• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

[DigiTimes] TSMC 10 nm trial production in 2015, mass production in 2016

Page 13 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Idontcare said:
Intel doesn't report financials broken out by process node, so we really don't have comparable data that would speak to how good or bad their 14nm ramp is going beyond the subjective claims of Intel management.

My subjective opinion is that TSMC is probably 18 months behind Intel in terms of delivering a transistor with comparable Ion/Ioff at the same operating voltage, leakage and reliability. And they are even farther behind in delivering air-gap interconnect at the design rules Intel is using for 14nm.

Who cares though? It only matters if (1) Intel decides to get serious about taking on TSMC as a foundry, and/or (2) Intel is able to use its process advantage in a meaningful way to compete with TSMC's fabless customers.

I hear a lot of commotion in regards to both those bullet points, and the opportunity certainly exists, but Intel has to implement and execute to get anything done...and 14nm isn't exactly a shining step forward of implementing and executing to date.

Something tells me TSMC's fabless customers will continue to do just fine in the business realm despite the lag between their foundry's process nodes and those of Intel's.
Click to expand...

Great points. Intel really has made poor use of its process technology in the mobile chip market.
 
III-V said:
It only has to do with Apple becoming TSMC's customer.

That's because it's not powered by the Snapdragon 810 😉
Click to expand...


So I don't get it, what allows for such low TDP in the Air 2 if it's not the process? And why is Qualcomm's new chip running so hot?


From what I gather you're saying that Apple's A8/X revenue stream allowed for that graph to be made (revenue/time x nm ramp) and TSMCs 20nm process has nothing to do with why Apple's new chips run so cool? I guess it's possible.
 
Idontcare said:
Intel doesn't report financials broken out by process node, so we really don't have comparable data that would speak to how good or bad their 14nm ramp is going beyond the subjective claims of Intel management.

My subjective opinion is that TSMC is probably 18 months behind Intel in terms of delivering a transistor with comparable Ion/Ioff at the same operating voltage, leakage and reliability. And they are even farther behind in delivering air-gap interconnect at the design rules Intel is using for 14nm.

Who cares though? It only matters if (1) Intel decides to get serious about taking on TSMC as a foundry, and/or (2) Intel is able to use its process advantage in a meaningful way to compete with TSMC's fabless customers.

I hear a lot of commotion in regards to both those bullet points, and the opportunity certainly exists, but Intel has to implement and execute to get anything done...and 14nm isn't exactly a shining step forward of implementing and executing to date.

Something tells me TSMC's fabless customers will continue to do just fine in the business realm despite the lag between their foundry's process nodes and those of Intel's.
Click to expand...

18 months? So 22nm is on par with 28HP planar transistors? You've got to joking...

I don't see how Qualcomm and MediaTek are going to stay competitive with such a big disadvantage if Intel's aiming to provide the absolute state of the art in terms of feature set and performance, like they do in the PC space. SoFIA, which will deliver unparalleled performance in the low-end phone space, is the first sign of that.

It of course depends on what you see as "doing fine", but for Qualcomm it can only go downhill.
 
witeken said:
18 months? So 22nm is on par with 28HP planar transistors? You've got to joking...

I don't see how Qualcomm and MediaTek are going to stay competitive with such a big disadvantage if Intel's aiming to provide the absolute state of the art in terms of feature set and performance, like they do in the PC space. SoFIA, which will deliver unparalleled performance in the low-end phone space, is the first sign of that.

It of course depends on what you see as "doing fine", but for Qualcomm it can only go downhill.
Click to expand...

For example Snapdragon 615 and MT6752 are cheap and seems to perform pretty well with just a bunch of Cortex A53, and they are on the market today, with LTE modem included.
So this is Intels problem, they are always too late when it comes to smartphones. There are already cheap complete SOC's available that gives a smooth user experience. I doubt that SoFIA 3G/4G will make a big impact on the market. Unless Intel sell them at a loss though, that will keep the volume high of course.
 
teejee said:
For example Snapdragon 615 and MT6752 are cheap and seems to perform pretty well with just a bunch of Cortex A53, and they are on the market today, with LTE modem included.
So this is Intels problem, they are always too late when it comes to smartphones. There are already cheap complete SOC's available that gives a smooth user experience. I doubt that SoFIA 3G/4G will make a big impact on the market. Unless Intel sell them at a loss though, that will keep the volume high of course.
Click to expand...

615 has a ton of problems.
 
witeken said:
I don't see how Qualcomm and MediaTek are going to stay competitive with such a big disadvantage if Intel's aiming to provide the absolute state of the art in terms of feature set and performance, like they do in the PC space. SoFIA, which will deliver unparalleled performance in the low-end phone space, is the first sign of that.
Click to expand...

First gen SoFIA isn't going to change anything. It leaves Intel dealing with the inherent x86 efficiency problems without any process advantage to help them, and any Android apps with native ARM binaries are going to run pretty poorly on it.

Maybe the 2nd gen on 14nm can turn things around, if they ship the thing in a timely fashion.
 
TreVader said:
So I don't get it, what allows for such low TDP in the Air 2 if it's not the process? And why is Qualcomm's new chip running so hot?


From what I gather you're saying that Apple's A8/X revenue stream allowed for that graph to be made (revenue/time x nm ramp) and TSMCs 20nm process has nothing to do with why Apple's new chips run so cool? I guess it's possible.
Click to expand...
The process is certainly part of it, but it's ultimately up to Apple how "cool" their chip runs. You could have the same chip on 28nm running just as cool... if you clocked it really low.

