There are actually more than one 20 nm processes out there. In the case of TSMC Apple is using their low power process that is actually shipping product. The high power process, that you might find GPU manufactures using for example, isn't up and ready to go. My impression is that TSMC does have other customers using the low power process.
What is interesting here is that the entire industry has prioritized low power processes over the performance processes.
I'll allow that last bit. But if that's true, then the entire industry needs a serious revamp. Apple basically owns the 20nm market right now, no one else can get 20nm anything produced.
This isn't really true. If you want to produce on a low power process you can.
So what happens next year when they dominate the 14nm market as well?
If things go to plan they won't dominate. We should see Samsung and Global on line early in 2015 followed by TSMC. We may actually see a faster 14nm ramp in 2015 than is normal.
In any event it is actually a good thing that Apple is able to stay on this bleeding edge as it makes for some of the best handheld products out there.
Getting a handle on what Intel is actually going to do is a bit like pushing a worm - you can try but it will lead to frustration. In any event my understanding was that Skylake would have Intel refocusing on mainstream processors, that is laptop and desktop parts. Those where suppose to hit in mid 2015. That does not appear to be the case anymore.
I was under the impression that Skylake is on the same schedule cadence as Broadwell, it'll be launched later in 2015 starting with ULV mobile versions and ramping up to higher-performing parts into 2016. Meaning, we probably won't see Macs using Skylake starting until early 2016, like Broadwell. I have yet to read any reliable reports of Skylake starting to ship in summer 2015.
Rumor has it that Skylake is a finished design for the most part. It would be a management decision to go with it in mid 2015.
The Exynos 5430 and 5433 in the Galaxy Alpha and Galaxy Note 4 are 20 nm SoCs from Samsung's 20 nm. They may ship in one or two of Samsung's countless tablet models too. Not the same volumes as the A8 SoCs, but it's in the millions at least. Heck, the 5433 is a 64-bit A57/A53 big.LITTLE SoC, but Samsung is running it in 32-bit mode for various reasons. Maybe they'll advertise it at 64-bit when they ship Lollipop for it.
I believe Qualcomm has 20 nm parts shipping too. I could be wrong as I don't follow them closely but there are a lot of chips shipping on 20 nm now. Chips running on the low power 20 nm processes aren't a problem, it is a different story with the performance optimized processes though.
AMD has confirmed that it will be shipping 20nm chips(APU's) in 2015, but 16nm/14nm won't come till Zen and K12 which is 2016 at earliest.
Well yeah that is what the road maps say. However AMD has a new management staff in place with a very strong engineering background. These people want to reestablish AMD as a viable processor supplier. I suspect that they will push very hard to get something out on 14 nm before the end of 2015. It might not be K12. It could be an ARM based chip or even something new in the brazos family.
It would do AMD wonders to get something out the door before 2016 that is built on a 14 nm process. It would make investors and customers alike happy. I doesn't have to be a mainstream processor, in just needs to demonstrate viability in AMDs engineering and management teams. I mentioned ARM because it is an easy architecture to port to a new process.
Getting a handle on what Intel is actually going to do is a bit like pushing a worm - you can try but it will lead to frustration. In any event my understanding was that Skylake would have Intel refocusing on mainstream processors, that is laptop and desktop parts. Those where suppose to hit in mid 2015. That does not appear to be the case anymore.
Rumor has it that Skylake is a finished design for the most part. It would be a management decision to go with it in mid 2015.
I see. If that's the case, then Intel should, as you mentioned in an earlier post, scrap Broadwell and jump straight to Skylake
Isn't their a rumor that Apple is already doing this with the A8 SoC? Well, at the Korean new mongers are saying this, with them saying Samsung is producing up to 40% of the A8 SoCs Apple has shipped.
There has been no proof of this, the tear downs of the SoC to date show TSMC as the manufacturers. My personal belief when we hear that Samsung has 40% of the SOC business it means of all of apple's SoC business so the 4s and 5 and 5s and 5c use chips manufactured by Samsung and this would be true. Anything which apples is still selling and shipping which does not have A8 in is a Samsung chip. In about 18 months time frame when Apple is not selling anything less than an A8 Samsung will be out of the pictures completely.
What is interesting here is that Intels screw ups here far exceed anything AMD has done and yet Intel gets thrown softballs from the press. You have to wonder how much longer this unethical support of Intel will go on.
