I dont understand this, Pentium 4 was what, 3.0+ Ghz. and Now all intel is making is under 2.5. I know its dual core now, but I thought this new techology could break the barriers and reach 4.0+ speeds?
It seems like there has been a shift from clock speed to higher cache and FSB speeds, as well as multiple cores. In essence, a higher cache and FSB negate the need for higher clock speed. That's my take on it, someone with more knowledge will probably correct me.
I think the most general explanation is that previous processors like the Pentium 4 achieved high clock speeds at the expense of efficiency (ie: less work done per clock cycle). The newer processors concentrate on performance efficiency and power consumption, at the expense of clock speed.
Yup, this was the MHZ Myth that Apple battled for years. The Pentiums were designed to be insanely fast on the cycling, but take a lot of cycles to do much. PowerPCs were designed to do a lot per cycle, but the clock was slower.
The Core designs are closer to the PPC approach in this.
Yeah, it's funny how "the megahertz myth" got widely derided in PC circles as RDF but now that Intel has switched gears the topic has just been dropped like a red-hot Pentium.
And, um ironic, or something, that it only happened after Apple switched, so we don't even get a proper "nyah, nyah nyah" out of it.
Oh, hell, IBM/Moto progress was so glacial there probably wasn't any "nyah, nyah, nyah" left in the thing anyway.
I've heard before that you can (at least in the days of the Pentium M) multiple the clock speed by 1.7 to see what it pretty much equals in the old Mhz mentality. I don't know how true that is or not. But if you apply that logic to say a 2Ghz Core Duo you get 3.4Ghz (old school ratings) x 2 cores. If that makes you feel any better.
Yeah there I go spreading stuff around
In truth though the new chips are faster but run at slow clock speeds. That's all that really matters.
The megahertz myth was never really a myth, just one way of measureing a chips performance. Unfortunately too much weight was put on that one measurement and it became a marketing term. From there they had to play to the market based on what the market understood. Once they hit a barrier both companies pretty had to much come clean and focus on a different way of doing things. Intel came up with an "interesting" chip naming convention where most people just chose to call them by their true Mhz speed (which is indeed slower then before) and AMD uses a "presumed" MHz rating (2000+, 3100+, and so on) to try and relate them to the old school measurement of MHz race.
I've heard before that you can (at least in the days of the Pentium M) multiple the clock speed by 1.7 to see what it pretty much equals in the old Mhz mentality. I don't know how true that is or not. But if you apply that logic to say a 2Ghz Core Duo you get 3.4Ghz (old school ratings) x 2 cores. If that makes you feel any better.
Intel came up with an "interesting" chip naming convention where most people just chose to call them by their true Mhz speed (which is indeed slower then before) and AMD uses a "presumed" MHz rating (2000+, 3100+, and so on) to try and relate them to the old school measurement of MHz race.
The rule of thumb I used was about 1.5, in relating Pentium M or AMD clocks to P4 clocks.
AMD has dropped their processor rating and moved to model numbers. Processor ratings only meant something as a comparison against the prevailing Intel product, but Intel switched to model numbers.
I don't know if that's helped though. Model numbers admit that there's more to it than clock but now that Netburst is going away, they are all a lot more comparable now, clock for clock.
As I posted in another topic, is the race to more and more cores the new Mhz myth? In the present term anyway, until the majority of programs are SMP aware? Does the average consumer really need 8 or 16 cores on a single die?
I'm sure the things I do (iLife, internet, school work, gaming, occasional movies) would barely tax a Dual G5, much less a Quad or Octo Intel! \
As I posted in another topic, is the race to more and more cores the new Mhz myth? In the present term anyway, until the majority of programs are SMP aware? Does the average consumer really need 8 or 16 cores on a single die?
I'm sure the things I do (iLife, internet, school work, gaming, occasional movies) would barely tax a Dual G5, much less a Quad or Octo Intel! \
Low end machines probably won't need anything more than dual for a while. I do hope that more software becomes more multi-core savvy soon.
As I posted in another topic, is the race to more and more cores the new Mhz myth? In the present term anyway, until the majority of programs are SMP aware? Does the average consumer really need 8 or 16 cores on a single die?
