Anyway, It's not really important to know if the G5 will reach 3 ghz or not (it's only 20 % more performance). The important point, pointed it out by nr9 is that the heat density has reach a limit who need solutions.
Some months ago, there was a rumor saying that IBM and others where studying diamond derived material to extract heat of chip in a very efficient way. Currently this rumor has vanished, the diamond way is perhaps a defective technology and chips designers have to choose an another way.
It's clear that the multicore way, is the future, and that programmation will recquiere some adadptation to benefit of the extra power.
Lets then hope Apple can produce better graphics cards, and OpenGL drivers to compensate for the lack of processor speed to make things faster for some of us.
Um... except faster bus, on-chip mem controller, SSOI instead of SOI, 1mb l2 cache... yah there is a lot that is different... oh yah... both are ppc and 130nm
Aparts from the fact you have 2 things wrong in that list (it has a larger L2 cache and doesn't use SSOI) what there do you picture as being radically new? It certainly isn't a faster bus or every speed bump of the PIV's has made it a radically new chip. Heavens help you if you learn to overclock too. On chip memory controllers are about moving existing silicon and integrating, a job, yes, but nothing radical about it.
Things changed but not substantially. The POWER4's design was refined and improved upon. The POWER5 is not and never has been a completely new design it has always been a refinement of the original POWER4 aimed at using resources better and you really need to only have seen IBMs original briefs to know that one since they emphasised it very strongly.
About the most substantial change was in the form of support for hardware partitioning, which a successor to the 970 won't see.
Those approaches are all well and good but licking the problem at the root is a smarter course of action.
Smarter courses of action aren't necessarily the ones that come to fruition. In the PPC world, the smarter course of action seems to be exception not the rule.
Quote:
New processes and geometries need to be icorporated to lower static power drastically and to address dynamic power.
True. They will. But personal computer power consumption has been on an upward trend since it began. There was a time when 5 Watts was what was budgeted for mobile processors. Then the power budget was 10 Watts, then 20 Watts. It's now 30 to 40 Watts with some desktop replacements using the 80+ Watts desktop processors. This same trend exists for the desktop markets well with 100 Watts being the norm in the next year.
The industry may retreat, but I think they will do what is necessary to make cooling systems be able handle the extra heat from next-gen processors. They will have to because multi-core processors won't alleviate power consumption and heat dissipation much at all.
Quote:
Multi processing serves to in crease performance in other ways, the industry needs to incorporate both advances to keep moving forward. So yes some in the industry hit a speed bump, the smart players are treating it just as that and are recovering and moving forward. I'm not sure what to think of the people who think they hit a wall.
Yes. Agree. I've always thought that multi-core was going to be normal at <90 nm because they would allow 150+ million transistor processors to be economical enough for personal computing, not that clock rate difficulties are forcing CPU manufacturers to use multi-core.
The industry may retreat, but I think they will do what is necessary to make cooling systems be able handle the extra heat from next-gen processors. They will have to because multi-core processors won't alleviate power consumption and heat dissipation much at all.
Yes. Agree. I've always thought that multi-core was going to be normal at <90 nm because they would allow 150+ million transistor processors to be economical enough for personal computing, not that clock rate difficulties are forcing CPU manufacturers to use multi-core.
You are wrong. Why are you so stubborn. Do you work in the CPU industry? Obviously not.
Multiple Core processors will allow us to stop increasing power dissipation / area and just keep adding extra cores. Why don't you get it? Frequency is going to stop goign up. That's a fact.
Multicore processors will completely alleviate the relevant measure of power consumption, which is power density/performance. If you have twice the # of cores, and you measure performance in terms of throughput of multiple threads, this measure of power consumption will be halved.
You are wrong. Why are you so stubborn. Do you work in the CPU industry? Obviously not.
Multiple Core processors will allow us to stop increasing power dissipation / area and just keep adding extra cores. Why don't you get it? Frequency is going to stop goign up. That's a fact.
