LG and Intel Developing A Mobile Chipset Using The 14nm FinFET Process?
I confirmed this yesterday. Really good move by LG as they have been in Samsung's shadow for years. Apple started this all by going vertical (making their own SOC) with Samsung and the Chinese smartphone companies following suit. Apple is partnered with ARM and TSMC. Samsung is partnered with ARM and has its own fabs. Huaweii and Xiaomi are "backed" by the Chinese Government and have a lock on the largest smartphone market. LG and Intel makes perfect sense and they can always toss in QCOM chips if they get delayed. This also gives LG a direct path at an integrated modem which will be standard issue in the next year or two.
Last edited:
No way does Intel work with ARM on an integrated SoC solution. The LG phone will be all Intel Inside. They will need that level of differentiation to beat Samsung.
True, but the foundry has to have a license to manufacture the ARM IP. And that is one negotiation I would like to listen to!
I
ippisl
Guest
Who knows, maybe now Intel will succeed. they had good processors, but had bad app compatibility. Since than, they've released and app compatibility layer, and since android 5 google have made changes that also support x86 by default for all apps, so maybe they're ready - and that's why LG decided to invest in that.
This is the next Sofia generation I assume. I think the direction Sofia took is definitely to use Atom cores, not ARM. All the X3, X5 and X7 chips contain Atom cores, even the ultra-low cost X3 which is made at TSMC on a 28nm process. Intel used to produce ARM cores for mobile and embedded applications but sold that business to Marvell in 2006. It wasn't a high enough margin business.
X86 in smartphones performs well in some benchmarks but in the Geekbench they are dogs. I think mobile needs better benchmarks than Geekbench. It goes back to the "difference" between Samsung and TSMC A9 parts--Apple says there is no difference beyond 2-3%. Yet Geekbench benchmarks of A9-Samsung vs. A9-TSMC somehow generate huge differences. That leads me to question Geekbench, not Apple.
So, I think, pending a better benchmark than Geekbench, the world is probably ready at long last for X86 smartphones. Maybe even providing a choice of OS. Does anyone plan to give a Windows 10 smartphone a try?
X86 in smartphones performs well in some benchmarks but in the Geekbench they are dogs. I think mobile needs better benchmarks than Geekbench. It goes back to the "difference" between Samsung and TSMC A9 parts--Apple says there is no difference beyond 2-3%. Yet Geekbench benchmarks of A9-Samsung vs. A9-TSMC somehow generate huge differences. That leads me to question Geekbench, not Apple.
So, I think, pending a better benchmark than Geekbench, the world is probably ready at long last for X86 smartphones. Maybe even providing a choice of OS. Does anyone plan to give a Windows 10 smartphone a try?
Seems unlikely. LG has already invested a lot of money in their Nuclan SoC, which is ARM based: LG Nuclun 2 SoC pushed to 2016
and Atom would be a downgrade to the Cortex A72 that they are going to use in the Nuclan 2.
and Atom would be a downgrade to the Cortex A72 that they are going to use in the Nuclan 2.
hist78
Well-known member
Yes, it makes sense. But does that mean LG will maintain two architectures in parallel? This is probably even more challenging than Apple's multi sourcing practice. But who knows that big companies always have some amazing or crazy ideas that average people like me can't imagine.
This goes back to my comments about benchmarks. Currently, the 2.4GHZ Z3580 outperform all A57, in real-world-based benchmarks (NOT GEEKBENCH). Geekbench is not credible to me.
LG won't even be the first to use A72, it looks like Mediatek will be first. So how do you differentiate? A 2.4GHZ+ 14nm X86 would be differentiated, that would kick ass. If you could run Win10 or Android Marshmallow (or both!) that would be differentiated.
Now that both Apple and Samsung are using ARM architecture licenses:
https://www.semiwiki.com/forum/content/5004-mongoose-making-samsungs-custom-cpu-core.html
LG and the other high end smartphone makers will have to do something custom to keep up in regards to power and performance. Otherwise the cheap Chinese phones using QCOM and Mediatek SoCs will eat their lunch?
