I don't see Apple necessarily designing super high-performing chipsets. Instead, they will use their (very expensive) IP, design team and ARM/PowerVR licenses to design parts that run their software within a desired performance/watt envelope at a predictable price. It's about tailoring components to precisely fit their particular needs, not raw performance. I wouldn't be surprised if they added hardware DRM to satisfy content providers.
There has to be a question mark about how worthwhile the costs of running an in-house chip design team will be in the long run. It's not even like the modern Apple to have such a thing, since they have been moving more and more towards outsourced, commodity components and focussing on their core competencies in software and industrial design (and now retail). I wonder what will happen if Tegra 2 (or 3, or 4) beats their best internal efforts? Will they still eat their own dogfood? I suppose the design licenses may give them some leverage in the market, or influence over the future direction of ARM and Power VR cores.
I don't see Apple necessarily designing super high-performing chipsets.
In the context of iPhone no, instead they will go System on Chip (SoC). While that is a fine line to draw, putting everything on one chip (as is technologically possible) gives them huge advantages. Especially if the I/O is tightly tailored to the requirements of the device, as it would save a lot of power.
Quote:
Instead, they will use their (very expensive) IP, design team and ARM/PowerVR licenses to design parts that run their software within a desired performance/watt envelope at a predictable price.
I was in an interesting thread with another guy discussing this very thing. I supported a position that kinda mirrors yours but the other poster was convincing in his argument that the chip would be a full custom implementation. Considering PA Semis background he may be right. One of his argument was that a full custom chip would be the avenue to the lowest power design. This would certainly be true, but it is also an expensive approach.
Quote:
It's about tailoring components to precisely fit their particular needs, not raw performance.
A chip targeted at the tablet would be all about performance at the lowest possible power point. IPhone chips are a different story, but even here the 3GS still needs a boost in both directions. That is lower power and higher performance.
Quote:
I wouldn't be surprised if they added hardware DRM to satisfy content providers.
That would hurt but to be honest I think many of the content providers are learning the hard way. Learning that pissing off your customers and forcing them to alternative entertainment is counter productive.
That being said a full custom approach allows Apple to do what ever they want. People forget about some of the technology Apple has purchased up in the past. Just one of them, the rumored FingerWorks technology might end up requiring a subsystem of its very own on the SoC. They could also tailor the ARM chip in ways that might not be possible otherwise, one idea being instructions to specifically speed up Objective C programs. They could also do things with ARM like deleting the vector unit and attaching a GPU core to each CPU. In effect tightly coupling an OpenCL chip to the ARM.
Or they could do something like put the frame buffer for the GPU right on chip with the GPU. That would lead to significant performance without the power sapping that going off chip for memory entails. The point is the possibilities are endless.
Quote:
There has to be a question mark about how worthwhile the costs of running an in-house chip design team will be in the long run.
This I agree with 100%. I'm not convinced Apple can keep up in the long run.
Quote:
It's not even like the modern Apple to have such a thing, since they have been moving more and more towards outsourced, commodity components and focussing on their core competencies in software and industrial design (and now retail).
Well yeah they are doing that or where, but that was more out of necessity than anything. Relying on the commodity world does leave you open to issues with being able to deliver more than competitive products. Especially for these small hand held devices. In many ways the ability to design a SoC is the equivalent of being able to design a TTL based system board in the days of the Apple 2.
In other words modern hardware requires the ability to innovate with SoC technology to remain competitive. If you think about it this really isn't a surprise, in this day and age you can put several Apple 2s on a single chip and have them all running may times faster.
Quote:
I wonder what will happen if Tegra 2 (or 3, or 4) beats their best internal efforts? Will they still eat their own dogfood?
That is a good question! I would have to say the SoC will have IP that is private to Apple and not easily copied by others. I've already mentioned the possibility of an interface subsystem to handle the FingerWorks technology. Apple though has the potential to add other IP or to customize the ARM CPU to meet their needs.
Another problem is that the TEGRA and like chips from other vendors seem to take a kitchen sink approach. That means they are likely to use more power. Even more so if the chips are ASIC put together from plug in IP blocks.
Quote:
I suppose the design licenses may give them some leverage in the market, or influence over the future direction of ARM and Power VR cores.
