...The brilliant part is the tech press at large doesn't even see, which means they fail to comprehend, the end around that Apple's doing with Intel.
Well, in the mobile space Intel can pretty much give up on Apple using their tech. The tablet if released this year may use Atom though. Don't be too optimistic, nonetheless for Macs Intel will be there for at least another good 3 to 5 years.
By the way, where I am right now (Malaysia) iPhone 3G[S] is listed for a July launch... So, can't wait. I don't need the "snappiness" necessarily but the photo and video improvements will be good. Not a big mobile gamer at all BTW. In any case within two months it's time for my current iPhone 3G 8GB to go to a good home. Also around the time the warranty runs out, and, touch wood, no problems since Day 1.
One more thing... Is there OpenCL on the iPhone 3G/S ???
...laugh out loud @ apple taking on Palm. try the other way around.
Palm's international push won't be as strong as the iPhone. Since global telcos are now heavily invested in rolling out the next sexy iPhone to continue generating interest in people using their networks. Even during the recession.
I think Atom + Tegra is more likely for the next AppleTV.
Atom and Tegra are mutually exclusive, as Tegra is a system-on-a-chip with dual ARM11 cores and a low(ish) power nvidia GPU. Do you mean the ION/9400 platform perhaps? The dual-core Atom 330 and Ion would be good. but even that is overkill. There are many embedded 1080P decoders available that can be integrated into a 900-1000mhz Cortex-A8 based system-on-a-chip. The AppleTV could be the size of a deck of cards and still run great!
Quote:
Originally Posted by GRevolution
If you ask me, Apple would be more future oriented with this thing, I think their using at least an equivalent SGX 530, though ideally they'd use the latest SGX 543 and the VXD for video playback. Also, why not use the much better A9? It's been available for at least a year. ...
It doesn't work that way. The embedded market moves far slower than PCs. Even when a processing core is finished, it has to go through extensive testing and verification. And then that's just the CPU core. Next, it has to be integrated with a handful of other components into a system-on-a-chip from one of the major manufacturers, and then that whole chip has to be tested, debugged, and validated. Finally, the system-on-a-chip has to be integrated with another handful of components into a full smartphone platform, and then tested/validated/debugged. All of this before a single product gets out.
The Cortex-A8 core was introduced in October 2005 and only now finding its way to production smartphones. Similarly, the multi-core Cortex-A9 architecture was finalized a long time ago, but companies using it like Texas Instruments (OMAP4) and Qualcomm (future Snapdragon) aren't even sampling chips yet. Last I heard, OMAP4 samples in Q3/Q4 2009 and will be in volume production H2 2010.
Also, The Cortex-A8 chip in the iPhone 3GS is 65nm.. A power efficient dual-core Cortex-A9 will need to be in 45nm to get a decent battery life.
And even if they could, why would they jump so far ahead when what they have now is so far ahead of the competition. A Super-iPhone running a dual-core Cortex-A9 running at 1.0+ Ghz and PowerVR SGX40+ would also create major platform problems with regards to making games and applications compatible with the legacy iPhones and iPods.
Atom and Tegra are mutually exclusive, as Tegra is a system-on-a-chip with dual ARM11 cores and a low(ish) power nvidia GPU. Do you mean the ION/9400 platform perhaps? The dual-core Atom 330 and Ion would be good. but even that is overkill. There are many embedded 1080P decoders available that can be integrated into a 900-1000mhz Cortex-A8 based system-on-a-chip. The AppleTV could be the size of a deck of cards and still run great!
.
i think he did mean atom + ion. I've seen people mix up the tegra and ion chips a lot lately.
It doesn't work that way. The embedded market moves far slower than PCs. Even when a processing core is finished, it has to go through extensive testing and verification. And then that's just the CPU core. Next, it has to be integrated with a handful of other components into a system-on-a-chip from one of the major manufacturers, and then that whole chip has to be tested, debugged, and validated. Finally, the system-on-a-chip has to be integrated with another handful of components into a full smartphone platform, and then tested/validated/debugged. All of this before a single product gets out.
to making games and applications compatible with the legacy iPhones and iPods.
This is exactly why I don't see Intel or Nvidia making any inroads to iPhone/iPods/ Apple TV etc.
