It's quite a bit of work to bring up a chip of a given process. Apple trying to shift silicon production away from Samsung is likely and Apple dual sourcing is also reasonable. But, making SoC at 3 different foundries each with distinct processes? That seems excessive, expensive, and of little benefit.
Now, institutional investors reportedly believe that Samsung will receive about half of Apple's "A7" orders, while TSMC will take 40 percent, and Intel will grab the remaining 10 percent.
I can imagine the phone calls that happened:
To Intel: "This is Tim. I'm going to throw the old dog a bone. You get 10% of our A7 business. Live it up Fido."
To Samsung: "This is Tim. I'm cutting your A7 order down to 50%. And I'm cutting your A8 order down to 0%, chumps."
To TSMC: "This is Tim. Congrats. You get 40% of our A7 orders. But Intel just might outbid you. Don't get cocky."
Am I the only one who thinks building the same chip on three different processes like this sounds absurd? Apple's chips are custom built, this means they have to have people layout the transistors etc. Sure the tools help with the process, but it is by no means cheap. Up until now, each chip has had a team make it for a single process technology (a shrink in the tech node means a new chip design, even if it contains the same "logical" circuitry.
Also, Samsung and TSMCs technologies are likely very similar, as they cross license much of the technology needed to manufacture chips. Intel's technology is quite a bit different, as they don't tend to license their cutting edge technology to the others. Intel's process technologies are also primarily used in-house, so it's unlikely that third party placement and routing tools are made to handle them.
The amount of time and effort required of apple's design teams to learn these new tools and use them likely isn't worth it for 10% of their chip orders... It just doesn't make sense unless Apple's making a very expensive gamble to test the waters. Each technology that apple produces a chip on adds greatly to their fixed costs. Mask sets alone cost tens of millions of dollars. Design teams easily cost more to make the new chip.
Basically the point of this is that it's not a "simple" process for apple to change manufacturers of their chips. There's a huge lead-time required to get the team ready to use new technologies, and huge costs associated with this. Unless apple's been secretly hiring hundreds of design engineers, I doubt the A chips will be produced on vastly different technologies.
That's right, Samsung are currently making a lot of money with what is essentially a rip off of the iPhone/iPad line. Steve Jobs thought he made a mistake not being aggressive enough patenting the Mac, allowing MS to come in, so with the iPhone he patented all he could. Now Tim is learning that the legal system doesn't protect you even if you *do* patent, so Steve did not make a mistake after all, there is simply no-one who will stand up for the innovators.
Great post. Short, relevant, and highlights an extremely important aspect of what's happened in mobile over the past few years. Perhaps THE most important aspect.
Am I the only one who thinks building the same chip on three different processes like this sounds absurd?
...
Also, Samsung and TSMCs technologies are likely very similar, as they cross license much of the technology needed to manufacture chips. Intel's technology is quite a bit different, as they don't tend to license their cutting edge technology to the others. Intel's process technologies are also primarily used in-house, so it's unlikely that third party placement and routing tools are made to handle them.
The amount of time and effort required of apple's design teams to learn these new tools and use them likely isn't worth it for 10% of their chip orders... It just doesn't make sense unless Apple's making a very expensive gamble to test the waters. Each technology that apple produces a chip on adds greatly to their fixed costs. Mask sets alone cost tens of millions of dollars. Design teams easily cost more to make the new chip.
Basically the point of this is that it's not a "simple" process for apple to change manufacturers of their chips. There's a huge lead-time required to get the team ready to use new technologies, and huge costs associated with this. Unless apple's been secretly hiring hundreds of design engineers, I doubt the A chips will be produced on vastly different technologies.
Phil
No, it's not simple.
Yes, it costs a lot.
Yes, it takes time.
On the other hand the single biggest risk is SoC supply. Apple has plenty of money so that tens of millions to mitigate supply risk is worthwhile. Also, while Apple may not have a full team spun up on intels tools they've been working with intel for a while and certainly Intel has the expertise and manpower if they desired to get it done.
Personally, I still think that if intel agrees to make A7 for Apple they're going to want a mobile design win for Atom. That might be an iPad or it might be a touch enabled 11" MBA running an iOS/OSX hybrid.
