management. How good is it? How about the rest of the circuitry? Will a much faster bus draw much more power and generate much more heat? How much power does Apple's ASIC draw?
Certainly, some engineers at Apple are looking at this and working in tandem with IBM to develop solutions. My guess is it will September 2004 till we see G5 PBs in full production. The announcement might come earlier.
Until then we'll see a few more speed bumps and improvements in the G4. Nothing real dramatic but helpful nonetheless.
IBM has spent a lot at fishkill, but they're own statements reveal that the money has been spent in FAB and DESIGN expertise, not neccessarily PPC fab and design
PPC research and design is not location specific. Lots goes on in Austin, and elsewhere. IBM has a major govt research grant on low-power computing, and is also doing major research on super dense servers. Both of these, along with the stated intent of entering a PPC chip into the blade market seems to point to a very rich-future for PPC design and research. Also note that if Moto continues to screw the pooch, a die-shrink 970 would make the 970 a major hit in many of Moto's embedded markets.
The L3 cache is not on the die of the G4, right. Would the power consumptions for the MPC7455 and MPC7457 include the power consumption attributable to the L3 cache?? I would think not, but that's why I'm asking.
After all, I would think that 1MB or 2MB's of L3 caches may consume some power, wouldn't they?
Collated and extrapolated for you viewing pleasure. And yes, the PPC 970 has all the same power modes as the 750fx (per some interview with IBM guy I read somewhere), which pretty means it's the same as the 7455. Neither has the power management abilities that Centrino has, combined voltage and frequency cycling, but there is always hope for the future (of the 970). The 970 really needs it compared to the 7457.
Code:
Typical Power Consumption in Watts
Clock Rate
Volt 0.87 1.00 1.30 1.80 GHz
---- ---- ---- ---- ----
7455 1.6 18.5 21.3 27.7*
7455 1.3 12.1* 14.0*
7457 1.3 14.8 15.8 18.7 25.9*
7457 1.1 10.6* 11.3* 13.4*
970 1.3 26.1* 33.9* 47
970 1.1 18.7* 24.3* 33.7*
Maximum Power Consumption in Watts
Clock Rate
Volt 0.87 1.00 1.30 1.80 GHz
---- ---- ---- ---- ----
7455 1.6 26.0 30.0 39*
7455 1.3 17.2* 19.9*
7457 1.3 21.0 22.0 26.0 36*
7457 1.1 15.0* 15.8* 18.8*
970 1.3 36.3* 47.2* 65.3*
970 1.1 26.0* 33.8* 46.8*
"*" means extrapolated. The 47 Watts @ 1.8 GHz for the 970 is the single official IBM info I have seen for the 970 (courtesy of Rickag). I'm assuming it is a typical number. All PPC 970 numbers are extrapolated from that number. Maximum 970 numbers are extrapolated based on 7457 typical/maximum ratio. My conclusion, at 1.3 GHz maximum power consumption for the 1.1V 970 is 34 Watts while the 1.3V 7457 is 26 Watts. Put the 970 in the iMac and Powerbook.
The L3 cache is not on the die of the G4, right. Would the power consumptions for the MPC7455 and MPC7457 include the power consumption attributable to the L3 cache?? I would think not, but that's why I'm asking.
Of course not, Rickag.
Quote:
After all, I would think that 1MB or 2MB's of L3 caches may consume some power, wouldn't they?
Around 1 to 2 Watts maximum, I think. Maybe some of the high MHz SRAM chips will reach 5 Watts, not sure.
What I mean to say is that to get a good reliable 1.2Ghz part for a notebook, you might just need a 1.8 level part anyway (with the same costs). From what I've understood it isn't a simple matter of using lower spec parts when it comes to the heat and volatge requirements of a laptop.
It's true that low-voltage CPUs have to be culled from among the highest quality parts. They're not cheap. Neither are PowerBooks.
Basically, every chip is flawed to some degree. If you think of electricity as water, voltage is sort of like the speed of the current. A chip that can be run at low voltage is one that runs efficiently enough that you a relatively lazy current is sufficient to turn the wheel, so to speak. Lower-clocked chips are inefficient enough that the current needs a little more oomph.