The amount of work this it takes to design and debug SoC designs is what keeps so many vendors out the game and buying components "off the shelf" but Apple has already signaled their intentions to go this route with the PA Semi acquisition and acquisition of IBM and AMD talent.
If there was ever a time for Apple to eat its own dogfood ..now is that time.
This is exactly why I don't see Intel or Nvidia making any inroads to iPhone/iPods/ Apple TV etc.
The amount of work this it takes to design and debug SoC designs is what keeps so many vendors out the game and buying components "off the shelf" but Apple has already signaled their intentions to go this route with the PA Semi acquisition and acquisition of IBM and AMD talent.
If there was ever a time for Apple to eat its own dogfood ..now is that time.
Given that Intel is not your average company and has built ARM products in the past I don't see why you expect Intel not able to make inroads if they really really wanted to.
Especially given they already run in the AppleTV.
It seems that they really really want to but are taking their time about it. They aren't competing against ARM until at least Pineview/PineTrail (SoC) and that will go against the A9.
Intel is also going to have a process advantage over most other folks. If I remember right Atom is going 32nm this year.
But the MID market is DOA so these will go into netbooks and smartphones. Maybe an Apple tablet...
Given that Intel is not your average company and has built ARM products in the past I don't see why you expect Intel not able to make inroads if they really really wanted to.
Especially given they already run in the AppleTV.
It seems that they really really want to but are taking their time about it. They aren't competing against ARM until at least Pineview/PineTrail (SoC) and that will go against the A9.
Intel is also going to have a process advantage over most other folks. If I remember right Atom is going 32nm this year.
But the MID market is DOA so these will go into netbooks and smartphones. Maybe an Apple tablet...
Intel the company that said "clockspeed is king" and then promptly cancelled Tejas after spending millions?
Intel the company that yammered about 60" LCOS displays for a grand and then promptly cancelled their development?
Intel the company that has announced the Atom which isn't close to be a mobile platform. Hell the 945 bridge eats up wattage like it has prader-willi ?
Intel the company that has been talking about delivering a decent GPU for years and the best they can do is the awful GMA 4xxx crap?
I don't think the process matters as much as the overall design. Intel is simply not a player in the ultra low wattage arena they are the plucky challenger hoping to unseat ARM and not the other way around.
Intel the company that said "clockspeed is king" and then promptly cancelled Tejas after spending millions?
Intel the company that yammered about 60" LCOS displays for a grand and then promptly cancelled their development?
Intel the company that has announced the Atom which isn't close to be a mobile platform. Hell the 945 bridge eats up wattage like it has prader-willi ?
Intel the company that has been talking about delivering a decent GPU for years and the best they can do is the awful GMA 4xxx crap?
I don't think the process matters as much as the overall design. Intel is simply not a player in the ultra low wattage arena they are the plucky challenger hoping to unseat ARM and not the other way around.
Yep, Intel which is now executing well and dominating the CPU arena as it has in the past. Not quite the same Intel you detail above which couldn't get much right and was getting clobbered by AMD.
Atom is currently targeted toward a different market than ARM....netbooks/MID vs phones. But ARM is scaling up while Intel is scaling down. The A9 will be more power hungry than the A8 while Pineview will be less power hungry than Silverthorne.
IMHO, it's more defensive than offensive on Intel's part. I think they prefer to cannibalize their own notebook CPU sales with Atom than let ARM have an entry into the netbook arena with higher performing ARMs.
Regarding GPUs, amusingly, the Intel GMA500 is a licensed variant of the PowerVR SGX and the GMA4500 isn't that bad. The guys working on Larrabee is the same shop that worked on Nehalem. Maybe you've heard of that?
Larrabee should be a very interesting and with the number of cores they say it has the process steps really DO matter quite a bit. Whether it is a successful design remains to be seen but as a hugely cored CPU pretending to be a GPU it plays to Intel's strengths in x86 CPU design (heh) and compiler design (Intel's compilers are some of the best).
If they can beat ARM then they've got one hell of a platform because it's going to take a lot of work.
