I'm confused about Intels 14nm process lead

[SOFTengCOMPelec]

I thought that Intels 14nm process, was many years ahead of the competition. So that it would be at least 5 to 7 years before the competitors could put 14nm (approx) onto the market.

Intels main 14nm cpu (desktop) seems to be Skylake, which has not even been released yet (as far as I know, I do know Broadwell is coming out a bit sooner, but Skylake is probably the cpu to aim for, if performance is your primary objective).

So how come the potential competitors (such as Samsung and TSMC), are so close to 14nm large scale production ?

If you are curious, THIS THREAD, initiated my question. I did not put it into that thread, because someone was objecting to mentioning INTEL, in that thread. i.e. I thought it was better to ask the question in a new thread.

Second part to the question. Has Intel possibly had their 14nm process IP "leaked out". Possibly in a similar way to the Samsung/TSMC issue ?

 

[jpiniero]

Samsung and TSMC are doing a bit of process name inflation. TSMC's "16" should really be called "20FF" and Samsung's "14" is actually worse than TSMC's "16" but is further along. It's not really comparable to Intel's 14.

 

[Arachnotronic]

It's basically marketing shenanigans.

Intel's 14nm process features tighter gate pitches, metal pitches, higher drive currents, and probably higher yields at this point in time than TSMC 16FF+.

 

[Phynaz]

It's not about the 14nm label, but the characteristics of the manufacturing process.

One characteristic we know of is clock speed. Nobody else has shown a 14nm process that can manufacture a large CPU that clocks at 4Ghz at a financially viable cost.

Most other things about the processes we will never know about - Cost, yields, binning, wafer throughput, design rules, etc.. What you could call the "secret sauce" of a process.

 

[witeken]

Because node names are irrelevant! What matters are the objective semiconductor feature sizes (approx. 2-3 year lead or 1 node) and transistor characteristics (approx. 3.5 years or 2 nodes). E.g. When Intel put out its 22nm node, first gen FinFET, TSMC had only shortly before released its 1st gen HKMG.

Similarly, TSMC's 1st gen FinFET hits the market in Q1'16 or so, and not long after Intel will release its 1st gen III-V post-silicon. But Intel's lead will increase because 20->16 was shorter than usual because feature sizes stayed flat with only the transistor changed. In other words: FinFETs were pulled in, but I doubt they'll do that a second time.

There are also other things that can be compared, like the interconnect. So I don't think you can just attach one number and say this is Intel's lead; you have to do apples to apples comparisons for the different process node features (although you could take the average, I guess).

 

[SOFTengCOMPelec]

Thanks for all the very rapid responses. I understand it much better now.

Because they (Samsung/TSMC etc) are calling it 14nm, to many people, they will think 14nm = 14nm = the same.

i.e. I'm probably not the only person, who is going to mistake the inferior 14nm processes (from Intels competitors), from Intels much better quality 14nm one.

 

[Arachnotronic]

Because they (Samsung/TSMC etc) are calling it 14nm, to many people, they will think 14nm = 14nm = the same.

i.e. I'm probably not the only person, who is going to mistake the inferior 14nm processes (from Intels competitors), from Intels much better quality 14nm one.

That's exactly what the foundries want and have done a very good job of

 

[Arachnotronic]

Similarly, TSMC's 1st gen FinFET hits the market in Q1'16 or so, and not long after Intel will release its 1st gen III-V post-silicon.

Got a source on the III-V post silicon from Intel?

 

[witeken]

Got a source on the III-V post silicon from Intel?

Sure.

1. Patterns: Intel always does 2 generations before it introduces a new technology.
2. See my signature.
3. http://download.intel.com/newsroom/...esearch_Enabling_Breakthroughs_Technology.pdf As you can see, after FinFET comes III-V. There are also other PDFs out there. It seems III-V (or maybe for some companies SiGe) is generally accepted to be the next step, and 10nm is the obvious node.
4. Paul Otellini in 2009 said silicon was in its last 3 generations before the post-silicon era (= 10nm).

 

[frozentundra123456]

It's not about the 14nm label, but the characteristics of the manufacturing process.

One characteristic we know of is clock speed. Nobody else has shown a 14nm process that can manufacture a large CPU that clocks at 4Ghz at a financially viable cost.

Most other things about the processes we will never know about - Cost, yields, binning, wafer throughput, design rules, etc.. What you could call the "secret sauce" of a process.