Multicore processors will completely alleviate the relevant measure of power consumption, which is power density/performance. If you have twice the # of cores, and you measure performance in terms of throughput of multiple threads, this measure of power consumption will be halved.
We all know that CPU's will reach a certain speed and will have to resort to adding cores to it. But, I think it won't be at 2.5 Ghz or 4 Ghz in the Intel world. The more likely area is 6-15 Ghz.
We all know that CPU's will reach a ceratin speed and will have to resort to adding cores to it. But, I think it won't be at 2.5 Ghz or 4 Ghz in the Intel world. The more likely area is 6-15 Ghz.
Well didn't the indrustry think 400 Mhz(or was it around 500 Mhz) was the fastest CPU's could go. Well in went faster. This is the same story. People think 2.5 Ghz G5 is the fastest the PPC can go before adding another core and 4 Ghz on the P4. Trust me one this CPU's will go faster then 4 Ghz including the PPC on a single core. It will be around 6 Ghz when we will hit trouble.
Well didn't the indrustry think 400 Mhz(or was it around 500 Mhz) was the fastest CPU's could go. Well in went faster. This is the same story. People think 2.5 Ghz G5 is the fastest the PPC can go before adding another core and 4 Ghz on the P4. Trust me one this CPU's will go faster then 4 Ghz including the PPC on a single core. It will be around 6 Ghz when we will hit trouble.
How is this the same story, based on what? I haven't been following the thread too closely, and especially since much of this is above me, but the power problems that are popping up with Intel processors and others are causing processor speed increases to nearly halt. It's not like you can just wave a magic wand and all of a sudden processor speeds will start rapidly moving upwards again without any power problems.
Why do you think everyone has dual core on their roadmaps? If it was as simple as just upping frequencies as usual for the next couple years to hit anything close to 6+ GHz, don't you think they'd be doing that?
Nr9 is a knuckle-head. Quad-core 440s eh? Go away, sludge.
Anyway, it isn't necessary the power factor as it is the leakage factor, it will be solved. It is just the situation right now. 90nm is in it's infancy and obviously, so is your acceptance of this fact.
Why do you think everyone has dual core on their roadmaps? If it was as simple as just upping frequencies as usual for the next couple years to hit anything close to 6+ GHz, don't you think they'd be doing that?
Because adding another cpu is a lot easier, cheaper, and efficient than getting a 33%-66% speed increase on single core... Speeds will go over 2.5ghz within the next year. I bet my dual 2.0 RevB Powermac.
How is this the same story, based on what? I haven't been following the thread too closely, and especially since much of this is above me, but the power problems that are popping up with Intel processors and others are causing processor speed increases to nearly halt. It's not like you can just wave a magic wand and all of a sudden processor speeds will start rapidly moving upwards again without any power problems.
Why do you think everyone has dual core on their roadmaps? If it was as simple as just upping frequencies as usual for the next couple years to hit anything close to 6+ GHz, don't you think they'd be doing that?
Back when the G4's were new the industry analysis were saying they wouldn't be able to break a certain MHz speed. I forgot but, I think it was 400 Mhz. Maybe it was the 1 year stall from Moto on the G4. I forgot but, remember the basic story. They were saying because of the heat and the size of the CPU we wouldn't be able to make it smaller and faster. Certainly CPU's went faster. Now, we have this misinformed person preaching we won't be able to break 2.5 Ghz on the PPC and 4 Ghz on the P4 because of the heat and size. Most likely CPU's will go faster. I don't know how much faster we can push it but, certainly more then 2.5 Ghz and 4 Ghz. I say we won't be able to push the CPU more then 15 Ghz. But, that is my guess. But, I say the window of the fastest single core CPU will be between 6-15 Ghz. Then we will be adding cores to make the CPU faster.