While I think that Intel will always think of domination, I really do think their expectation to be much more limited (rational) for the next few years vis a vis the embedded CPU space. ARM is the big leader but it works differently than Intel or AMD. More like a successful Sparc with many licensees (and not just a couple).
Quote:
The thing I love about the mobile arena is this.
Wintel doesn't dominate.
Microsoft's software is not the defacto and Office isn't going to help them here.
Intel's deep pockets and considerable design skills in desktop/server CPU won't apply here
Intel has made some missteps but I give them kudos for executing very well once they left Netburst architecture.
Can they deliver with Larabee? That's going to be an answer I cannot wait to see.
The Wintel dominance helped more than it hurt in the PC space. They drove computing to a commodity market which neither IBM nor Apple was inclined to do. In an IBM or Apple dominated world we'd still be paying...well...Apple prices for computers.
That's okay today given we have the choice of cheaper Wintel boxes but if that was the ONLY choice then the PC market would be a heck of a lot smaller than today and the internet not nearly as evolved.
They see where the market is heading and know they need a product that can be competitive in the smartphone arena.
There is no doubt in my mind that Intel will use the process node as the 'hammer' to beat on ARM. We know intel will get to 22nm. Will ARM? Can ARM get below 22nm?
No matter what process node you look at, Intel will get there first. That's going to be an advantage. Whether its enough to chip away at ARMs dominance in embedded chips remains to be seen.
They see where the market is heading and know they need a product that can be competitive in the smartphone arena.
There is no doubt in my mind that Intel will use the process node as the 'hammer' to beat on ARM. We know intel will get to 22nm. Will ARM? Can ARM get below 22nm?
No matter what process node you look at, Intel will get there first. That's going to be an advantage. Whether its enough to chip away at ARMs dominance in embedded chips remains to be seen.
Well, ARM not so much as the fabs the licensees use. TSMC will go 22nm. Of course, how much of that capabiltiy will get used by Intel to build Atom remains to be seen.
Ti has said no to 32nm.
IBM however will go 22nm and they have an agreement with ARM. Samsung is part of that alliance I think.
Below 22nm? Heh, let's let IBM and Intel figure out 22nm first. IBM has supposedly done so already.
Well, ARM not so much as the fabs the licensees use. TSMC will go 22nm. Of course, how much of that capabiltiy will get used by Intel to build Atom remains to be seen.
Ti has said no to 32nm.
IBM however will go 22nm and they have an agreement with ARM. Samsung is part of that alliance I think.
Below 22nm? Heh, let's let IBM and Intel figure out 22nm first. IBM has supposedly done so already.
I thought that Ti had said they weren't going to 32nm. I didn't know that TSMC was committed to going to 22nm and that IBM had an agreement with ARM.
It'll be interesting to see how this plays out. At a similar process node I would think that ARM will have an advantage over the x86 designs of Intel. However, if Intel can keep their x86 chips one process node ahead of ARM who knows?
I thought that Ti had said they weren't going to 32nm. I didn't know that TSMC was committed to going to 22nm and that IBM had an agreement with ARM.
It'll be interesting to see how this plays out. At a similar process node I would think that ARM will have an advantage over the x86 designs of Intel. However, if Intel can keep their x86 chips one process node ahead of ARM who knows?
Um, I think you misread my Ti comment. Sorry if it was unclear.
TSMC: On page 2 you see the Atom deal with Intel mentioned.
I don't expect to see Intel really trying to play against ARM until 32nm Medfield which is Intel's first announced smartphone offering. That's 2011 according to most folks.
ARM will push hard into netbooks and nettops before then I think with Android and possibly OSX.
I think Intel was being pretty forward looking in seeing ARM push into its market and seeing past AMD as a singular threat. Of course, with AMD being so-so at the moment, that's not too hard either.
Since there's no defense like a good offense, there's no reason not to push low power x86 into ARM's bread and butter markets even if they will only be moderately successful.
Comments
...The brilliant part is the tech press at large doesn't even see, which means they fail to comprehend, the end around that Apple's doing with Intel.
Well, in the mobile space Intel can pretty much give up on Apple using their tech. The tablet if released this year may use Atom though. Don't be too optimistic, nonetheless for Macs Intel will be there for at least another good 3 to 5 years.