How low the processor's TDP is is a rather meaningless number in and of itself. It only matters when you look at the performance and cost as well.
NTMBK said:
Really? First I heard about that- I thought it was the 810 that was having problems?
Click to expand...
They're the same silicon, I think.
NTMBK said:
First gen SoFIA isn't going to change anything. It leaves Intel dealing with the inherent x86 efficiency problems without any process advantage to help them, and any Android apps with native ARM binaries are going to run pretty poorly on it.

Maybe the 2nd gen on 14nm can turn things around, if they ship the thing in a timely fashion.
Click to expand...
Silvermont should look pretty nice, even on TSMC's process. It'll be a bit dated, as Bay Trail already is, but since it's a more budget-friendly offering, it should do fine.
 
Last edited:
witeken said:
18 months? So 22nm is on par with 28HP planar transistors? You've got to joking...

I don't see how Qualcomm and MediaTek are going to stay competitive with such a big disadvantage if Intel's aiming to provide the absolute state of the art in terms of feature set and performance, like they do in the PC space. SoFIA, which will deliver unparalleled performance in the low-end phone space, is the first sign of that.

It of course depends on what you see as "doing fine", but for Qualcomm it can only go downhill.
Click to expand...

The problem is that Intel keeps "aiming" but it doesn't actually hit the target 😉

I think the 14nm mobile generation should be pretty strong for Intel if it executes...it's whether it actually will that's up in the air.
 
Arachnotronic said:
The problem is that Intel keeps "aiming" but it doesn't actually hit the target 😉

I think the 14nm mobile generation should be pretty strong for Intel if it executes...it's whether it actually will that's up in the air.
Click to expand...
It'd already be very strong if Intel were actually delivering. Right now, Intel should have their 14nm phone SoCs on the market... not just barely delivering their 22nm ones.
 
NTMBK said:
First gen SoFIA isn't going to change anything. It leaves Intel dealing with the inherent x86 efficiency problems without any process advantage to help them, and any Android apps with native ARM binaries are going to run pretty poorly on it.

Maybe the 2nd gen on 14nm can turn things around, if they ship the thing in a timely fashion.
Click to expand...

There's no such thing as x86 efficiency problems. Silvermont is a stellar architecture.
 
III-V said:
It'd already be very strong if Intel were actually delivering. Right now, Intel should have their 14nm phone SoCs on the market... not just barely delivering their 22nm ones.
Click to expand...

The good thing, though, is that once they have their offerings up to snuff -- i.e. they have successfully entered the market with a healthy product pipeline like you see with Core in the server, laptop and desktop space -- then we'll see the iterations coming at a normal pace without any delays, utilizing their full process advantage from top to low-end bottom.

I mean: in a bit more than a year from now (if Broxton and SoFIA 2 and MID launch at CES), Bay Trail won't exist anymore. All products will be on 14nm, the platforms will be in place, there will be no BOM deficiencies and contra-revenue, and the roadmap will be full for the foreseeable future, with 10nm and new architectures following quickly, and with the modem on-par with QQ.

Brian Krzanich took control only 20 months ago and SoFIA's been on the roadmap for even less time.
 
Last edited:
witeken said:
There's no such thing as x86 efficiency problems. Silvermont is a stellar architecture.
Click to expand...
There certainly is such a thing, it's just well-mitigated these days.
witeken said:
The good thing, though, is that once they have their offerings up to snuff -- i.e. they have successfully entered the market with a healthy product pipeline like you see with Core in the server, laptop and desktop space -- then we'll see the iterations coming at a normal pace without any delays, utilizing their full process advantage from top to low-end bottom.
Click to expand...
Hopefully.
 
Last edited:
witeken said:
18 months? So 22nm is on par with 28HP planar transistors? You've got to joking...
Click to expand...

Just guessing, but maybe IDC was looking ahead when he said that Intel is approximately 18 months ahead of TSMC? I.e. comparing the state of TSMC 20/16FF/16FF+/10 nm with Intel 14/10 nm?
 
witeken said:
There's no such thing as x86 efficiency problems. Silvermont is a stellar architecture.
Click to expand...

Stellar on 22nm FinFET. Doubt it's going to do so well on 28nm planar. You're the one always telling us how magical Intel's transistors are 😉

Should be fun... We get to compare Intel's arch with ARM cores on the same process, and compare TSMC 28nm with Intel 22nm on the same architecture. Finally its apples to apples!
 
NTMBK said:
Stellar on 22nm FinFET. Doubt it's going to do so well on 28nm planar. You're the one always telling us how magical Intel's transistors are 😉

Should be fun... We get to compare Intel's arch with ARM cores on the same process, and compare TSMC 28nm with Intel 22nm on the same architecture. Finally its apples to apples!
Click to expand...

With stellar I mean high dynamic range, something pretty much all ARM cores ever except Krait lack.
 
witeken said:
18 months? So 22nm is on par with 28HP planar transistors? You've got to joking...

I don't see how Qualcomm and MediaTek are going to stay competitive with such a big disadvantage if Intel's aiming to provide the absolute state of the art in terms of feature set and performance, like they do in the PC space. SoFIA, which will deliver unparalleled performance in the low-end phone space, is the first sign of that.

It of course depends on what you see as "doing fine", but for Qualcomm it can only go downhill.
Click to expand...

Was talking 10nm, thread title
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top