Well didn't the indrustry think 400 Mhz(or was it around 500 Mhz) was the fastest CPU's could go. Well in went faster. This is the same story. People think 2.5 Ghz G5 is the fastest the PPC can go before adding another core and 4 Ghz on the P4. Trust me one this CPU's will go faster then 4 Ghz including the PPC on a single core. It will be around 6 Ghz when we will hit trouble.
I tend to agree with quagmire. Just because you can not see the future does not mean it is not there.
Nr9, reminds me of chicken little.
If memory serves me correct he was the one that told everyone the sky was falling.
I think research, and technology will push on, and the door to higher clock cycles will be opened once again in the near future.
Until that time I think Apple may have a head start on optimizing their computers performance by other means weather it be multiprocessing, clustering, code optimization, and/or utilizing the GPU to increase performance.
The only set back is if they intend to use the GPU to strengthen system performance they have to get a grip on OpenGL for once, and also start providing better graphics card solutions. If they don't they will remain dead in the water because the X86 side totally outperforms in these categories already, and once they (x86 side) start to utilize the same things Apple will look all the worse.
[edit]
Which brings me back to something I said in an older thread about Apple needing to change the way the Mac graphics cards are, and develop a motherboard that will enable plug, and play with PC graphics cards.
I think it's time they (apple) hire a new OpenGL team dedicated to porting PC graphics drivers to start putting them in the system updates as they become available. Also, incorporate a PC graphics card compatibility list on their web site, and just keep adding cards to the list as the team finishes the drivers.
Something needs to be done in this area if Apple intends not only to survive, but to compete which is what they should be striving for.
[second edit]
Then again. It shouldn't be Apples job in it's entirety to supply graphics drivers for Macs. Nvidia, and ATI should both participate if Apple opens up to a more standard graphics card system.
You are wrong. Why are you so stubborn. Do you work in the CPU industry? Obviously not.
Obviously not. But what does it matter. This is a message board consisting of Apple fans, and we're all allowed to participate.
Quote:
Multiple Core processors will allow us to stop increasing power dissipation / area and just keep adding extra cores. Why don't you get it? Frequency is going to stop goign up. That's a fact.
That's a prediction. It won't be a fact until companies give up trying to do it (on CMOS based processes). I think everyone has another GHz or so before it becomes impossible.
Quote:
Multicore processors will completely alleviate the relevant measure of power consumption, which is power density/performance. If you have twice the # of cores, and you measure performance in terms of throughput of multiple threads, this measure of power consumption will be halved.
Single threaded performance is still too attractive for companies to stop trying to produce higher clock rate cores, and why would I want to limit performance benchmarking to just multi-threaded performance?
I'm a techno-optimist, I think companies and users are perfectly willing to put up with using higher performance cooling systems in higher power computers for the same cost as their prior systems.
But, that is my guess. But, I say the window of the fastest single core CPU will be between 6-15 Ghz. Then we will be adding cores to make the CPU faster.
Except that the major companies have dual core in their plans over the next 12-18 months. This isn't something coming way down the line, it's in the plans now. If they really thought they could move to those high speeds in the next 2-3 years, they wouldn't have roadmaps showing dual core.
Except that the major companies have dual core in their plans over the next 12-18 months. This isn't something coming way down the line, it's in the plans now. If they really thought they could move to those high speeds in the next 2-3 years, they wouldn't have roadmaps showing dual core.
That doesn't mean they will not make each core faster. I was trying to say is that we will hit the final speed in a core. Which I think is 6-15 Ghz. Then we will only have the technique of adding cores to make the CPU faster. But, then we wouldn't be able to add anymore cores. What then? Only the future can tell us.
Based on several companies planning dual core CPUs, some folks are trying to make a case that this means we will not see higher clock rates. It ain't necessarily so. There can be other reasons for seeing dual core chips now. Here is one possibility:
Each time the chip process got smaller in the past, more transistors were added to improve performance, adding features like SMT. With the latest change to 90nm, the best use for those extra transistors may now be a second core. It can be that simple. Do you have other features you would like to see added? If there is a better use for the higher transistor count I'd like to hear it.