Quote:
That means that they have to wait on the G5 a bit, the pessimist in me says Q3-4 '04, but summer '04 seems right about on as far as a G5 PB goes, mebbe it will even feature a .13u G5 (a more expensive, and not particularly high clock one if it comes to that)
Anything that late will have a 90nm part unless something goes horribly wrong at Fishkill. Note that Fishkill was designed to be a 90nm fab. The 130nm manufacturing was always intended to be a short term stop gap before the plant really got going.
Quote:
And, though it seems illogical, I'm pulling for Moto to get their stuff together by the time they move to .09u. IBM may be the bees knees right now, but a second option is needed, if only as a back-up.
Heck, I'm pulling for Mot as well. I hope Crolles is everything they hope it is. Options are good.
Quote:
I've argued this before, they can switch at any time, or redirect the PPC to areas not of use to Apple, ie the POWER end of the equation.
Interestingly, IBM is currently doing the opposite, and redirecting the POWER line toward the PowerPC. The POWER5 has been widely reported as being small and cheap and scalable. The lumbering-monster strategy appears to be going out with the POWER4. In other words, the difference between a POWER chip and a PowerPC might come down to issues like a few dedicated units, high-end busses, and gate thickness, and supporting a PowerPC platform would therefor become really inexpensive.
Of course, Apple has CPU designers in house, and they have had since AIM started (it's not widely known - or perhaps, it's widely denied - that the lead architect for what became AltiVec was an Apple employee). So although they don't appear to have the resources to become their own fabless design shop, they do know what they want and they can and do work with their partners at a planning and design level to get it. So IBM's Fishkill strategy and POWER strategy both look very good for Apple.
My conclusion, at 1.3 GHz maximum power consumption for the 1.1V 970 is 34 Watts while the 1.3V 7457 is 26 Watts. Put the 970 in the iMac and Powerbook.
Good work THT, of course Im sure these figures are fairly "rough" based on your assumptions, even IBM's docs have the 970 @ 1.2GHz @ 19W *estimated. Unless Moto can supply 7457's now and get them to 2GHz within 6 months - the 970 is likely to find its way into quite a few of apple's product line.
"*" means extrapolated. The 47 Watts @ 1.8 GHz for the 970 is the single official IBM info I have seen for the 970 (courtesy of Rickag).
Hannibal over at arstechnica usually has pretty good info, and he lists a 1.3v 1.8GHz 970 at 42W. He also lists a 1.2GHz (no voltage mentioned) at 19W. I think he got his info from the IBM presentation of the unveiling of the 970 (at that microprocessor show). This could have been estimates on IBMs part, or values from preproduction chips or something, and the true value is [email protected]. Does Rickag work for IBM (sorry if that is a stupid question), and if not, where did he get the info?
If we assumed these 2 values to be the typical, how would that change your chart (I would guess having 2 values would be better than 1 when doing the kind of extrapolation you did).
They are listed as typical, and are listed as estimates. So Rickag could be very right, but if so I would love to knwo where he got the info from (assuming he can say).
Lamentably, I don't work for IBM, I would be making a ton more money and have much more knowledge about computers.
I get my information about the 970 from IBM documents and the internet just like most everyone else here.
Then would you mind pointing me to where you read the 1.8GHz 970 would be 47W. THT mentioned that you had provided the 47W figure, so that is why I thought you might work for IBM or something
If it turns out you misread, the table THT created could be off. And with the 1.2GHz figure, that might help create a more realistic table (having two data-points for extrapolation would be better than one). Or you could be very right, I could be wrong, and the table could be pretty close.
Then would you mind pointing me to where you read the 1.8GHz 970 would be 47W. THT mentioned that you had provided the 47W figure, so that is why I thought you might work for IBM or something
The 47 Watts comes from IBM's PowerPC Quick Reference Guide. The PPC 970 is on page 9 of 10. It's the only non-estimated number I've seen. Being 5 Watts off is reasonable. The typical 10 to 15% error for estimates.
