The unified code-base concerns is very much irrelevant, since the code is compiled, and the x86->ARM toolchains are very mature. The high-range TI OMAPs are already at 65nm, and can be expected to be at 45nm at the same time as this chip debuts.
I'll be surprised if the forthcoming Intel chip is cost-competitive with the 2009 OMAPs. I'll also be surprised if it's much better, and flabbergasted if it has anywhere near as capable low-power operation.
The unified code-base concerns is very much irrelevant, since the code is compiled, and the x86->ARM toolchains are very mature. The high-range TI OMAPs are already at 65nm, and can be expected to be at 45nm at the same time as this chip debuts.
I'll be surprised if the forthcoming Intel chip is cost-competitive with the 2009 OMAPs. I'll also be surprised if it's much better, and flabbergasted if it has anywhere near as capable low-power operation.
Have you read the article over at Ars? Hannibal suggests that apps written to run on x86 processors would then be able to run on these embedded chips and will be a powerful force that will make these chips more attractive than other embedded chips, ie ARM.
He predicts that these x86 chips will dominate the embedded space in the years to come. Of course we don't know if apple will even allow non-apple apps on the future iPhones or other similar devices that may be forthcoming but it would seem that Apple are moving and should move in that direction.
I don't know why you would assume that this chip is x86 based. x86 is a big outdated instruction set which is not exactly energy efficent (especially not in intel's hands). And speaking of that, would they be going for 32bit or 64 bit x86? Obviously if Intel is making a SoC with mobile devices in mind they are going to want to create a instruction set which is efficent.
That being said.. as pointed out by others, this chip is still far too large and power consuming to be in an iPhone sized (and powered) device.
I don't know why you would assume that this chip is x86 based. x86 is a big outdated instruction set which is not exactly energy efficent (especially not in intel's hands). And speaking of that, would they be going for 32bit or 64 bit x86? Obviously if Intel is making a SoC with mobile devices in mind they are going to want to create a instruction set which is efficent.
That being said.. as pointed out by others, this chip is still far too large and power consuming to be in an iPhone sized (and powered) device.
All this talk of x86 may be a distraction from broader possibilities. I wonder if it's a reference to the ARM chip that Intel is still making, last I heard. They had a line of ARM chips that they sold, but they were still fabbing them for the company that "bought" the product line. I forget who was making the iPhone's chips, for a while, it was rumored to be Intel.
I don't know why you would assume that this chip is x86 based. x86 is a big outdated instruction set which is not exactly energy efficent (especially not in intel's hands). And speaking of that, would they be going for 32bit or 64 bit x86? Obviously if Intel is making a SoC with mobile devices in mind they are going to want to create a instruction set which is efficent.
That being said.. as pointed out by others, this chip is still far too large and power consuming to be in an iPhone sized (and powered) device.
Because that's what the article at Ars says. Give it a read it's quite interesting.
As far as power consumption is concerned, Ars feels the initial Menlow platform will be capable of 6 hours of usage and 10 hours standby. When Moorestown is delivered it's supposed to be even more power efficient.
Because that's what the article at Ars says. Give it a read it's quite interesting.
As far as power consumption is concerned, Ars feels the initial Menlow platform will be capable of 6 hours of usage and 10 hours standby. When Moorestown is delivered it's supposed to be even more power efficient.
Yes.. Ok.. So x86 it is.. however those devices have very large batteries The iPhone would have to be the size of a small briefcase to get 6 hours of usage.. I'm going on 9 hours and I'm not even at 50% battery
All this talk of x86 may be a distraction from broader possibilities. I wonder if it's a reference to the ARM chip that Intel is still making, last I heard. They had a line of ARM chips that they sold, but they were still fabbing them for the company that "bought" the product line. I forget who was making the iPhone's chips, for a while, it was rumored to be Intel.
Samsung manufactures the iPhones processor. ARM processors are made by many different companies. Hitachi, Intel, Samsung, even IBM was fabbing them for a while.