Comments
Some months ago, there was a rumor saying that IBM and others where studying diamond derived material to extract heat of chip in a very efficient way. Currently this rumor has vanished, the diamond way is perhaps a defective technology and chips designers have to choose an another way.
It's clear that the multicore way, is the future, and that programmation will recquiere some adadptation to benefit of the extra power.
I truly belive in muticore future.
Originally posted by Nr9
plans change
But people don't, apparently...
Originally posted by emig647
Um... except faster bus, on-chip mem controller, SSOI instead of SOI, 1mb l2 cache... yah there is a lot that is different... oh yah... both are ppc and 130nm
Aparts from the fact you have 2 things wrong in that list (it has a larger L2 cache and doesn't use SSOI) what there do you picture as being radically new? It certainly isn't a faster bus or every speed bump of the PIV's has made it a radically new chip. Heavens help you if you learn to overclock too. On chip memory controllers are about moving existing silicon and integrating, a job, yes, but nothing radical about it.
Things changed but not substantially. The POWER4's design was refined and improved upon. The POWER5 is not and never has been a completely new design it has always been a refinement of the original POWER4 aimed at using resources better and you really need to only have seen IBMs original briefs to know that one since they emphasised it very strongly.
About the most substantial change was in the form of support for hardware partitioning, which a successor to the 970 won't see.
Also I was almost positive that the Power5 uses SSOI... that was one of the reasons they gained so much more out of it....
Originally posted by wizard69
Those approaches are all well and good but licking the problem at the root is a smarter course of action.
Smarter courses of action aren't necessarily the ones that come to fruition. In the PPC world, the smarter course of action seems to be exception not the rule.
New processes and geometries need to be icorporated to lower static power drastically and to address dynamic power.
True. They will. But personal computer power consumption has been on an upward trend since it began. There was a time when 5 Watts was what was budgeted for mobile processors. Then the power budget was 10 Watts, then 20 Watts. It's now 30 to 40 Watts with some desktop replacements using the 80+ Watts desktop processors. This same trend exists for the desktop markets well with 100 Watts being the norm in the next year.
The industry may retreat, but I think they will do what is necessary to make cooling systems be able handle the extra heat from next-gen processors. They will have to because multi-core processors won't alleviate power consumption and heat dissipation much at all.
Multi processing serves to in crease performance in other ways, the industry needs to incorporate both advances to keep moving forward. So yes some in the industry hit a speed bump, the smart players are treating it just as that and are recovering and moving forward. I'm not sure what to think of the people who think they hit a wall.
Yes. Agree. I've always thought that multi-core was going to be normal at <90 nm because they would allow 150+ million transistor processors to be economical enough for personal computing, not that clock rate difficulties are forcing CPU manufacturers to use multi-core.
Originally posted by THT
The industry may retreat, but I think they will do what is necessary to make cooling systems be able handle the extra heat from next-gen processors. They will have to because multi-core processors won't alleviate power consumption and heat dissipation much at all.
Yes. Agree. I've always thought that multi-core was going to be normal at <90 nm because they would allow 150+ million transistor processors to be economical enough for personal computing, not that clock rate difficulties are forcing CPU manufacturers to use multi-core.
You are wrong. Why are you so stubborn. Do you work in the CPU industry? Obviously not.
Multiple Core processors will allow us to stop increasing power dissipation / area and just keep adding extra cores. Why don't you get it? Frequency is going to stop goign up. That's a fact.
Multicore processors will completely alleviate the relevant measure of power consumption, which is power density/performance. If you have twice the # of cores, and you measure performance in terms of throughput of multiple threads, this measure of power consumption will be halved.
Originally posted by Nr9
You are wrong. Why are you so stubborn. Do you work in the CPU industry? Obviously not.
Multiple Core processors will allow us to stop increasing power dissipation / area and just keep adding extra cores. Why don't you get it? Frequency is going to stop goign up. That's a fact.