But one has to note what typical really means, since I'm sure IBM, Motorola, and any other CPU vendor have different instruction mixes used for their "typical" power consumption numbers.
Someone from PowerLogix has been active on this forum talking about future upgrades. What I get from his posts is that they have 7457's at speed's of up to 1.6 Ghz for testing, and they are designing a daughter card for the Cube that would handle 2 of these. He dosnt give much in the way of when he expects to have these chips available for production. Given the history of Motorolla giving chips to Apple first, as much as 6-12 months before upgrade manufacturers get ahold of them, I would imagine that Apple has these chips in hand. They may be waiting on "sustainable" supplies from Moto so they dont relive the "G4 500" fiasco.
Other interesting quotes:
Quote:
Interestingly, as chip designs advance, power requirements go down and speeds go up...this is what is happening with the transition from 7455 to 7457. The 7457s will draw less power than an equivalent (or even greater) speed 7455.
A 1.6 7457 (when they are available) will draw far less power than a 1.4 7455.
Quote:
(panphage @ Jul 29 2003, 08:57 PM)
I keep reading about the power gains from the die shrink. But looking at moto's pdfs, the 7455 @1.0Ghz uses a max of 22W while the 7457 @ 1.0Ghz uses the exact same 22W max. That would be a power savings of exactly nothing. And in fact the typical usage of the 7457 @1.0 Ghz is 15.8W while the 7455 at the same speed is rated 15.0W typical. Am I reading these wrong? If I'm not and these numbers are true, a dual 1.6 ain't gonna happen. I mean in a cube.
Basically, that's incorrect info. I'll post something later to update.
No update yet, but it sounds like he does have som, or has access to some data that is different than the pdf the Moto publically released, possibly real world, and is confidant on the power usage of the 7457's.
Quote:
I expect to see slower 7457s (slower than 1.6 that is), yes. Not sure exactly what speeds yet, we are waiting on Motorola.
Wouldn't it be sensible for MOT to consider contracting the 7457 to Fishkill and save some face and get things back on track for hardware updates?
If start-up is as fast and easy as we have heard, then its good for everyone, sure MOT gives up some profit, but every day that goes by, the chips they have are worth less as alternatives become more viable.
Let's find out how fast IBM can gear up its new wafer plant at .13u?
Comments
Originally posted by smalM
1.2 GHz 970 19W @ 1.1V typical
1.3 GHz 970 30W @ 1.3V typical
TYPICAL
First, how do you know those numbers are the "TYPICAL" (not that I doubt you, just curious).
Also, where do you get the "1.3 GHz 970 30W @ 1.3V"? You have said it twice, but it doesn't make it any more true unless we see some good figures (do they even make a 1.3GHz 970?)
The power 4 chips is the result of a 4 years developpement project. This is a great amount of research for a chip. The result is an oustanding core. And this core his higly pipelined, and have the room to further developpement without having to make a new design from scratch.
The PPC 970 is the single core VMX enhanced version of the power 4. When you compare it to the 7457, you understand that he is way above it, despite the same amount of transistors, the same fabbing process, and nearly the same power consumption at equal mhz.
The PPC 9xx will evoluate in parallelar with the power chips, because the PPC 9XX are the light version of the power chips family. Most of the R&D, spend for the power line, is integrated in the PPC 9xx line. Futhermore, with the new products of developpement (see the programmer posts), both improvement will be made at the same time : it means that the next generation of PPC 9xx chip could be released only a couple of months after the power 5.
Some features of the new PPC chip can be extrapolated from various rumors and interviews.
- the chip will have Hyperthreading
- it will be built upon 90 nm process
- it will have 1 MB of L2 cache
- the VMX engine will be enhanced, like the one of the 745X series.
Originally posted by kupan787
First, how do you know those numbers are the "TYPICAL" (not that I doubt you, just curious).
Also, where do you get the "1.3 GHz 970 30W @ 1.3V"? You have said it twice, but it doesn't make it any more true unless we see some good figures (do they even make a 1.3GHz 970?)