Given the number of folks with a keyboard for the full sized iPad I think that would sell well in the business market.
On the other hand the single biggest risk is SoC supply. Apple has plenty of money so that tens of millions to mitigate supply risk is worthwhile. Also, while Apple may not have a full team spun up on intels tools they've been working with intel for a while and certainly Intel has the expertise and manpower if they desired to get it done.
Personally, I still think that if intel agrees to make A7 for Apple they're going to want a mobile design win for Atom. That might be an iPad or it might be a touch enabled 11" MBA running an iOS/OSX hybrid.
Given the number of folks with a keyboard for the full sized iPad I think that would sell well in the business market.
Sorry, but it is more cost effective for Apple to buy GlobalFoundries than to invest and switch over to Intel who themselves would have to sink well over $10 billion to be on the road to ARM plant certification.
Am I the only one who thinks building the same chip on three different processes like this sounds absurd? Apple's chips are custom built, this means they have to have people layout the transistors etc. Sure the tools help with the process, but it is by no means cheap. Up until now, each chip has had a team make it for a single process technology (a shrink in the tech node means a new chip design, even if it contains the same "logical" circuitry.
Also, Samsung and TSMCs technologies are likely very similar, as they cross license much of the technology needed to manufacture chips. Intel's technology is quite a bit different, as they don't tend to license their cutting edge technology to the others. Intel's process technologies are also primarily used in-house, so it's unlikely that third party placement and routing tools are made to handle them.
The amount of time and effort required of apple's design teams to learn these new tools and use them likely isn't worth it for 10% of their chip orders... It just doesn't make sense unless Apple's making a very expensive gamble to test the waters. Each technology that apple produces a chip on adds greatly to their fixed costs. Mask sets alone cost tens of millions of dollars. Design teams easily cost more to make the new chip.
Basically the point of this is that it's not a "simple" process for apple to change manufacturers of their chips. There's a huge lead-time required to get the team ready to use new technologies, and huge costs associated with this. Unless apple's been secretly hiring hundreds of design engineers, I doubt the A chips will be produced on vastly different technologies.
Phil
Can we all get this through our heads: Global Foundries, TSMC and Samsung partnered to build a fabrication process for ARM and the same FinFET tech to compete against Intel.
Comments
It's quite a bit of work to bring up a chip of a given process. Apple trying to shift silicon production away from Samsung is likely and Apple dual sourcing is also reasonable. But, making SoC at 3 different foundries each with distinct processes? That seems excessive, expensive, and of little benefit.
Originally Posted by AppleInsider
Now, institutional investors reportedly believe that Samsung will receive about half of Apple's "A7" orders, while TSMC will take 40 percent, and Intel will grab the remaining 10 percent.
I can imagine the phone calls that happened:
To Intel: "This is Tim. I'm going to throw the old dog a bone. You get 10% of our A7 business. Live it up Fido."
To Samsung: "This is Tim. I'm cutting your A7 order down to 50%. And I'm cutting your A8 order down to 0%, chumps."
To TSMC: "This is Tim. Congrats. You get 40% of our A7 orders. But Intel just might outbid you. Don't get cocky."
Also, Samsung and TSMCs technologies are likely very similar, as they cross license much of the technology needed to manufacture chips. Intel's technology is quite a bit different, as they don't tend to license their cutting edge technology to the others. Intel's process technologies are also primarily used in-house, so it's unlikely that third party placement and routing tools are made to handle them.
The amount of time and effort required of apple's design teams to learn these new tools and use them likely isn't worth it for 10% of their chip orders... It just doesn't make sense unless Apple's making a very expensive gamble to test the waters. Each technology that apple produces a chip on adds greatly to their fixed costs. Mask sets alone cost tens of millions of dollars. Design teams easily cost more to make the new chip.
Basically the point of this is that it's not a "simple" process for apple to change manufacturers of their chips. There's a huge lead-time required to get the team ready to use new technologies, and huge costs associated with this. Unless apple's been secretly hiring hundreds of design engineers, I doubt the A chips will be produced on vastly different technologies.