Multicore processors will completely alleviate the relevant measure of power consumption, which is power density/performance. If you have twice the # of cores, and you measure performance in terms of throughput of multiple threads, this measure of power consumption will be halved.
We all know that CPU's will reach a certain speed and will have to resort to adding cores to it. But, I think it won't be at 2.5 Ghz or 4 Ghz in the Intel world. The more likely area is 6-15 Ghz.
Originally posted by quagmire
We all know that CPU's will reach a ceratin speed and will have to resort to adding cores to it. But, I think it won't be at 2.5 Ghz or 4 Ghz in the Intel world. The more likely area is 6-15 Ghz.
Well put.
Originally posted by quagmire
Well didn't the indrustry think 400 Mhz(or was it around 500 Mhz) was the fastest CPU's could go. Well in went faster. This is the same story. People think 2.5 Ghz G5 is the fastest the PPC can go before adding another core and 4 Ghz on the P4. Trust me one this CPU's will go faster then 4 Ghz including the PPC on a single core. It will be around 6 Ghz when we will hit trouble.
How is this the same story, based on what? I haven't been following the thread too closely, and especially since much of this is above me, but the power problems that are popping up with Intel processors and others are causing processor speed increases to nearly halt. It's not like you can just wave a magic wand and all of a sudden processor speeds will start rapidly moving upwards again without any power problems.
Why do you think everyone has dual core on their roadmaps? If it was as simple as just upping frequencies as usual for the next couple years to hit anything close to 6+ GHz, don't you think they'd be doing that?
An overview on the power problems is over at Ars:
http://arstechnica.com/cpu/004/presc...rescott-1.html
Anyway, it isn't necessary the power factor as it is the leakage factor, it will be solved. It is just the situation right now. 90nm is in it's infancy and obviously, so is your acceptance of this fact.
Originally posted by MCQ
Why do you think everyone has dual core on their roadmaps? If it was as simple as just upping frequencies as usual for the next couple years to hit anything close to 6+ GHz, don't you think they'd be doing that?
Because adding another cpu is a lot easier, cheaper, and efficient than getting a 33%-66% speed increase on single core... Speeds will go over 2.5ghz within the next year. I bet my dual 2.0 RevB Powermac.
Originally posted by MCQ
How is this the same story, based on what? I haven't been following the thread too closely, and especially since much of this is above me, but the power problems that are popping up with Intel processors and others are causing processor speed increases to nearly halt. It's not like you can just wave a magic wand and all of a sudden processor speeds will start rapidly moving upwards again without any power problems.
Why do you think everyone has dual core on their roadmaps? If it was as simple as just upping frequencies as usual for the next couple years to hit anything close to 6+ GHz, don't you think they'd be doing that?
An overview on the power problems is over at Ars:
http://arstechnica.com/cpu/004/presc...rescott-1.html
Back when the G4's were new the industry analysis were saying they wouldn't be able to break a certain MHz speed. I forgot but, I think it was 400 Mhz. Maybe it was the 1 year stall from Moto on the G4. I forgot but, remember the basic story. They were saying because of the heat and the size of the CPU we wouldn't be able to make it smaller and faster. Certainly CPU's went faster. Now, we have this misinformed person preaching we won't be able to break 2.5 Ghz on the PPC and 4 Ghz on the P4 because of the heat and size. Most likely CPU's will go faster. I don't know how much faster we can push it but, certainly more then 2.5 Ghz and 4 Ghz. I say we won't be able to push the CPU more then 15 Ghz. But, that is my guess. But, I say the window of the fastest single core CPU will be between 6-15 Ghz. Then we will be adding cores to make the CPU faster.
Originally posted by quagmire
Well didn't the indrustry think 400 Mhz(or was it around 500 Mhz) was the fastest CPU's could go. Well in went faster. This is the same story. People think 2.5 Ghz G5 is the fastest the PPC can go before adding another core and 4 Ghz on the P4. Trust me one this CPU's will go faster then 4 Ghz including the PPC on a single core. It will be around 6 Ghz when we will hit trouble.