There's nothing AMD can do to hurt Intel in the near future. As long as Intel is a node ahead and has a competitive design, they are basically in a monopolist position in the laptop, desktop and server world, mistakes are eminently survivable, and a boring media story. And we all know that there's nothing more than the media hates is a boring story.
Intel's problems in the handset world is heavily taxed by their business model and strategy. They can ship the most performant ARM SoC today if they wanted to, but they are sticking with x86 and trying to thread the needle of not shipping a too powerful handset x86 SoC at $30 that won't nuke their $300 processors. 10 years ago, when PocketPC PDAs were the "next big thing", they participated with StrongARM/Xscale, but unfortunately, the near-death experience from Netbust made them pretty conservative on the non-x86 front. If they kept with ARM, they could be the dominate ARM SoC provider today.
Actually if Intel wants to maintain any respect in the industry they really need to consider scrapping Broadwell and simply go with Skylake. Beyond that AMD could be shipping 14 nm chips before 2016, probably chips built on a new architecture. In the end Intel can't really afford to hold off on Skylake.
Hmm, or they are doubling-down on the stupid. Skylake better be done and provably better than Broadwell in order for them to shorten Broadwell's life cycle. If Skylake is not done, it could mean 2 Holiday seasons where laptops and desktops really haven't gotten anywhere. Intel can make mistakes, but they can't make them forever.
I can see a 5 to 10 W TDP Apple Ax SoC making a very nice laptop or desktop product. Fanless, super thin, and only costing Apple $50 instead of $300 from Intel. The notional, Platonic ideal for the iMac imo is a 10 mm thin device. It's like 30 mm to 40 mm today, but thinning that down further to 10 mm, going with SSDs only, USB 3.1, etc, sounds like a nice machine that looks like it is from the future. These comments on Thunderbolt from the other thread or what have you I/O is immaterial. Thunderbolt is nice, yes, but Apple would drop it with nary a regret if it was holding it back. They can just go with USB 3.1 and HDMI 2.0 or a customized HDMI 2.0 and call it a day.
On the subject of Samsung Semi and 14/16 nm, Apple wants to have more then one vendor be able to manufacture their components. Better yet, three. It's only a good thing. If there was actually a third GPU vendor, Apple would have tried to shore it up and shipped them in their Macs. The two discrete GPU vendors today kind of suck. Not the two companies, but the situation of having only to choose components from 2 vendors. So, it's pretty vital for Apple to have both Samsung Semi and TSMC be able to fab their chips.
Well yeah that is what the road maps say. However AMD has a new management staff in place with a very strong engineering background. These people want to reestablish AMD as a viable processor supplier. I suspect that they will push very hard to get something out on 14 nm before the end of 2015. It might not be K12. It could be an ARM based chip or even something new in the brazos family.
It would do AMD wonders to get something out the door before 2016 that is built on a 14 nm process. It would make investors and customers alike happy. I doesn't have to be a mainstream processor, in just needs to demonstrate viability in AMDs engineering and management teams. I mentioned ARM because it is an easy architecture to port to a new process.
Oh yea, I agree that they're going to push it, just not going to make it pre 2016. 2015 is a critical year for AMD in the launch of its Skybridge initiative to make it's X86 and ARM offerings pin compatible with each other on a socket. Zen and K12's effectiveness is driven hugely by this. Dr Lisa su even says she wants AMD to be mean, lean and fast, but is not willing to rush product launches.
2015 will be Puma+ cores for X86, and a low power 64-bit A57 for ARM. Then by Q3-Q4 we should hear more about Zen and K12, both of which are high frequency yet power efficient designs.
For me? I'm just hoping to see a 16nm GPU from then in 2016. If Zen could at least match Haswell on IPC, then I would also pretty much pick up an AMD CPU to go along with it in my next build. Otherwise it'll be intel+AMD GPU for me.
Apple shouldn't use Samsung anymore. Hundreds of workers in their semiconductor fab got leukemia because they were leaking harmful chemicals and Samsung denies that those are work related. Even last they had a gas leak and some worker died. And they are so arrogant that they didn't even report to the police for a whole day. Everybody should boycott any products with Samsung components in them.
Apple shouldn't use Samsung anymore. Hundreds of workers in their semiconductor fab got leukemia because they were leaking harmful chemicals and Samsung denies that those are work related. Even last they had a gas leak and some worker died. And they are so arrogant that they didn't even report to the police for a whole day. Everybody should boycott any products with Samsung components in them.