One more thing... Is there OpenCL on the iPhone 3G/S ???
I think Atom + Tegra is more likely for the next AppleTV.
Cheapo Core 2 (Celeron/Pentium) and Nvidia Ion? Maybe. Just throwing it out there.
...laugh out loud @ apple taking on Palm. try the other way around.
Palm's international push won't be as strong as the iPhone. Since global telcos are now heavily invested in rolling out the next sexy iPhone to continue generating interest in people using their networks. Even during the recession.
I think Atom + Tegra is more likely for the next AppleTV.
Atom and Tegra are mutually exclusive, as Tegra is a system-on-a-chip with dual ARM11 cores and a low(ish) power nvidia GPU. Do you mean the ION/9400 platform perhaps? The dual-core Atom 330 and Ion would be good. but even that is overkill. There are many embedded 1080P decoders available that can be integrated into a 900-1000mhz Cortex-A8 based system-on-a-chip. The AppleTV could be the size of a deck of cards and still run great!
If you ask me, Apple would be more future oriented with this thing, I think their using at least an equivalent SGX 530, though ideally they'd use the latest SGX 543 and the VXD for video playback. Also, why not use the much better A9? It's been available for at least a year. ...
It doesn't work that way. The embedded market moves far slower than PCs. Even when a processing core is finished, it has to go through extensive testing and verification. And then that's just the CPU core. Next, it has to be integrated with a handful of other components into a system-on-a-chip from one of the major manufacturers, and then that whole chip has to be tested, debugged, and validated. Finally, the system-on-a-chip has to be integrated with another handful of components into a full smartphone platform, and then tested/validated/debugged. All of this before a single product gets out.
The Cortex-A8 core was introduced in October 2005 and only now finding its way to production smartphones. Similarly, the multi-core Cortex-A9 architecture was finalized a long time ago, but companies using it like Texas Instruments (OMAP4) and Qualcomm (future Snapdragon) aren't even sampling chips yet. Last I heard, OMAP4 samples in Q3/Q4 2009 and will be in volume production H2 2010.
Also, The Cortex-A8 chip in the iPhone 3GS is 65nm.. A power efficient dual-core Cortex-A9 will need to be in 45nm to get a decent battery life.
And even if they could, why would they jump so far ahead when what they have now is so far ahead of the competition. A Super-iPhone running a dual-core Cortex-A9 running at 1.0+ Ghz and PowerVR SGX40+ would also create major platform problems with regards to making games and applications compatible with the legacy iPhones and iPods.
Atom and Tegra are mutually exclusive, as Tegra is a system-on-a-chip with dual ARM11 cores and a low(ish) power nvidia GPU. Do you mean the ION/9400 platform perhaps? The dual-core Atom 330 and Ion would be good. but even that is overkill. There are many embedded 1080P decoders available that can be integrated into a 900-1000mhz Cortex-A8 based system-on-a-chip. The AppleTV could be the size of a deck of cards and still run great!
.
i think he did mean atom + ion. I've seen people mix up the tegra and ion chips a lot lately.
i think he did mean atom + ion. I've seen people mix up the tegra and ion chips a lot lately.
Yes, that is exactly what I did, not several hours after correcting someone on a different board for making that mistake.
It doesn't work that way. The embedded market moves far slower than PCs. Even when a processing core is finished, it has to go through extensive testing and verification. And then that's just the CPU core. Next, it has to be integrated with a handful of other components into a system-on-a-chip from one of the major manufacturers, and then that whole chip has to be tested, debugged, and validated. Finally, the system-on-a-chip has to be integrated with another handful of components into a full smartphone platform, and then tested/validated/debugged. All of this before a single product gets out.
to making games and applications compatible with the legacy iPhones and iPods.
This is exactly why I don't see Intel or Nvidia making any inroads to iPhone/iPods/ Apple TV etc.
The amount of work this it takes to design and debug SoC designs is what keeps so many vendors out the game and buying components "off the shelf" but Apple has already signaled their intentions to go this route with the PA Semi acquisition and acquisition of IBM and AMD talent.
If there was ever a time for Apple to eat its own dogfood ..now is that time.
This is exactly why I don't see Intel or Nvidia making any inroads to iPhone/iPods/ Apple TV etc.
