Apple & Samsung could be only smartphone makers with 7nm chips in 2018
Apple and Samsung could potentially become the only smartphone vendors and chip designers to adopt 7-nanometer processors for 2018 phones, owing to the costs involved, a report noted on Wednesday.
The slowdown of smartphone industry growth is pushing down chip prices, in turn forcing chip suppliers to "seriously weigh in" on the cost of new chipsets, DigiTimes sources said. Higher wafer foundry costs might eat into declining profit margins.
Sources estimate that a chipset manufacturer may need to ship as many as 120 million to 150 million 7-nanometer chips just to break even on development. As the world's biggest smartphone vendors, Apple and Samsung could be the only ones ordering on that scale -- even Chinese giant Huawei is allegedly having trouble catching up in processor size because of cost issues.
Qualcomm and MediaTek could also be in a position to order 7-nanometer chips, but those companies simply develop processors rather than make complete devices. DigiTimes added that Qualcomm is sticking to Samsung's 10-nanometer process for its Snapdragon 845 processor, while MediaTek is hanging on to TSMC's 12- and 16-nanometer nodes.
Apple is typically quick to adopt more efficient chips, looking to simultaneously shrink devices and improve power efficiency even as it adds new features. The A11 Bionic processor in the iPhone 8 and X uses TSMC's 10-nanometer process, and multiple reports have pointed to a 7-nanometer "A12" appearing in next year's iPhones. Those models may include 5.8- and 6.5-inch OLED devices and a 6.1-inch LCD product.
The slowdown of smartphone industry growth is pushing down chip prices, in turn forcing chip suppliers to "seriously weigh in" on the cost of new chipsets, DigiTimes sources said. Higher wafer foundry costs might eat into declining profit margins.
Sources estimate that a chipset manufacturer may need to ship as many as 120 million to 150 million 7-nanometer chips just to break even on development. As the world's biggest smartphone vendors, Apple and Samsung could be the only ones ordering on that scale -- even Chinese giant Huawei is allegedly having trouble catching up in processor size because of cost issues.
Qualcomm and MediaTek could also be in a position to order 7-nanometer chips, but those companies simply develop processors rather than make complete devices. DigiTimes added that Qualcomm is sticking to Samsung's 10-nanometer process for its Snapdragon 845 processor, while MediaTek is hanging on to TSMC's 12- and 16-nanometer nodes.
Apple is typically quick to adopt more efficient chips, looking to simultaneously shrink devices and improve power efficiency even as it adds new features. The A11 Bionic processor in the iPhone 8 and X uses TSMC's 10-nanometer process, and multiple reports have pointed to a 7-nanometer "A12" appearing in next year's iPhones. Those models may include 5.8- and 6.5-inch OLED devices and a 6.1-inch LCD product.
Comments
Maybe they'll convince Intel to include their modem logic onboard? I believe right now these are discrete parts and, thus, consume more real estate and more power, since they're very likely not 7nm parts nor will be next year).
But at the end of 2018 we may see Kirin 980 at 7nm too. Transistor count for Kirin 970 is already 5.5 billion and chip size is ~97mm² - that's very big for a non-Apple SoC.
I think HiSilcon will aim for that density improvement 7nm has to offer.
@Tjwolf ;
The modem doesn't have to be integrated on die - Apple could integrate the modem into the PoP (SoC + RAM + modem). Intel is rumoured to bring production of the next-gen modem inhouse (10nm process, same desity as TSMC's 7nm).
Samsung, Globalfoundries and Intel are all trying to woo HiSilicon to be that second supplier. Rumours say Samsung is throwing OLED and other IP into the offer to sweeten the deal.
The Kirin 980 is apparently very near finished so no, they are not reverse engineering anything. Your mention of that was no doubt tongue in cheek. There are also rumours of a HiSilicon designed GPU that might not be far off either.
We'll see.
To make reverse engineering the A-11 worthwhile it would also be necessary to reverse engineer iOS and all other supporting chips.
Ain't going to happen.
Bwahaha
I'm assuming you're joking, right?
I don't think it's really necessary for most consumers to have a need for 7nm-based smartphones. I'm sure the 10nm SoCs are good for another year or two for most smartphone manufacturers selling mainly low- and mid-range smartphones. Most current flagships have more than enough power for most tasks consumers need to perform. Apple is probably the only company that can easily afford to keep pushing forward in chip density and probably has the most to gain from it.
I don't think Samsung would have a 7nm SoC while Qualcomm having a 10nm SoC in Android world in the same year. Samsung still needs Qualcomm SoCs in their international variants of Flagship phones (read S and Note series) for USA and China. Exynos SoC is used only in Korea, India and few other countries. Year 2015 was an exception when Samsung used Exynos 7420 all over the world, due to over-heating problems with SD 810. But that was an exception, not the norm. Samsung is again back to using Qualcomm SoCs for USA and China. It is going to 10nm SoCs in 2018 for ALL Android OEMs, including Samsung.