I'm sure the things I do (iLife, internet, school work, gaming, occasional movies) would barely tax a Dual G5, much less a Quad or Octo Intel! \
So when you render your final movie in iMovie it happens instantaneously then? No? Then there's still a need for faster chips with more cores.
Is iPhoto "faster" enough yet? I still run into spinning icons at startup and progress bars after editing my 6.3MP RAW images.
So when you render your final movie in iMovie it happens instantaneously then? No? Then there's still a need for faster chips with more cores.
Is iPhoto "faster" enough yet? I still run into spinning icons at startup and progress bars after editing my 6.3MP RAW images.
iMovie currently won't scale beyond two cores, many actions will only use the equivalent of one core. That's a software problem that I was hoping they would fix for iLife '06, especially when they knew that their entire computer line-up would be dual-core. So faster CPUs and more cores won't help the consumer unless someone bothers to make consumer software that scales. iPhoto's editing can scale, but I think that will mean that they will have to move away from the idea of using the graphics processor. Assuming breaking down the tasks for multiple CPUs net faster results. It would be nice to have it both ways so the program can choose the quickest based on what the user has.
Program startup time is more a function of the hard drive than CPU power. A faster processor helps a little bit, but I doubt more cores would help that.
The main problem is that as frequency rises, inductor reactance becomes a problem. As chips get larger, this makes chip design especially difficult as designers struggle to minimize the fact that every wire in the the chip, and especially every loop, is behaving like an antenna.
We'll have to wait until superconductors and nanotube switches hit the mainstream before we'll see another jump in available clock frequencies. By the time that happens, though, it's entirely possible that CPUs will be predominantly asynchronous.
Yup, this was the MHZ Myth that Apple battled for years. The Pentiums were designed to be insanely fast on the cycling, but take a lot of cycles to do much. PowerPCs were designed to do a lot per cycle, but the clock was slower.
The Core designs are closer to the PPC approach in this.
The problem was, the pentiums were almost always faster even with the mac's greater performance/cycle.
iMovie currently won't scale beyond two cores, many actions will only use the equivalent of one core. That's a software problem that I was hoping they would fix for iLife '06, especially when they knew that their entire computer line-up would be dual-core. So faster CPUs and more cores won't help the consumer unless someone bothers to make consumer software that scales. iPhoto's editing can scale, but I think that will mean that they will have to move away from the idea of using the graphics processor. Assuming breaking down the tasks for multiple CPUs net faster results.
The problem is that some software can't be broken down easily. I mean, you can only multithread things that have to happen at the same time. You can't have Core A working on something and Core B working on something that requires Core A's answer (if that makes sense). Now I agree that programmers can get on top of that and do a lot, there's some stuff that can't be multithreaded.
Also, with multiple cores, FSBs and chipsets become an issue. Having two cores moving down one FSB is OK if the FSB is really fast, but when we move to 4 cores and the FSB doesn't get any larger, you're in trouble because you can't get the processors information all that fast. the solution is to do what AMD does and move to hypertransport or other point-to-point techs, because it's amazingly high-bandwidth.
Comments
The Core designs are closer to the PPC approach in this.
And, um ironic, or something, that it only happened after Apple switched, so we don't even get a proper "nyah, nyah nyah" out of it.
Oh, hell, IBM/Moto progress was so glacial there probably wasn't any "nyah, nyah, nyah" left in the thing anyway.
Yeah there I go spreading stuff around
In truth though the new chips are faster but run at slow clock speeds. That's all that really matters.
The megahertz myth was never really a myth, just one way of measureing a chips performance. Unfortunately too much weight was put on that one measurement and it became a marketing term. From there they had to play to the market based on what the market understood. Once they hit a barrier both companies pretty had to much come clean and focus on a different way of doing things. Intel came up with an "interesting" chip naming convention where most people just chose to call them by their true Mhz speed (which is indeed slower then before) and AMD uses a "presumed" MHz rating (2000+, 3100+, and so on) to try and relate them to the old school measurement of MHz race.
I've heard before that you can (at least in the days of the Pentium M) multiple the clock speed by 1.7 to see what it pretty much equals in the old Mhz mentality. I don't know how true that is or not. But if you apply that logic to say a 2Ghz Core Duo you get 3.4Ghz (old school ratings) x 2 cores. If that makes you feel any better.