Phil
Quote:
Originally Posted by ascii
That's right, Samsung are currently making a lot of money with what is essentially a rip off of the iPhone/iPad line. Steve Jobs thought he made a mistake not being aggressive enough patenting the Mac, allowing MS to come in, so with the iPhone he patented all he could. Now Tim is learning that the legal system doesn't protect you even if you *do* patent, so Steve did not make a mistake after all, there is simply no-one who will stand up for the innovators.
Great post. Short, relevant, and highlights an extremely important aspect of what's happened in mobile over the past few years. Perhaps THE most important aspect.
No, it's not simple.
Yes, it costs a lot.
Yes, it takes time.
On the other hand the single biggest risk is SoC supply. Apple has plenty of money so that tens of millions to mitigate supply risk is worthwhile. Also, while Apple may not have a full team spun up on intels tools they've been working with intel for a while and certainly Intel has the expertise and manpower if they desired to get it done.
Personally, I still think that if intel agrees to make A7 for Apple they're going to want a mobile design win for Atom. That might be an iPad or it might be a touch enabled 11" MBA running an iOS/OSX hybrid.
Given the number of folks with a keyboard for the full sized iPad I think that would sell well in the business market.
Quote:
Originally Posted by SolipsismX
Only 10%? Then it's highly unlikely they'll be the 22nm lithography which would kill what I'd think is the primary benefit of going with Intel.
Off topic: Ever wonder how a silicon wafer is "grown" or how the exact same CPU can have performance and thermal variances?
Cool video. Thanks for the link.
Quote:
Originally Posted by nht
No, it's not simple.
Yes, it costs a lot.
Yes, it takes time.
On the other hand the single biggest risk is SoC supply. Apple has plenty of money so that tens of millions to mitigate supply risk is worthwhile. Also, while Apple may not have a full team spun up on intels tools they've been working with intel for a while and certainly Intel has the expertise and manpower if they desired to get it done.
Personally, I still think that if intel agrees to make A7 for Apple they're going to want a mobile design win for Atom. That might be an iPad or it might be a touch enabled 11" MBA running an iOS/OSX hybrid.
Given the number of folks with a keyboard for the full sized iPad I think that would sell well in the business market.
Sorry, but it is more cost effective for Apple to buy GlobalFoundries than to invest and switch over to Intel who themselves would have to sink well over $10 billion to be on the road to ARM plant certification.
Quote:
Originally Posted by philgar
Am I the only one who thinks building the same chip on three different processes like this sounds absurd? Apple's chips are custom built, this means they have to have people layout the transistors etc. Sure the tools help with the process, but it is by no means cheap. Up until now, each chip has had a team make it for a single process technology (a shrink in the tech node means a new chip design, even if it contains the same "logical" circuitry.
Also, Samsung and TSMCs technologies are likely very similar, as they cross license much of the technology needed to manufacture chips. Intel's technology is quite a bit different, as they don't tend to license their cutting edge technology to the others. Intel's process technologies are also primarily used in-house, so it's unlikely that third party placement and routing tools are made to handle them.
The amount of time and effort required of apple's design teams to learn these new tools and use them likely isn't worth it for 10% of their chip orders... It just doesn't make sense unless Apple's making a very expensive gamble to test the waters. Each technology that apple produces a chip on adds greatly to their fixed costs. Mask sets alone cost tens of millions of dollars. Design teams easily cost more to make the new chip.
Basically the point of this is that it's not a "simple" process for apple to change manufacturers of their chips. There's a huge lead-time required to get the team ready to use new technologies, and huge costs associated with this. Unless apple's been secretly hiring hundreds of design engineers, I doubt the A chips will be produced on vastly different technologies.
Phil
Can we all get this through our heads: Global Foundries, TSMC and Samsung partnered to build a fabrication process for ARM and the same FinFET tech to compete against Intel.
http://semimd.com/blog/tag/14nm/
There are far better places to figure out what is actually going on in the industry than AI Rumors.
Quote:
Originally Posted by zoffdino
Last year Apple sold 130M iOS devices.
Last year Apple sold 230M iOS devices.