<p>Apple shouldn't use Samsung anymore. Hundreds of workers in their semiconductor fab got leukemia because they were leaking harmful chemicals and Samsung denies that those are work related. Even last they had a gas leak and some worker died. And they are so arrogant that they didn't even report to the police for a whole day. Everybody should boycott any products with Samsung components in them.</p>
This isn't even something new. The Korean newspapers reported back in Oct. 2013 that Samsung had won back the majority of Apple's 2015 A9 production by virtue of being first to roll out a properly performing 14nm die, well ahead of TSMC. Yeah Apple works on these contracts years in advance.
EDIT: Here you go. Take a look at the date of this even earlier article.
The Korean newspapers also reported Samsung had won the majority of the A8 production. Those reports were proven false. There is no proof A9 reports are telling the truth. One thing is certain, the reports are succeeding at making Samsung look good until the reports are proven false. When the reports are proven false, new reports surface that Samsung won the next A production.
The last three paragraphs of the so-called report are the most important.
2014/12/12 By Bae Ok-jin
Samsung Electronics began production of 'A9,' the application processor (AP) for Apple's next-generation smartphone. It applies the 14nm FinFET, the cutting-edge microprocess for system semiconductors, for the first time. SEC's foundry business picked up momentum again.
According to industry insiders on December 11, SEC began production of Apple's A9 in the Austin plant in the US using the 14nm FinFET technology. Samsung has production lines capable of FinFET process production in Austin, US and Giheung, Korea, but began to produce A9 only in Austin as it is in the initial stage.
SEC originally said it would start production of the 14nm FinFET chip at the end of this year, but did not disclose whether the company received an order from Apple for the production of A9 chips or whether the production line is actually running. SEC has shown confidence in the yield of the 14nm FinFET process, and supplied samples as good as finished products early enough.
The Austin plant began official production first at Apple's request, and industry insiders said it is a move to produce the chip in the US, not Korea. They guessed that the Austin plant was chosen because of the next-generation chip's problems with performance security and supply.
The initiation of the A9 chip production enabled SEC to recover the foundry quantities from Apple, which have been discontinued for some time, and get the upper hand in the 14nm FinFET technology competition with TSMC, killing two birds with one stone.
Apple and SEC virtually stopped AP production as they were embroiled in patent litigation back in 2012. SEC preoccupation of the 14nm FinFET technology led to resumption of the cooperation between the two tech giants.
However, it's too early to relax yet. Taiwan's TSMC began the risk production of the 16nm FinFET plus (16FF+) process, and began to produce chips in July earlier than originally anticipated Q3, getting ready in a hurry. As Apple is adjusting the A9 chip production quantity shuttling between SEC and TSMC, if TSMC's production line is stabilized in the future, there is no knowing how SEC will respond.
SEC's foundry business was hit hard when Apple orders stopped. Although the entire semiconductor business is booming, securities companies predict that the system LSI business, including the foundry business, will suffer a loss to the tune of KRW800 billion this year. SEC is expected to recover sales loss to a certain extent with the production of Apple A9.
“We said we would inject production wafers when we announced Q3 performance,” said SEC. “But we cannot confirm whether we received orders from Apple or at which plant it will be produced.”
Comments
What is interesting here is that the entire industry has prioritized low power processes over the performance processes.
This isn't really true. If you want to produce on a low power process you can.
If things go to plan they won't dominate. We should see Samsung and Global on line early in 2015 followed by TSMC. We may actually see a faster 14nm ramp in 2015 than is normal.
In any event it is actually a good thing that Apple is able to stay on this bleeding edge as it makes for some of the best handheld products out there.
Rumor has it that Skylake is a finished design for the most part. It would be a management decision to go with it in mid 2015.
I believe Qualcomm has 20 nm parts shipping too. I could be wrong as I don't follow them closely but there are a lot of chips shipping on 20 nm now. Chips running on the low power 20 nm processes aren't a problem, it is a different story with the performance optimized processes though.
Well yeah that is what the road maps say. However AMD has a new management staff in place with a very strong engineering background. These people want to reestablish AMD as a viable processor supplier. I suspect that they will push very hard to get something out on 14 nm before the end of 2015. It might not be K12. It could be an ARM based chip or even something new in the brazos family.
It would do AMD wonders to get something out the door before 2016 that is built on a 14 nm process. It would make investors and customers alike happy. I doesn't have to be a mainstream processor, in just needs to demonstrate viability in AMDs engineering and management teams. I mentioned ARM because it is an easy architecture to port to a new process.