Yes.. Ok.. So x86 it is.. however those devices have very large batteries The iPhone would have to be the size of a small briefcase to get 6 hours of usage.. I'm going on 9 hours and I'm not even at 50% battery
The graphs are a bit confusing to me but it appears to me that Silverthorn (the cpu in the Menlow platform) has less than 0.55 power usage even under load.
Time will tell if silverthorn and the menlow/moorestown platforms live up to Intel's predictions. But it looks like Intel is going after the UMPC and handheld device market in a big way.
The graphs are a bit confusing to me but it appears to me that Silverthorn (the cpu in the Menlow platform) has less than 0.55 power usage even under load.
Time will tell if silverthorn and the menlow/moorestown platforms live up to Intel's predictions. But it looks like Intel is going after the UMPC and handheld device market in a big way.
Does appear that way, but there is a big difference between this going into a UMPC or even a handheld device and going into a cellphone.. the cellphone has the added power draw of the cellular network.. Far as I see the power useage of these is based on data network usage which is idle when not transfering data.. voice service is generally always trying to make connections with towers (or discover new ones) and uses more power.
Hey.. I'm not anti intel.. just don't think that this is something you're going to see apple using in a phone.. perhaps they have something else up their sleeve.. or they are just looking down the road for the future.
Does appear that way, but there is a big difference between this going into a UMPC or even a handheld device and going into a cellphone.. the cellphone has the added power draw of the cellular network.. Far as I see the power useage of these is based on data network usage which is idle when not transfering data.. voice service is generally always trying to make connections with towers (or discover new ones) and uses more power.
Hey.. I'm not anti intel.. just don't think that this is something you're going to see apple using in a phone.. perhaps they have something else up their sleeve.. or they are just looking down the road for the future.
We didn't think that Apple would go Intel at all. People were still kicking and screaming that it wasn't possible after it happened.
Whatever Intel is going to do, you can be sure that Apple and other large OEM's already know about it.
You might remember that when Jobs showed the chart of Intel's power/performance vs. IBM's, we didn't bel;ieve that either. But, it was true.
The power used by this chip is pretty low, and it includes technologies that now use plenty of power. You might be surprised.
You might remember that when Jobs showed the chart of Intel's power/performance vs. IBM's, we didn't bel;ieve that either. But, it was true.
Actually, I don't think it is true yet. X86 has only just passed the PowerPC in terms of performance. It has been more power efficient, but not the multipliers Jobs was quoting.
These new chips may finally make that so.
Apple also spent some significant R&D on it's ARM LLVM compiler tools which we've not seen yet.
The other problem you've got with the iPhone OS v Mac OS is that from v1.1.1 the binaries are supposedly signed, encrypted and have checksums. One would hope Apple isn't stupid enough to insist on that on MacOS. If they did they'd kill open source and shareware development on MacOS too.
Actually, I don't think it is true yet. X86 has only just passed the PowerPC in terms of performance. It has been more power efficient, but not the multipliers Jobs was quoting.
These new chips may finally make that so.
Apple also spent some significant R&D on it's ARM LLVM compiler tools which we've not seen yet.
The other problem you've got with the iPhone OS v Mac OS is that from v1.1.1 the binaries are supposedly signed, encrypted and have checksums. One would hope Apple isn't stupid enough to insist on that on MacOS. If they did they'd kill open source and shareware development on MacOS too.
I would say we passed PPC performance a while ago, almost from the beginning.
Sure, we've now got 8-core Mac Pros but the old Quad G5 sometimes beats the Quad Intel and vice versa. There's no particular reason why you should feel second class on a late G5.
Further down the line in the laptops and iMac there's been much greater gains, particularly as the G4 was just so rubbish and sticking two cores in certainly makes you a believer. That's been the only real place we've seen PPC surpassed in performance.
Really, I still think Job's performance per watt thing was smoke and mirrors to get past the fact at the high end Intel weren't quite as fast as the graphs would show. If you weren't bothered so much about how much electricity your CPU used, the graphs were silly.
It'll be great if they can get the power down to the 5w and 0.5w levels they were claiming back then without making them less powerful than today's chips but it's a little early to count out the ARM in the iPhone and iPod.