Nothing prevent IBM to downclock a chip. Any PPC 970 can be downclocked. However, decrease the voltage is a much harder job.
Originally posted by hotboxd
Nope, nor will it be.
I still maintain that there is no real technical reason Apple can't put the current 970 in Powerbooks right now @ 1.0 and 1.2Ghz. The heat and power consumption characteristics would be somewhat less than optimal, but not unfeasible. Power and heat will gradually decrease in magnitude as the efficiency of the production line ramps up and more of the processes switch to 90 nm. So 970 Powerbooks by next spring, guaranteed
I agree with Amorph. We know very little of the details of the 970 to be able to second guess Apple's engineers. You have to imagine that they have been prototyping a G5 PB since they first got evaluation samples. I would venture that not only is it too hot but it may be missing some key features that are needed in a PB. Can a 970 go to sleep? Maybe it needs a constant clock to prevent losing some items. It appears to have some power management. How good is it? How about the rest of the circuitry? Will a much faster bus draw much more power and generate much more heat? How much power does Apple's ASIC draw?
Certainly, some engineers at Apple are looking at this and working in tandem with IBM to develop solutions. My guess is it will September 2004 till we see G5 PBs in full production. The announcement might come earlier.
Until then we'll see a few more speed bumps and improvements in the G4. Nothing real dramatic but helpful nonetheless.
IBM has spent a lot at fishkill, but they're own statements reveal that the money has been spent in FAB and DESIGN expertise, not neccessarily PPC fab and design
PPC research and design is not location specific. Lots goes on in Austin, and elsewhere. IBM has a major govt research grant on low-power computing, and is also doing major research on super dense servers. Both of these, along with the stated intent of entering a PPC chip into the blade market seems to point to a very rich-future for PPC design and research. Also note that if Moto continues to screw the pooch, a die-shrink 970 would make the 970 a major hit in many of Moto's embedded markets.
The L3 cache is not on the die of the G4, right. Would the power consumptions for the MPC7455 and MPC7457 include the power consumption attributable to the L3 cache?? I would think not, but that's why I'm asking.
After all, I would think that 1MB or 2MB's of L3 caches may consume some power, wouldn't they?
Typical Power Consumption in Watts
Clock Rate
Volt 0.87 1.00 1.30 1.80 GHz
---- ---- ---- ---- ----
7455 1.6 18.5 21.3 27.7*
7455 1.3 12.1* 14.0*
7457 1.3 14.8 15.8 18.7 25.9*
7457 1.1 10.6* 11.3* 13.4*
970 1.3 26.1* 33.9* 47
970 1.1 18.7* 24.3* 33.7*
Maximum Power Consumption in Watts
Clock Rate
Volt 0.87 1.00 1.30 1.80 GHz
---- ---- ---- ---- ----
7455 1.6 26.0 30.0 39*
7455 1.3 17.2* 19.9*
7457 1.3 21.0 22.0 26.0 36*
7457 1.1 15.0* 15.8* 18.8*
970 1.3 36.3* 47.2* 65.3*
970 1.1 26.0* 33.8* 46.8*
"*" means extrapolated. The 47 Watts @ 1.8 GHz for the 970 is the single official IBM info I have seen for the 970 (courtesy of Rickag). I'm assuming it is a typical number. All PPC 970 numbers are extrapolated from that number. Maximum 970 numbers are extrapolated based on 7457 typical/maximum ratio. My conclusion, at 1.3 GHz maximum power consumption for the 1.1V 970 is 34 Watts while the 1.3V 7457 is 26 Watts. Put the 970 in the iMac and Powerbook.
Originally posted by rickag
The L3 cache is not on the die of the G4, right. Would the power consumptions for the MPC7455 and MPC7457 include the power consumption attributable to the L3 cache?? I would think not, but that's why I'm asking.
Of course not, Rickag.
After all, I would think that 1MB or 2MB's of L3 caches may consume some power, wouldn't they?
Around 1 to 2 Watts maximum, I think. Maybe some of the high MHz SRAM chips will reach 5 Watts, not sure.