The amount of work this it takes to design and debug SoC designs is what keeps so many vendors out the game and buying components "off the shelf" but Apple has already signaled their intentions to go this route with the PA Semi acquisition and acquisition of IBM and AMD talent.
If there was ever a time for Apple to eat its own dogfood ..now is that time.
Given that Intel is not your average company and has built ARM products in the past I don't see why you expect Intel not able to make inroads if they really really wanted to.
Especially given they already run in the AppleTV.
It seems that they really really want to but are taking their time about it. They aren't competing against ARM until at least Pineview/PineTrail (SoC) and that will go against the A9.
Intel is also going to have a process advantage over most other folks. If I remember right Atom is going 32nm this year.
But the MID market is DOA so these will go into netbooks and smartphones. Maybe an Apple tablet...
Given that Intel is not your average company and has built ARM products in the past I don't see why you expect Intel not able to make inroads if they really really wanted to.
Especially given they already run in the AppleTV.
It seems that they really really want to but are taking their time about it. They aren't competing against ARM until at least Pineview/PineTrail (SoC) and that will go against the A9.
Intel is also going to have a process advantage over most other folks. If I remember right Atom is going 32nm this year.
But the MID market is DOA so these will go into netbooks and smartphones. Maybe an Apple tablet...
Intel the company that said "clockspeed is king" and then promptly cancelled Tejas after spending millions?
Intel the company that yammered about 60" LCOS displays for a grand and then promptly cancelled their development?
Intel the company that has announced the Atom which isn't close to be a mobile platform. Hell the 945 bridge eats up wattage like it has prader-willi ?
Intel the company that has been talking about delivering a decent GPU for years and the best they can do is the awful GMA 4xxx crap?
I don't think the process matters as much as the overall design. Intel is simply not a player in the ultra low wattage arena they are the plucky challenger hoping to unseat ARM and not the other way around.
I've got my fingers crossed
Intel the company that said "clockspeed is king" and then promptly cancelled Tejas after spending millions?
Intel the company that yammered about 60" LCOS displays for a grand and then promptly cancelled their development?
Intel the company that has announced the Atom which isn't close to be a mobile platform. Hell the 945 bridge eats up wattage like it has prader-willi ?
Intel the company that has been talking about delivering a decent GPU for years and the best they can do is the awful GMA 4xxx crap?
I don't think the process matters as much as the overall design. Intel is simply not a player in the ultra low wattage arena they are the plucky challenger hoping to unseat ARM and not the other way around.
Yep, Intel which is now executing well and dominating the CPU arena as it has in the past. Not quite the same Intel you detail above which couldn't get much right and was getting clobbered by AMD.
Atom is currently targeted toward a different market than ARM....netbooks/MID vs phones. But ARM is scaling up while Intel is scaling down. The A9 will be more power hungry than the A8 while Pineview will be less power hungry than Silverthorne.
IMHO, it's more defensive than offensive on Intel's part. I think they prefer to cannibalize their own notebook CPU sales with Atom than let ARM have an entry into the netbook arena with higher performing ARMs.
Regarding GPUs, amusingly, the Intel GMA500 is a licensed variant of the PowerVR SGX and the GMA4500 isn't that bad. The guys working on Larrabee is the same shop that worked on Nehalem. Maybe you've heard of that?
Larrabee should be a very interesting and with the number of cores they say it has the process steps really DO matter quite a bit. Whether it is a successful design remains to be seen but as a hugely cored CPU pretending to be a GPU it plays to Intel's strengths in x86 CPU design (heh) and compiler design (Intel's compilers are some of the best).
If they can beat ARM then they've got one hell of a platform because it's going to take a lot of work.
The thing I love about the mobile arena is this.
Wintel doesn't dominate.
Microsoft's software is not the defacto and Office isn't going to help them here.
Intel's deep pockets and considerable design skills in desktop/server CPU won't apply here
Intel has made some missteps but I give them kudos for executing very well once they left Netburst architecture.
Can they deliver with Larabee? That's going to be an answer I cannot wait to see.
I look forward to the battle.
If they can beat ARM then they've got one hell of a platform because it's going to take a lot of work.