I tend to agree with quagmire. Just because you can not see the future does not mean it is not there.
Nr9, reminds me of chicken little.
If memory serves me correct he was the one that told everyone the sky was falling.
I think research, and technology will push on, and the door to higher clock cycles will be opened once again in the near future.
Until that time I think Apple may have a head start on optimizing their computers performance by other means weather it be multiprocessing, clustering, code optimization, and/or utilizing the GPU to increase performance.
The only set back is if they intend to use the GPU to strengthen system performance they have to get a grip on OpenGL for once, and also start providing better graphics card solutions. If they don't they will remain dead in the water because the X86 side totally outperforms in these categories already, and once they (x86 side) start to utilize the same things Apple will look all the worse.
[edit]
Which brings me back to something I said in an older thread about Apple needing to change the way the Mac graphics cards are, and develop a motherboard that will enable plug, and play with PC graphics cards.
I think it's time they (apple) hire a new OpenGL team dedicated to porting PC graphics drivers to start putting them in the system updates as they become available. Also, incorporate a PC graphics card compatibility list on their web site, and just keep adding cards to the list as the team finishes the drivers.
Something needs to be done in this area if Apple intends not only to survive, but to compete which is what they should be striving for.
[second edit]
Then again. It shouldn't be Apples job in it's entirety to supply graphics drivers for Macs. Nvidia, and ATI should both participate if Apple opens up to a more standard graphics card system.
Originally posted by Nr9
You are wrong. Why are you so stubborn. Do you work in the CPU industry? Obviously not.
Obviously not. But what does it matter. This is a message board consisting of Apple fans, and we're all allowed to participate.
Multiple Core processors will allow us to stop increasing power dissipation / area and just keep adding extra cores. Why don't you get it? Frequency is going to stop goign up. That's a fact.
That's a prediction. It won't be a fact until companies give up trying to do it (on CMOS based processes). I think everyone has another GHz or so before it becomes impossible.
Multicore processors will completely alleviate the relevant measure of power consumption, which is power density/performance. If you have twice the # of cores, and you measure performance in terms of throughput of multiple threads, this measure of power consumption will be halved.
Single threaded performance is still too attractive for companies to stop trying to produce higher clock rate cores, and why would I want to limit performance benchmarking to just multi-threaded performance?
I'm a techno-optimist, I think companies and users are perfectly willing to put up with using higher performance cooling systems in higher power computers for the same cost as their prior systems.
Originally posted by quagmire
But, that is my guess. But, I say the window of the fastest single core CPU will be between 6-15 Ghz. Then we will be adding cores to make the CPU faster.
Except that the major companies have dual core in their plans over the next 12-18 months. This isn't something coming way down the line, it's in the plans now. If they really thought they could move to those high speeds in the next 2-3 years, they wouldn't have roadmaps showing dual core.
http://www.anandtech.com/tradeshows/...spx?i=2203&p=1
Originally posted by MCQ
Except that the major companies have dual core in their plans over the next 12-18 months. This isn't something coming way down the line, it's in the plans now. If they really thought they could move to those high speeds in the next 2-3 years, they wouldn't have roadmaps showing dual core.
http://www.anandtech.com/tradeshows/...spx?i=2203&p=1
That doesn't mean they will not make each core faster. I was trying to say is that we will hit the final speed in a core. Which I think is 6-15 Ghz. Then we will only have the technique of adding cores to make the CPU faster. But, then we wouldn't be able to add anymore cores. What then? Only the future can tell us.
Each time the chip process got smaller in the past, more transistors were added to improve performance, adding features like SMT. With the latest change to 90nm, the best use for those extra transistors may now be a second core. It can be that simple. Do you have other features you would like to see added? If there is a better use for the higher transistor count I'd like to hear it.