Originally posted by Matsu
What I mean to say is that to get a good reliable 1.2Ghz part for a notebook, you might just need a 1.8 level part anyway (with the same costs). From what I've understood it isn't a simple matter of using lower spec parts when it comes to the heat and volatge requirements of a laptop.
It's true that low-voltage CPUs have to be culled from among the highest quality parts. They're not cheap. Neither are PowerBooks.
Basically, every chip is flawed to some degree. If you think of electricity as water, voltage is sort of like the speed of the current. A chip that can be run at low voltage is one that runs efficiently enough that you a relatively lazy current is sufficient to turn the wheel, so to speak. Lower-clocked chips are inefficient enough that the current needs a little more oomph.
That means that they have to wait on the G5 a bit, the pessimist in me says Q3-4 '04, but summer '04 seems right about on as far as a G5 PB goes, mebbe it will even feature a .13u G5 (a more expensive, and not particularly high clock one if it comes to that)
Anything that late will have a 90nm part unless something goes horribly wrong at Fishkill. Note that Fishkill was designed to be a 90nm fab. The 130nm manufacturing was always intended to be a short term stop gap before the plant really got going.
And, though it seems illogical, I'm pulling for Moto to get their stuff together by the time they move to .09u. IBM may be the bees knees right now, but a second option is needed, if only as a back-up.
Heck, I'm pulling for Mot as well. I hope Crolles is everything they hope it is. Options are good.
I've argued this before, they can switch at any time, or redirect the PPC to areas not of use to Apple, ie the POWER end of the equation.
Interestingly, IBM is currently doing the opposite, and redirecting the POWER line toward the PowerPC. The POWER5 has been widely reported as being small and cheap and scalable. The lumbering-monster strategy appears to be going out with the POWER4. In other words, the difference between a POWER chip and a PowerPC might come down to issues like a few dedicated units, high-end busses, and gate thickness, and supporting a PowerPC platform would therefor become really inexpensive.
Of course, Apple has CPU designers in house, and they have had since AIM started (it's not widely known - or perhaps, it's widely denied - that the lead architect for what became AltiVec was an Apple employee). So although they don't appear to have the resources to become their own fabless design shop, they do know what they want and they can and do work with their partners at a planning and design level to get it. So IBM's Fishkill strategy and POWER strategy both look very good for Apple.
Originally posted by THT
[B]Collated and extrapolated for you viewing pleasure.
Just dumped my post reading your table
Two remarks:
- I think the maximum numbers for the 970 are higher. This chips has to handle much more data when charched to capicity then a 7457.
- IBM claimed the 970 can reach only 1.2GHz @ 1.1V. But this is the future hardware forum so you meant the .09u version
Originally posted by THT
My conclusion, at 1.3 GHz maximum power consumption for the 1.1V 970 is 34 Watts while the 1.3V 7457 is 26 Watts. Put the 970 in the iMac and Powerbook.
Good work THT, of course Im sure these figures are fairly "rough" based on your assumptions, even IBM's docs have the 970 @ 1.2GHz @ 19W *estimated. Unless Moto can supply 7457's now and get them to 2GHz within 6 months - the 970 is likely to find its way into quite a few of apple's product line.
Thank you for the information. While 1-2 watts may not be that significant for desktops, it may affect some decisions Apple makes regarding laptops.
Originally posted by THT
"*" means extrapolated. The 47 Watts @ 1.8 GHz for the 970 is the single official IBM info I have seen for the 970 (courtesy of Rickag).
Hannibal over at arstechnica usually has pretty good info, and he lists a 1.3v 1.8GHz 970 at 42W. He also lists a 1.2GHz (no voltage mentioned) at 19W. I think he got his info from the IBM presentation of the unveiling of the 970 (at that microprocessor show). This could have been estimates on IBMs part, or values from preproduction chips or something, and the true value is [email protected]. Does Rickag work for IBM (sorry if that is a stupid question), and if not, where did he get the info?