Getting a handle on what Intel is actually going to do is a bit like pushing a worm - you can try but it will lead to frustration. In any event my understanding was that Skylake would have Intel refocusing on mainstream processors, that is laptop and desktop parts. Those where suppose to hit in mid 2015. That does not appear to be the case anymore.
Rumor has it that Skylake is a finished design for the most part. It would be a management decision to go with it in mid 2015.
I see. If that's the case, then Intel should, as you mentioned in an earlier post, scrap Broadwell and jump straight to Skylake
Isn't their a rumor that Apple is already doing this with the A8 SoC? Well, at the Korean new mongers are saying this, with them saying Samsung is producing up to 40% of the A8 SoCs Apple has shipped.
There has been no proof of this, the tear downs of the SoC to date show TSMC as the manufacturers. My personal belief when we hear that Samsung has 40% of the SOC business it means of all of apple's SoC business so the 4s and 5 and 5s and 5c use chips manufactured by Samsung and this would be true. Anything which apples is still selling and shipping which does not have A8 in is a Samsung chip. In about 18 months time frame when Apple is not selling anything less than an A8 Samsung will be out of the pictures completely.
What is interesting here is that Intels screw ups here far exceed anything AMD has done and yet Intel gets thrown softballs from the press. You have to wonder how much longer this unethical support of Intel will go on.
There's nothing AMD can do to hurt Intel in the near future. As long as Intel is a node ahead and has a competitive design, they are basically in a monopolist position in the laptop, desktop and server world, mistakes are eminently survivable, and a boring media story. And we all know that there's nothing more than the media hates is a boring story.
Intel's problems in the handset world is heavily taxed by their business model and strategy. They can ship the most performant ARM SoC today if they wanted to, but they are sticking with x86 and trying to thread the needle of not shipping a too powerful handset x86 SoC at $30 that won't nuke their $300 processors. 10 years ago, when PocketPC PDAs were the "next big thing", they participated with StrongARM/Xscale, but unfortunately, the near-death experience from Netbust made them pretty conservative on the non-x86 front. If they kept with ARM, they could be the dominate ARM SoC provider today.
Hmm, or they are doubling-down on the stupid. Skylake better be done and provably better than Broadwell in order for them to shorten Broadwell's life cycle. If Skylake is not done, it could mean 2 Holiday seasons where laptops and desktops really haven't gotten anywhere. Intel can make mistakes, but they can't make them forever.
I can see a 5 to 10 W TDP Apple Ax SoC making a very nice laptop or desktop product. Fanless, super thin, and only costing Apple $50 instead of $300 from Intel. The notional, Platonic ideal for the iMac imo is a 10 mm thin device. It's like 30 mm to 40 mm today, but thinning that down further to 10 mm, going with SSDs only, USB 3.1, etc, sounds like a nice machine that looks like it is from the future. These comments on Thunderbolt from the other thread or what have you I/O is immaterial. Thunderbolt is nice, yes, but Apple would drop it with nary a regret if it was holding it back. They can just go with USB 3.1 and HDMI 2.0 or a customized HDMI 2.0 and call it a day.
On the subject of Samsung Semi and 14/16 nm, Apple wants to have more then one vendor be able to manufacture their components. Better yet, three. It's only a good thing. If there was actually a third GPU vendor, Apple would have tried to shore it up and shipped them in their Macs. The two discrete GPU vendors today kind of suck. Not the two companies, but the situation of having only to choose components from 2 vendors. So, it's pretty vital for Apple to have both Samsung Semi and TSMC be able to fab their chips.
Well yeah that is what the road maps say. However AMD has a new management staff in place with a very strong engineering background. These people want to reestablish AMD as a viable processor supplier. I suspect that they will push very hard to get something out on 14 nm before the end of 2015. It might not be K12. It could be an ARM based chip or even something new in the brazos family.
It would do AMD wonders to get something out the door before 2016 that is built on a 14 nm process. It would make investors and customers alike happy. I doesn't have to be a mainstream processor, in just needs to demonstrate viability in AMDs engineering and management teams. I mentioned ARM because it is an easy architecture to port to a new process.
Oh yea, I agree that they're going to push it, just not going to make it pre 2016. 2015 is a critical year for AMD in the launch of its Skybridge initiative to make it's X86 and ARM offerings pin compatible with each other on a socket. Zen and K12's effectiveness is driven hugely by this. Dr Lisa su even says she wants AMD to be mean, lean and fast, but is not willing to rush product launches.