Sure, we've now got 8-core Mac Pros but the old Quad G5 sometimes beats the Quad Intel and vice versa. There's no particular reason why you should feel second class on a late G5.
Really, I still think Job's performance per watt thing was smoke and mirrors to get past the fact at the high end Intel weren't quite as fast as the graphs would show. If you weren't bothered so much about how much electricity your CPU used, the graphs were silly.
What is the power consumption of the G5 quad? Last I heard it was a real heater. I don't remember for sure but it seems like I remember a figure that was something like 500W of actual wall-power draw. Mac Pro runs at around half that. I think both figures are max load, IIRC, Mac Pro at idle is 175W. Not great, but still better than a lot of the G5s.
Sure, we've now got 8-core Mac Pros but the old Quad G5 sometimes beats the Quad Intel and vice versa. There's no particular reason why you should feel second class on a late G5.
Further down the line in the laptops and iMac there's been much greater gains, particularly as the G4 was just so rubbish and sticking two cores in certainly makes you a believer. That's been the only real place we've seen PPC surpassed in performance.
Really, I still think Job's performance per watt thing was smoke and mirrors to get past the fact at the high end Intel weren't quite as fast as the graphs would show. If you weren't bothered so much about how much electricity your CPU used, the graphs were silly.
It'll be great if they can get the power down to the 5w and 0.5w levels they were claiming back then without making them less powerful than today's chips but it's a little early to count out the ARM in the iPhone and iPod.
That one test you peoduced shows the Quad as having an advantage in a very few areas. Those areas could be ascribed to programming as much as machine speed.
For almost everything else, the Intel machines rack up the scores.
As we go forward, we will see even more of that. IBM simply had no economic reason to update the G5 according to Apple's needs. It's too bad, because it was a superior architecture, but no longer.
What is the power consumption of the G5 quad? Last I heard it was a real heater. I don't remember for sure but it seems like I remember a figure that was something like 500W of actual wall-power draw. Mac Pro runs at around half that. I think both figures are max load, IIRC, Mac Pro at idle is 175W. Not great, but still better than a lot of the G5s.
Apple and Intel were talking about tenfold increases in performance per watt back when they switched. At best we're talking about double performance per watt in the Mac Pros but not double the performance. Performance stayed roughly the same as the previous Quads even if power consumption dropped to half.
If what you were after was more performance, Apple's slight of hand didn't deliver.
The G4 and PowerPC clones other than the G5 is oddly still better performance per watt than anything from Intel. It's problem is it's max performance is so poor, not it's performance per watt.
Just to be clear, I'm not saying the switch wasn't a bad thing, just the reason given was economical with the truth. It's clearly been good for Apple in it's core market - laptops and the iMac where the PowerPC was just wrong.
Quote:
Originally Posted by melgross
As we go forward, we will see even more of that. IBM simply had no economic reason to update the G5 according to Apple's needs. It's too bad, because it was a superior architecture, but no longer.
It was never a question of economics, just timescales from what I was told at the time by some IBM guys I know. But hey, history now.
It was never a question of economics, just timescales from what I was told at the time by some IBM guys I know. But hey, history now.
Economics determines timescales. It would have taken IBM much longer to pay off that R&D because of the low number of chips sold. It gets to the point where it doesn't pay at all. What happens is that the R&D gets stretched out to the point that the product no longer is competitive.
Apple and Intel were talking about tenfold increases in performance per watt back when they switched. At best we're talking about double performance per watt in the Mac Pros but not double the performance. Performance stayed roughly the same as the previous Quads even if power consumption dropped to half.
If what you were after was more performance, Apple's slight of hand didn't deliver.
The way I understood the charts, there would be a smaller short term difference but the long term difference would grow over time. Somehow I had the impression that the huge difference was something like five or ten years from the announcement. I don't remember the charts saying anything specifically about raw performance. Most of Apple's products needed low power, high performance chips to work satisfactorily. Personally, I'm happy with a much quieter Mac Pro than I was with either of my G5 towers that I had, and it's a better machine than those things in just about every way.