If we assumed these 2 values to be the typical, how would that change your chart (I would guess having 2 values would be better than 1 when doing the kind of extrapolation you did).
[EDIT] Ok in this PDF: http://www-3.ibm.com/chips/techlib/t...56C5200611780/$file/PPC970_MPF2002.pdf on page 14 it says:
42W @ 1.8GHz, 1.3v
19W @ 1.2GHz, 1.1v
They are listed as typical, and are listed as estimates. So Rickag could be very right, but if so I would love to knwo where he got the info from (assuming he can say).
Lamentably, I don't work for IBM, I would be making a ton more money and have much more knowledge about computers.
I get my information about the 970 from IBM documents and the internet just like most everyone else here.
Originally posted by rickag
kupan787
Lamentably, I don't work for IBM, I would be making a ton more money and have much more knowledge about computers.
I get my information about the 970 from IBM documents and the internet just like most everyone else here.
Then would you mind pointing me to where you read the 1.8GHz 970 would be 47W. THT mentioned that you had provided the 47W figure, so that is why I thought you might work for IBM or something
If it turns out you misread, the table THT created could be off. And with the 1.2GHz figure, that might help create a more realistic table (having two data-points for extrapolation would be better than one). Or you could be very right, I could be wrong, and the table could be pretty close.
Originally posted by kupan787
Then would you mind pointing me to where you read the 1.8GHz 970 would be 47W. THT mentioned that you had provided the 47W figure, so that is why I thought you might work for IBM or something
The 47 Watts comes from IBM's PowerPC Quick Reference Guide. The PPC 970 is on page 9 of 10. It's the only non-estimated number I've seen. Being 5 Watts off is reasonable. The typical 10 to 15% error for estimates.
But one has to note what typical really means, since I'm sure IBM, Motorola, and any other CPU vendor have different instruction mixes used for their "typical" power consumption numbers.
Originally posted by rickag
Another reference to the 47 watt figure.
This is the same reference quide as above only presented in another way.
The numbers in the Performance section are the same as those from the MPF 2002. So could 47W be a typo and we still have to wait for real numbers?
Someone from PowerLogix has been active on this forum talking about future upgrades. What I get from his posts is that they have 7457's at speed's of up to 1.6 Ghz for testing, and they are designing a daughter card for the Cube that would handle 2 of these. He dosnt give much in the way of when he expects to have these chips available for production. Given the history of Motorolla giving chips to Apple first, as much as 6-12 months before upgrade manufacturers get ahold of them, I would imagine that Apple has these chips in hand. They may be waiting on "sustainable" supplies from Moto so they dont relive the "G4 500" fiasco.
Other interesting quotes:
Interestingly, as chip designs advance, power requirements go down and speeds go up...this is what is happening with the transition from 7455 to 7457. The 7457s will draw less power than an equivalent (or even greater) speed 7455.
A 1.6 7457 (when they are available) will draw far less power than a 1.4 7455.
(panphage @ Jul 29 2003, 08:57 PM)
I keep reading about the power gains from the die shrink. But looking at moto's pdfs, the 7455 @1.0Ghz uses a max of 22W while the 7457 @ 1.0Ghz uses the exact same 22W max. That would be a power savings of exactly nothing. And in fact the typical usage of the 7457 @1.0 Ghz is 15.8W while the 7455 at the same speed is rated 15.0W typical. Am I reading these wrong? If I'm not and these numbers are true, a dual 1.6 ain't gonna happen. I mean in a cube.
Basically, that's incorrect info. I'll post something later to update.
No update yet, but it sounds like he does have som, or has access to some data that is different than the pdf the Moto publically released, possibly real world, and is confidant on the power usage of the 7457's.
I expect to see slower 7457s (slower than 1.6 that is), yes. Not sure exactly what speeds yet, we are waiting on Motorola.
If start-up is as fast and easy as we have heard, then its good for everyone, sure MOT gives up some profit, but every day that goes by, the chips they have are worth less as alternatives become more viable.
Let's find out how fast IBM can gear up its new wafer plant at .13u?
MOT may learn a thing or two in the process.