Here's my theory about Intel Custom Foundry and ARM chips. I believe ICF is mostly like a pure play foundry except for a few secret rules which include: 1) Do not build AMD chips 2) Do not build Apple chips 3) Do not build ARM chips
I'm considering the argument lefty makes, which is that Altera uses ARM cores as part of the FPGA leading to the conclusion that Intel will soon be producing ARM cores for Altera, and so why not LG as well? My speculation is that Altera ARM-based products are undergoing the same transformation to X86 as Sofia did when Intel acquired that product from Infineon.
Intel has little ARM core expertise, but has abundant X86 core expertise. Intel could only provide a weak ARM-based solution to Altera or LG, which is why they won't.
I'm considering the argument lefty makes, which is that Altera uses ARM cores as part of the FPGA leading to the conclusion that Intel will soon be producing ARM cores for Altera, and so why not LG as well? My speculation is that Altera ARM-based products are undergoing the same transformation to X86 as Sofia did when Intel acquired that product from Infineon.
Intel has little ARM core expertise, but has abundant X86 core expertise. Intel could only provide a weak ARM-based solution to Altera or LG, which is why they won't.
I agree completely. Here's the thing about developing leading edge chips, you need a VERY close relationship with the foundries and IP providers. It is not just a collaboration, it is a joint development agreement (JDA) that will define future products. ARM and TSMC are the best example. I can assure you that Apple is intimately involved in TSMC process and ARM product development. That is why Apple was first to TSMC 20nm and FinFETs. That is also why Apple was first to a 64-bit ARM SoC. And I would bet that when Intel officially acquires Altera, ARM and Xilinx will jump in bed with a JDA as a defensive measure, if they haven't already.
I really don't see Intel having a JDA type relationship with ARM even if they wanted to. I see X86 as the focus for the Intel Custom Foundry business. Otherwise I see little reason for Intel to be in the foundry business. Sound reasonable?
dmehta1759
New member
Don't like this news one bit. Oh boy, more bad news for QCOM.
I don't see Intel and LG coming up viable low power solution soon. If at all, LG will slowly introduce low cost Intel chips into to target its low end market. Would probably be atleast 2017.
I don't see Intel and LG coming up viable low power solution soon. If at all, LG will slowly introduce low cost Intel chips into to target its low end market. Would probably be atleast 2017.
Latest rumour says the deal is off. It seems LG taped out their ARM based Nuclun 2 on both TSMC's 16nm FinFET+ and Intel's 14nm. The Intel version was much faster, but Intel did not get the contract because they didn't have the manufacturing capacity.
Report: TSMC To Manufacture LG's Nuclun 2 SoC, Not Intel | Androidheadlines.com
Report: TSMC To Manufacture LG's Nuclun 2 SoC, Not Intel | Androidheadlines.com
I heard a bit about this at ARM TechCon. ARM and Intel will not work together so if LG insists on using ARM @ Intel it will not happen. And I have no doubt that Intel 14nm can offer more speed than TSMC 16FFC but it cannot match the low power or low cost. 10nm SoCs will tape out next month so it will be Intel 14nm vs TSMC/Samsung 10nm and that will not end well for Intel 14nm in the SoC world, absolutely.
I will be in Taiwan next week so I should know more after that.
hist78
Well-known member
I read that androidheadlines.com article but can't find where is their source for the report. If it's true, then I'd question the future of Intel Foundry business. If an Intel foundry customer needs to compete against Intel's own products for Intel's 14nm capacity, it's not a good sign. Unless Intel is halfhearted on the foundry business?
Another possibility is just like Daniel said that Intel doesn't want to deal with ARM. So Intel uses capacity constraint as an excuse.
Last edited:
By the way, that is how the fabless semiconductor industry started. IDMs sold open fab space to fabless companies until they filled their fabs back up which left an opening for TSMC and UMC. So yes, IDM fabs that also offer foundry business are a risk with capacity. This may also have happened with Samsung and Apple for 14nm. Samsung used capacity to make their own SoCs so Apple had to port the design to TSMC 16nm.