Apple would most likely be the first one to move to 7nm SoCs, either in 2018 or in 2019, because they only can invest so much in R&D costs AND can get the ROI due to the volume of high-end phone shipments. SoCs used in Android phones will be far behind Apple SoCs going forward.
https://www.notebookcheck.net/Intel-Globalfoundries-and-Samsung-reportedly-competing-for-contract-to-produce-Huawei-s-Kirin-ARM-SoCs.259548.0.html
Apple needs as much power in the iPhone as it can get. A corollary to their stance on privacy is that a lot more intelligence has to reside on the device. The neural engine silicone they added in the A11 is just the beginning. Google, which makes 95% of its profits by 'knowing' its users and selling targeted ads based on that knowledge wants to do most intelligent functions on their servers. So Android phones don't have quite as pressing a need for better silicone.
We tend to associate them with Huawei and Honor Mobile phones and tablets but they are present in many other areas ranging from general communications, surveillance cameras and storage systems, displays, set top boxes, networking devices and more. In terms of units and the smallest edge of the wedge in terms of sales, Kirin's mounted on reference boards are very popular.
https://www.96boards.org/products/
A 970 board is on the cards.
I would say that apart from the contracted capacity they have with their current supplier, the fact that they seem to be looking to diversify suppliers and that candidates appear to be pitching attractive offers to HiSilicon perhaps means they have high expectations.
The latest Kirin970 wasn't actually developed with the latest technology available. They deliberately chose more mature but optimised ARM technology. The clear focus was on efficiency and the NPU.
It might be that the 980 is a little closer to release than some think. Then there are rumours of an in house GPU.
I really don't know how things will play out.
In terms of sheer unit sales numbers, Huawei overtook Apple mid year 2017 and has repeatedly stated that it wants to be the world's number one producer around 2020/21.
Seeing how quickly the top end 970 arrived on the Honor (lower priced platform), it looks like unit sales of 970 SoCs might scale faster than was first imagined. I also wonder if they plan to stop using Snapdragon chips on some models if Kirin output can be increased.
I was not aware of the other uses of Kirin SoCs, apart from smartphones (Mate, P and Honor series). Thanks for sharing the info.
In terms of sheer unit sales numbers, Huawei overtook Apple mid year 2017 - Yes, but the majority of them were due to mid-range Kirin SoCs (Kirin 655/659 in Honor mid-range phones) and few Snapdragon mid-range devices (Nova series come to mind). Sales number for phones with Kirin 960/970 (their top tier SoCs in 2016/17 - Mate 9/9 Pro/10/10 Pro, Honor 8 Pro/9, view 10, P10/P10 Plus) - All of them put together would come around 20-30 million at the max. And this has a huge negative impact on Samsung, not Apple. So my question is - why would Samsung want to HELP their biggest rival (even more than Apple) since they are eating into their core business (devices in $300 to $600 range) all over the world.
I also wonder if they plan to stop using Snapdragon chips on some models if Kirin output can be increased. - They have already literally STOPPED using Snapdragon SoCs in their models. 9 out of 10 models launched by Huawei have Kirin SoCs only. And they are using Snapdragon SoCs only in mid-range phones ($300 to $400). They are NOT using Snapdragon in SoCs in high-end phones (Mate, P series) at all from the beginning. So they have already maxed out on the number of phones which can use their latest and greatest SoCs. Growth, if any, comes only with mid-range priced (value for money flagship) Honor series phones. Not sure, how this would help in increasing their already meager profits, justifying the additional investment on the latest process nodes for SoC manufacturing.
It might be that the 980 is a little closer to release than some think. - I don't think so. Kirin 980 will be out only next year Sep-Oct in its usual timeframe, with the next Mate flagship being the first one to use it. I don't think so that Huawei will use the Kirin 980 SoC for P series next year (with Mate and Honor series phones already out with Kirin 970).
"Qualcomm and MediaTek could also be in a position to order 7-nanometer chips, but those companies simply develop processors rather than make complete devices."
Huh? That doesn't mean anything. Quite the contrary, the opposite is true. Qualcomm doesn't need to rely on a single vendor to move 120 million 7 nm chips. So long as Samsung, LG, Motorola, Google, Xiaomi, Oppo etc. all buy the chips then they make a profit regardless of who makes the phone. In fact, they don't need the phones to sell to make a profit. If LG buys 30 million Qualcomm chips and only sells 25 million LG G8s and LG V30s, that is LG's loss. Qualcomm gets paid the same. So this analysis makes no sense.
Qualcomm could have gotten the 7 nm design this year if they had prioritized it. Instead, they prioritized getting the AI, AR, VR, HDR, graphics and biometrics features better, as well as increasing efficiency with an existing 10 nm design. Meaning that they focused on new features. Had they focused merely on getting to 10 nm, the new features would have had to wait. And it was the right decision. Smartphone owners don't care much about 10 nm versus 7 nm, especially when the new 10 nm design will improve power consumption (as will the new release of Android). But they do care about better cameras, video recording etc.