I wouldn't be surprised if Steve pressure-cooked the numbers too, Apple has tended to take the numbers that present them in the best possible light to an extreme of being less credible to anyone that's not under the RDF.
Comments
I'll be surprised if the forthcoming Intel chip is cost-competitive with the 2009 OMAPs. I'll also be surprised if it's much better, and flabbergasted if it has anywhere near as capable low-power operation.
The unified code-base concerns is very much irrelevant, since the code is compiled, and the x86->ARM toolchains are very mature. The high-range TI OMAPs are already at 65nm, and can be expected to be at 45nm at the same time as this chip debuts.
I'll be surprised if the forthcoming Intel chip is cost-competitive with the 2009 OMAPs. I'll also be surprised if it's much better, and flabbergasted if it has anywhere near as capable low-power operation.
Have you read the article over at Ars? Hannibal suggests that apps written to run on x86 processors would then be able to run on these embedded chips and will be a powerful force that will make these chips more attractive than other embedded chips, ie ARM.
He predicts that these x86 chips will dominate the embedded space in the years to come. Of course we don't know if apple will even allow non-apple apps on the future iPhones or other similar devices that may be forthcoming but it would seem that Apple are moving and should move in that direction.
That being said.. as pointed out by others, this chip is still far too large and power consuming to be in an iPhone sized (and powered) device.
I don't know why you would assume that this chip is x86 based. x86 is a big outdated instruction set which is not exactly energy efficent (especially not in intel's hands). And speaking of that, would they be going for 32bit or 64 bit x86? Obviously if Intel is making a SoC with mobile devices in mind they are going to want to create a instruction set which is efficent.
That being said.. as pointed out by others, this chip is still far too large and power consuming to be in an iPhone sized (and powered) device.
All this talk of x86 may be a distraction from broader possibilities. I wonder if it's a reference to the ARM chip that Intel is still making, last I heard. They had a line of ARM chips that they sold, but they were still fabbing them for the company that "bought" the product line. I forget who was making the iPhone's chips, for a while, it was rumored to be Intel.
I don't know why you would assume that this chip is x86 based. x86 is a big outdated instruction set which is not exactly energy efficent (especially not in intel's hands). And speaking of that, would they be going for 32bit or 64 bit x86? Obviously if Intel is making a SoC with mobile devices in mind they are going to want to create a instruction set which is efficent.
That being said.. as pointed out by others, this chip is still far too large and power consuming to be in an iPhone sized (and powered) device.
Because that's what the article at Ars says. Give it a read it's quite interesting.
As far as power consumption is concerned, Ars feels the initial Menlow platform will be capable of 6 hours of usage and 10 hours standby. When Moorestown is delivered it's supposed to be even more power efficient.
Because that's what the article at Ars says. Give it a read it's quite interesting.
As far as power consumption is concerned, Ars feels the initial Menlow platform will be capable of 6 hours of usage and 10 hours standby. When Moorestown is delivered it's supposed to be even more power efficient.
Yes.. Ok.. So x86 it is.. however those devices have very large batteries
All this talk of x86 may be a distraction from broader possibilities. I wonder if it's a reference to the ARM chip that Intel is still making, last I heard. They had a line of ARM chips that they sold, but they were still fabbing them for the company that "bought" the product line. I forget who was making the iPhone's chips, for a while, it was rumored to be Intel.
Samsung manufactures the iPhones processor. ARM processors are made by many different companies. Hitachi, Intel, Samsung, even IBM was fabbing them for a while.
Yes.. Ok.. So x86 it is.. however those devices have very large batteries
More reading if you're interested.
The graphs are a bit confusing to me but it appears to me that Silverthorn (the cpu in the Menlow platform) has less than 0.55 power usage even under load.
Time will tell if silverthorn and the menlow/moorestown platforms live up to Intel's predictions. But it looks like Intel is going after the UMPC and handheld device market in a big way.
More reading if you're interested.
The graphs are a bit confusing to me but it appears to me that Silverthorn (the cpu in the Menlow platform) has less than 0.55 power usage even under load.