Intel came up with an "interesting" chip naming convention where most people just chose to call them by their true Mhz speed (which is indeed slower then before) and AMD uses a "presumed" MHz rating (2000+, 3100+, and so on) to try and relate them to the old school measurement of MHz race.
The rule of thumb I used was about 1.5, in relating Pentium M or AMD clocks to P4 clocks.
AMD has dropped their processor rating and moved to model numbers. Processor ratings only meant something as a comparison against the prevailing Intel product, but Intel switched to model numbers.
I don't know if that's helped though. Model numbers admit that there's more to it than clock but now that Netburst is going away, they are all a lot more comparable now, clock for clock.
I'm sure the things I do (iLife, internet, school work, gaming, occasional movies) would barely tax a Dual G5, much less a Quad or Octo Intel!
As I posted in another topic, is the race to more and more cores the new Mhz myth? In the present term anyway, until the majority of programs are SMP aware? Does the average consumer really need 8 or 16 cores on a single die?
I'm sure the things I do (iLife, internet, school work, gaming, occasional movies) would barely tax a Dual G5, much less a Quad or Octo Intel!
Low end machines probably won't need anything more than dual for a while. I do hope that more software becomes more multi-core savvy soon.
As I posted in another topic, is the race to more and more cores the new Mhz myth? In the present term anyway, until the majority of programs are SMP aware? Does the average consumer really need 8 or 16 cores on a single die?
I'm sure the things I do (iLife, internet, school work, gaming, occasional movies) would barely tax a Dual G5, much less a Quad or Octo Intel!
So when you render your final movie in iMovie it happens instantaneously then? No? Then there's still a need for faster chips with more cores.
Is iPhoto "faster" enough yet? I still run into spinning icons at startup and progress bars after editing my 6.3MP RAW images.
So when you render your final movie in iMovie it happens instantaneously then? No? Then there's still a need for faster chips with more cores.
Is iPhoto "faster" enough yet? I still run into spinning icons at startup and progress bars after editing my 6.3MP RAW images.
iMovie currently won't scale beyond two cores, many actions will only use the equivalent of one core. That's a software problem that I was hoping they would fix for iLife '06, especially when they knew that their entire computer line-up would be dual-core. So faster CPUs and more cores won't help the consumer unless someone bothers to make consumer software that scales. iPhoto's editing can scale, but I think that will mean that they will have to move away from the idea of using the graphics processor. Assuming breaking down the tasks for multiple CPUs net faster results. It would be nice to have it both ways so the program can choose the quickest based on what the user has.
Program startup time is more a function of the hard drive than CPU power. A faster processor helps a little bit, but I doubt more cores would help that.
We'll have to wait until superconductors and nanotube switches hit the mainstream before we'll see another jump in available clock frequencies. By the time that happens, though, it's entirely possible that CPUs will be predominantly asynchronous.
Yup, this was the MHZ Myth that Apple battled for years. The Pentiums were designed to be insanely fast on the cycling, but take a lot of cycles to do much. PowerPCs were designed to do a lot per cycle, but the clock was slower.
The Core designs are closer to the PPC approach in this.
The problem was, the pentiums were almost always faster even with the mac's greater performance/cycle.
iMovie currently won't scale beyond two cores, many actions will only use the equivalent of one core. That's a software problem that I was hoping they would fix for iLife '06, especially when they knew that their entire computer line-up would be dual-core. So faster CPUs and more cores won't help the consumer unless someone bothers to make consumer software that scales. iPhoto's editing can scale, but I think that will mean that they will have to move away from the idea of using the graphics processor. Assuming breaking down the tasks for multiple CPUs net faster results.
The problem is that some software can't be broken down easily. I mean, you can only multithread things that have to happen at the same time. You can't have Core A working on something and Core B working on something that requires Core A's answer (if that makes sense). Now I agree that programmers can get on top of that and do a lot, there's some stuff that can't be multithreaded.
Also, with multiple cores, FSBs and chipsets become an issue. Having two cores moving down one FSB is OK if the FSB is really fast, but when we move to 4 cores and the FSB doesn't get any larger, you're in trouble because you can't get the processors information all that fast. the solution is to do what AMD does and move to hypertransport or other point-to-point techs, because it's amazingly high-bandwidth.