ARM, I'm not to sure about. There are a lot of ARM variants out there, but it seems like ARM leaves that to the implementers. Imagination is another story all together as it looks like Apple was a huge influence there with respect to the support of OpenCL.
In any event in 15 days we ought to at least get a hint.
Comments
There has to be a question mark about how worthwhile the costs of running an in-house chip design team will be in the long run. It's not even like the modern Apple to have such a thing, since they have been moving more and more towards outsourced, commodity components and focussing on their core competencies in software and industrial design (and now retail). I wonder what will happen if Tegra 2 (or 3, or 4) beats their best internal efforts? Will they still eat their own dogfood? I suppose the design licenses may give them some leverage in the market, or influence over the future direction of ARM and Power VR cores.
I don't see Apple necessarily designing super high-performing chipsets.
In the context of iPhone no, instead they will go System on Chip (SoC). While that is a fine line to draw, putting everything on one chip (as is technologically possible) gives them huge advantages. Especially if the I/O is tightly tailored to the requirements of the device, as it would save a lot of power.
Instead, they will use their (very expensive) IP, design team and ARM/PowerVR licenses to design parts that run their software within a desired performance/watt envelope at a predictable price.
I was in an interesting thread with another guy discussing this very thing. I supported a position that kinda mirrors yours but the other poster was convincing in his argument that the chip would be a full custom implementation. Considering PA Semis background he may be right. One of his argument was that a full custom chip would be the avenue to the lowest power design. This would certainly be true, but it is also an expensive approach.
It's about tailoring components to precisely fit their particular needs, not raw performance.
A chip targeted at the tablet would be all about performance at the lowest possible power point. IPhone chips are a different story, but even here the 3GS still needs a boost in both directions. That is lower power and higher performance.
I wouldn't be surprised if they added hardware DRM to satisfy content providers.
That would hurt but to be honest I think many of the content providers are learning the hard way. Learning that pissing off your customers and forcing them to alternative entertainment is counter productive.
That being said a full custom approach allows Apple to do what ever they want. People forget about some of the technology Apple has purchased up in the past. Just one of them, the rumored FingerWorks technology might end up requiring a subsystem of its very own on the SoC. They could also tailor the ARM chip in ways that might not be possible otherwise, one idea being instructions to specifically speed up Objective C programs. They could also do things with ARM like deleting the vector unit and attaching a GPU core to each CPU. In effect tightly coupling an OpenCL chip to the ARM.
Or they could do something like put the frame buffer for the GPU right on chip with the GPU. That would lead to significant performance without the power sapping that going off chip for memory entails. The point is the possibilities are endless.
There has to be a question mark about how worthwhile the costs of running an in-house chip design team will be in the long run.
This I agree with 100%. I'm not convinced Apple can keep up in the long run.
It's not even like the modern Apple to have such a thing, since they have been moving more and more towards outsourced, commodity components and focussing on their core competencies in software and industrial design (and now retail).
Well yeah they are doing that or where, but that was more out of necessity than anything. Relying on the commodity world does leave you open to issues with being able to deliver more than competitive products. Especially for these small hand held devices. In many ways the ability to design a SoC is the equivalent of being able to design a TTL based system board in the days of the Apple 2.
In other words modern hardware requires the ability to innovate with SoC technology to remain competitive. If you think about it this really isn't a surprise, in this day and age you can put several Apple 2s on a single chip and have them all running may times faster.
I wonder what will happen if Tegra 2 (or 3, or 4) beats their best internal efforts? Will they still eat their own dogfood?
That is a good question! I would have to say the SoC will have IP that is private to Apple and not easily copied by others. I've already mentioned the possibility of an interface subsystem to handle the FingerWorks technology. Apple though has the potential to add other IP or to customize the ARM CPU to meet their needs.
Another problem is that the TEGRA and like chips from other vendors seem to take a kitchen sink approach. That means they are likely to use more power. Even more so if the chips are ASIC put together from plug in IP blocks.
I suppose the design licenses may give them some leverage in the market, or influence over the future direction of ARM and Power VR cores.
ARM, I'm not to sure about. There are a lot of ARM variants out there, but it seems like ARM leaves that to the implementers. Imagination is another story all together as it looks like Apple was a huge influence there with respect to the support of OpenCL.
In any event in 15 days we ought to at least get a hint.
Dave