Time will tell if silverthorn and the menlow/moorestown platforms live up to Intel's predictions. But it looks like Intel is going after the UMPC and handheld device market in a big way.
Does appear that way, but there is a big difference between this going into a UMPC or even a handheld device and going into a cellphone.. the cellphone has the added power draw of the cellular network.. Far as I see the power useage of these is based on data network usage which is idle when not transfering data.. voice service is generally always trying to make connections with towers (or discover new ones) and uses more power.
Hey.. I'm not anti intel.. just don't think that this is something you're going to see apple using in a phone.. perhaps they have something else up their sleeve.. or they are just looking down the road for the future.
Does appear that way, but there is a big difference between this going into a UMPC or even a handheld device and going into a cellphone.. the cellphone has the added power draw of the cellular network.. Far as I see the power useage of these is based on data network usage which is idle when not transfering data.. voice service is generally always trying to make connections with towers (or discover new ones) and uses more power.
Hey.. I'm not anti intel.. just don't think that this is something you're going to see apple using in a phone.. perhaps they have something else up their sleeve.. or they are just looking down the road for the future.
We didn't think that Apple would go Intel at all. People were still kicking and screaming that it wasn't possible after it happened.
Whatever Intel is going to do, you can be sure that Apple and other large OEM's already know about it.
You might remember that when Jobs showed the chart of Intel's power/performance vs. IBM's, we didn't bel;ieve that either. But, it was true.
The power used by this chip is pretty low, and it includes technologies that now use plenty of power. You might be surprised.
You might remember that when Jobs showed the chart of Intel's power/performance vs. IBM's, we didn't bel;ieve that either. But, it was true.
Actually, I don't think it is true yet. X86 has only just passed the PowerPC in terms of performance. It has been more power efficient, but not the multipliers Jobs was quoting.
These new chips may finally make that so.
Apple also spent some significant R&D on it's ARM LLVM compiler tools which we've not seen yet.
The other problem you've got with the iPhone OS v Mac OS is that from v1.1.1 the binaries are supposedly signed, encrypted and have checksums. One would hope Apple isn't stupid enough to insist on that on MacOS. If they did they'd kill open source and shareware development on MacOS too.
Actually, I don't think it is true yet. X86 has only just passed the PowerPC in terms of performance. It has been more power efficient, but not the multipliers Jobs was quoting.
These new chips may finally make that so.
Apple also spent some significant R&D on it's ARM LLVM compiler tools which we've not seen yet.
The other problem you've got with the iPhone OS v Mac OS is that from v1.1.1 the binaries are supposedly signed, encrypted and have checksums. One would hope Apple isn't stupid enough to insist on that on MacOS. If they did they'd kill open source and shareware development on MacOS too.
I would say we passed PPC performance a while ago, almost from the beginning.
I would say we passed PPC performance a while ago, almost from the beginning.
Benchmarks would say otherwise.
http://barefeats.com/quad16.html
Sure, we've now got 8-core Mac Pros but the old Quad G5 sometimes beats the Quad Intel and vice versa. There's no particular reason why you should feel second class on a late G5.
Further down the line in the laptops and iMac there's been much greater gains, particularly as the G4 was just so rubbish and sticking two cores in certainly makes you a believer. That's been the only real place we've seen PPC surpassed in performance.
Really, I still think Job's performance per watt thing was smoke and mirrors to get past the fact at the high end Intel weren't quite as fast as the graphs would show. If you weren't bothered so much about how much electricity your CPU used, the graphs were silly.
It'll be great if they can get the power down to the 5w and 0.5w levels they were claiming back then without making them less powerful than today's chips but it's a little early to count out the ARM in the iPhone and iPod.
Benchmarks would say otherwise.
http://barefeats.com/quad16.html
Sure, we've now got 8-core Mac Pros but the old Quad G5 sometimes beats the Quad Intel and vice versa. There's no particular reason why you should feel second class on a late G5.