2015 will be Puma+ cores for X86, and a low power 64-bit A57 for ARM. Then by Q3-Q4 we should hear more about Zen and K12, both of which are high frequency yet power efficient designs.
For me? I'm just hoping to see a 16nm GPU from then in 2016. If Zen could at least match Haswell on IPC, then I would also pretty much pick up an AMD CPU to go along with it in my next build. Otherwise it'll be intel+AMD GPU for me.
Samsung is probably about to give a disastrous mid-Q update.
Apple shouldn't use Samsung anymore. Hundreds of workers in their semiconductor fab got leukemia because they were leaking harmful chemicals and Samsung denies that those are work related. Even last they had a gas leak and some worker died. And they are so arrogant that they didn't even report to the police for a whole day. Everybody should boycott any products with Samsung components in them.
Why am I not surprised that the source was from Korea? :roll eyes:
Calm down folks, I'm sure Apple got a great deal.
Surely A7 was a tock with 64-bit introduction, so A9 will be as well?
Why am I not surprised that the source was from Korea? :roll eyes:
Because they accurately reported the A7 will be made by TSMC?
Apple shouldn't use Samsung anymore. Hundreds of workers in their semiconductor fab got leukemia because they were leaking harmful chemicals and Samsung denies that those are work related. Even last they had a gas leak and some worker died. And they are so arrogant that they didn't even report to the police for a whole day. Everybody should boycott any products with Samsung components in them.
Not gonna happen with Apple.
This isn't even something new. The Korean newspapers reported back in Oct. 2013 that Samsung had won back the majority of Apple's 2015 A9 production by virtue of being first to roll out a properly performing 14nm die, well ahead of TSMC. Yeah Apple works on these contracts years in advance.
EDIT: Here you go. Take a look at the date of this even earlier article.
http://english.hankyung.com/news/apps/news.view?c1=&newscate=1&nkey=201307150718011
The Korean newspapers also reported Samsung had won the majority of the A8 production. Those reports were proven false. There is no proof A9 reports are telling the truth. One thing is certain, the reports are succeeding at making Samsung look good until the reports are proven false. When the reports are proven false, new reports surface that Samsung won the next A production.
The last three paragraphs of the so-called report are the most important.
2014/12/12 By Bae Ok-jin
Samsung Electronics began production of 'A9,' the application processor (AP) for Apple's next-generation smartphone. It applies the 14nm FinFET, the cutting-edge microprocess for system semiconductors, for the first time. SEC's foundry business picked up momentum again.
According to industry insiders on December 11, SEC began production of Apple's A9 in the Austin plant in the US using the 14nm FinFET technology. Samsung has production lines capable of FinFET process production in Austin, US and Giheung, Korea, but began to produce A9 only in Austin as it is in the initial stage.
SEC originally said it would start production of the 14nm FinFET chip at the end of this year, but did not disclose whether the company received an order from Apple for the production of A9 chips or whether the production line is actually running. SEC has shown confidence in the yield of the 14nm FinFET process, and supplied samples as good as finished products early enough.
The Austin plant began official production first at Apple's request, and industry insiders said it is a move to produce the chip in the US, not Korea. They guessed that the Austin plant was chosen because of the next-generation chip's problems with performance security and supply.
The initiation of the A9 chip production enabled SEC to recover the foundry quantities from Apple, which have been discontinued for some time, and get the upper hand in the 14nm FinFET technology competition with TSMC, killing two birds with one stone.
Apple and SEC virtually stopped AP production as they were embroiled in patent litigation back in 2012. SEC preoccupation of the 14nm FinFET technology led to resumption of the cooperation between the two tech giants.
However, it's too early to relax yet. Taiwan's TSMC began the risk production of the 16nm FinFET plus (16FF+) process, and began to produce chips in July earlier than originally anticipated Q3, getting ready in a hurry. As Apple is adjusting the A9 chip production quantity shuttling between SEC and TSMC, if TSMC's production line is stabilized in the future, there is no knowing how SEC will respond.
SEC's foundry business was hit hard when Apple orders stopped. Although the entire semiconductor business is booming, securities companies predict that the system LSI business, including the foundry business, will suffer a loss to the tune of KRW800 billion this year. SEC is expected to recover sales loss to a certain extent with the production of Apple A9.
“We said we would inject production wafers when we announced Q3 performance,” said SEC. “But we cannot confirm whether we received orders from Apple or at which plant it will be produced.”
http://english.etnews.com/electronics/2969056_1303.html