While I think that Intel will always think of domination, I really do think their expectation to be much more limited (rational) for the next few years vis a vis the embedded CPU space. ARM is the big leader but it works differently than Intel or AMD. More like a successful Sparc with many licensees (and not just a couple).
The thing I love about the mobile arena is this.
Wintel doesn't dominate.
Microsoft's software is not the defacto and Office isn't going to help them here.
Intel's deep pockets and considerable design skills in desktop/server CPU won't apply here
Intel has made some missteps but I give them kudos for executing very well once they left Netburst architecture.
Can they deliver with Larabee? That's going to be an answer I cannot wait to see.
The Wintel dominance helped more than it hurt in the PC space. They drove computing to a commodity market which neither IBM nor Apple was inclined to do. In an IBM or Apple dominated world we'd still be paying...well...Apple prices for computers.
That's okay today given we have the choice of cheaper Wintel boxes but if that was the ONLY choice then the PC market would be a heck of a lot smaller than today and the internet not nearly as evolved.
They see where the market is heading and know they need a product that can be competitive in the smartphone arena.
There is no doubt in my mind that Intel will use the process node as the 'hammer' to beat on ARM. We know intel will get to 22nm. Will ARM? Can ARM get below 22nm?
No matter what process node you look at, Intel will get there first. That's going to be an advantage. Whether its enough to chip away at ARMs dominance in embedded chips remains to be seen.
PowerVR MBX .... 2004?
http://www.imgtec.com/News/Release/index.asp?NewsID=77
'791
I wouldn't throw dirt on Intel just yet.
They see where the market is heading and know they need a product that can be competitive in the smartphone arena.
There is no doubt in my mind that Intel will use the process node as the 'hammer' to beat on ARM. We know intel will get to 22nm. Will ARM? Can ARM get below 22nm?
No matter what process node you look at, Intel will get there first. That's going to be an advantage. Whether its enough to chip away at ARMs dominance in embedded chips remains to be seen.
Well, ARM not so much as the fabs the licensees use. TSMC will go 22nm. Of course, how much of that capabiltiy will get used by Intel to build Atom remains to be seen.
Ti has said no to 32nm.
IBM however will go 22nm and they have an agreement with ARM. Samsung is part of that alliance I think.
Below 22nm? Heh, let's let IBM and Intel figure out 22nm first. IBM has supposedly done so already.
Well, ARM not so much as the fabs the licensees use. TSMC will go 22nm. Of course, how much of that capabiltiy will get used by Intel to build Atom remains to be seen.
Ti has said no to 32nm.
IBM however will go 22nm and they have an agreement with ARM. Samsung is part of that alliance I think.
Below 22nm? Heh, let's let IBM and Intel figure out 22nm first. IBM has supposedly done so already.
I thought that Ti had said they weren't going to 32nm. I didn't know that TSMC was committed to going to 22nm and that IBM had an agreement with ARM.
It'll be interesting to see how this plays out. At a similar process node I would think that ARM will have an advantage over the x86 designs of Intel. However, if Intel can keep their x86 chips one process node ahead of ARM who knows?
I thought that Ti had said they weren't going to 32nm. I didn't know that TSMC was committed to going to 22nm and that IBM had an agreement with ARM.
It'll be interesting to see how this plays out. At a similar process node I would think that ARM will have an advantage over the x86 designs of Intel. However, if Intel can keep their x86 chips one process node ahead of ARM who knows?
Um, I think you misread my Ti comment. Sorry if it was unclear.
TSMC: On page 2 you see the Atom deal with Intel mentioned.
http://www.eetimes.com/showArticle.j...leID=216900470
IBM and ARM
http://www.informationweek.com/news/...leID=210604549
I don't expect to see Intel really trying to play against ARM until 32nm Medfield which is Intel's first announced smartphone offering. That's 2011 according to most folks.
ARM will push hard into netbooks and nettops before then I think with Android and possibly OSX.
I think Intel was being pretty forward looking in seeing ARM push into its market and seeing past AMD as a singular threat. Of course, with AMD being so-so at the moment, that's not too hard either.
Since there's no defense like a good offense, there's no reason not to push low power x86 into ARM's bread and butter markets even if they will only be moderately successful.