Really, I still think Job's performance per watt thing was smoke and mirrors to get past the fact at the high end Intel weren't quite as fast as the graphs would show. If you weren't bothered so much about how much electricity your CPU used, the graphs were silly.
What is the power consumption of the G5 quad? Last I heard it was a real heater. I don't remember for sure but it seems like I remember a figure that was something like 500W of actual wall-power draw. Mac Pro runs at around half that. I think both figures are max load, IIRC, Mac Pro at idle is 175W. Not great, but still better than a lot of the G5s.
Benchmarks would say otherwise.
http://barefeats.com/quad16.html
Sure, we've now got 8-core Mac Pros but the old Quad G5 sometimes beats the Quad Intel and vice versa. There's no particular reason why you should feel second class on a late G5.
Further down the line in the laptops and iMac there's been much greater gains, particularly as the G4 was just so rubbish and sticking two cores in certainly makes you a believer. That's been the only real place we've seen PPC surpassed in performance.
Really, I still think Job's performance per watt thing was smoke and mirrors to get past the fact at the high end Intel weren't quite as fast as the graphs would show. If you weren't bothered so much about how much electricity your CPU used, the graphs were silly.
It'll be great if they can get the power down to the 5w and 0.5w levels they were claiming back then without making them less powerful than today's chips but it's a little early to count out the ARM in the iPhone and iPod.
That one test you peoduced shows the Quad as having an advantage in a very few areas. Those areas could be ascribed to programming as much as machine speed.
For almost everything else, the Intel machines rack up the scores.
As we go forward, we will see even more of that. IBM simply had no economic reason to update the G5 according to Apple's needs. It's too bad, because it was a superior architecture, but no longer.
What is the power consumption of the G5 quad? Last I heard it was a real heater. I don't remember for sure but it seems like I remember a figure that was something like 500W of actual wall-power draw. Mac Pro runs at around half that. I think both figures are max load, IIRC, Mac Pro at idle is 175W. Not great, but still better than a lot of the G5s.
Apple and Intel were talking about tenfold increases in performance per watt back when they switched. At best we're talking about double performance per watt in the Mac Pros but not double the performance. Performance stayed roughly the same as the previous Quads even if power consumption dropped to half.
If what you were after was more performance, Apple's slight of hand didn't deliver.
The G4 and PowerPC clones other than the G5 is oddly still better performance per watt than anything from Intel. It's problem is it's max performance is so poor, not it's performance per watt.
Just to be clear, I'm not saying the switch wasn't a bad thing, just the reason given was economical with the truth. It's clearly been good for Apple in it's core market - laptops and the iMac where the PowerPC was just wrong.
As we go forward, we will see even more of that. IBM simply had no economic reason to update the G5 according to Apple's needs. It's too bad, because it was a superior architecture, but no longer.
It was never a question of economics, just timescales from what I was told at the time by some IBM guys I know. But hey, history now.
It was never a question of economics, just timescales from what I was told at the time by some IBM guys I know. But hey, history now.
Economics determines timescales. It would have taken IBM much longer to pay off that R&D because of the low number of chips sold. It gets to the point where it doesn't pay at all. What happens is that the R&D gets stretched out to the point that the product no longer is competitive.
That was what was happening.
Apple and Intel were talking about tenfold increases in performance per watt back when they switched. At best we're talking about double performance per watt in the Mac Pros but not double the performance. Performance stayed roughly the same as the previous Quads even if power consumption dropped to half.
If what you were after was more performance, Apple's slight of hand didn't deliver.
The way I understood the charts, there would be a smaller short term difference but the long term difference would grow over time. Somehow I had the impression that the huge difference was something like five or ten years from the announcement. I don't remember the charts saying anything specifically about raw performance. Most of Apple's products needed low power, high performance chips to work satisfactorily. Personally, I'm happy with a much quieter Mac Pro than I was with either of my G5 towers that I had, and it's a better machine than those things in just about every way.
I wouldn't be surprised if Steve pressure-cooked the numbers too, Apple has tended to take the numbers that present them in the best possible light to an extreme of being less credible to anyone that's not under the RDF.