Well, Intel hasn't either, really. All we have seen is low voltage mobile processors.

 

[III-V]

Got a source on the III-V post silicon from Intel?

He's estimating. Paul Otellini said in 2009 we'd probably get "post-silicon" by 2017. It's probably fairly node-agnostic as far as insertion time goes (i.e., it's ready when it's ready -- once its issues have been worked out), but may or may not have been pushed out.

Oh, well he already responded.

Personally, I think it'll be around at 7 nm, but it's a subject that doesn't have a clear answer.

I just checked this presentation -- Idsats 4x higher than what Intel's getting on NMOS at 14nm, at a .4V Vgs and .7V Vds. That'll be nice. ~10x higher Fmax.

 

[Fjodor2001]

Ok, so we've all heard the Intel crowd's opinion now.

But what is the reason Samsung decided to call their process tech 14 nm in the first place? Surely they must have some logical explanation for that which they convey to the public (whether you agree with it or not)? Does anyone know how they justify calling it 14 nm? And similar for TSMC 16FF/16FF+?

 

[Exophase]

Got a source on the III-V post silicon from Intel?

I've argued with witeken on this before and I think he's totally wrong about III-V and Germanium being 10nm features. Check out this article:

http://www.eetimes.com/document.asp?doc_id=1263255

10nm is called "effectively done" and these two technologies are discussed as post-10nm strategies. Of course his interpretation is different from mine, I guess draw your own conclusions.

 

[hawtdawg]

Samsung and TSMC's claims are routinely compared to what Intel is shipping, so that's one reason. TSMC couldn't hit a barn with it's release predictions.

 

[oobydoobydoo]

Intel's 14nm process appears to be a disaster. If they had a lead at 22nm, it's gone now that TSMC's 20nm planar and 16nm FF are out.

 

[dahorns]

Intel's 14nm process appears to be a disaster. If they had a lead at 22nm, it's gone now that TSMC's 20nm planar and 16nm FF are out.

From what perspective? They are, for the first time, shipping Core processors that run in fan-less systems. Seems like that is validation of the process on some level.

Also, exactly how is 16nm FF out? Have I missed something?

 

[III-V]

Also, exactly how is 16nm FF out? Have I missed something?

Nope. It's not going to be "out" until 2017. Samsung's 14 nm is in production, though.

Ok, so we've all heard the Intel crowd's opinion now.

But what is the reason Samsung decided to call their process tech 14 nm in the first place? Surely they must have some logical explanation for that which they convey to the public (whether you agree with it or not)? Does anyone know how they justify calling it 14 nm? And similar for TSMC 16FF/16FF+?

It's just marketing. It doesn't always have to make sense.

I've argued with witeken on this before and I think he's totally wrong about III-V and Germanium being 10nm features. Check out this article:

http://www.eetimes.com/document.asp?doc_id=1263255

10nm is called "effectively done" and these two technologies are discussed as post-10nm strategies. Of course his interpretation is different from mine, I guess draw your own conclusions.

I'd think SiGe would be a good interim -- certainly it is easier to use than III-V or Ge in the channel, considering it's already in use in the source and drain.

 

[erunion]

Ok, so we've all heard the Intel crowd's opinion now.

But what is the reason Samsung decided to call their process tech 14 nm in the first place? Surely they must have some logical explanation for that which they convey to the public (whether you agree with it or not)? Does anyone know how they justify calling it 14 nm? And similar for TSMC 16FF/16FF+?

So you were provided an answer, had it confirmed by multiple people, but then asked for a different answer more to your liking.

In the past the foundries released node X with similar metrics to intel's node X, but delayed compared to intel. Now the foundries have shifted their naming schedules. Node X will be released at the same time as Intel's node X, but with inferior metrics.

 

[Fjodor2001]

So you were provided an answer, had it confirmed by multiple people, but then asked for a different answer more to your liking.

In the past the foundries released node X with similar metrics to intel's node X, but delayed compared to intel. Now the foundries have shifted their naming schedules. Node X will be released at the same time as Intel's node X, but with inferior metrics.

All I hear from you is an echo from the Intel crowd. Not something that TSMC or Samsung would use to justify calling their process 14 / 16 FF / 16FF+ nm.

 

[Fjodor2001]

It's just marketing. It doesn't always have to make sense.

So you're saying that the way TSMC and Samsung spins this to the market is that they just selected any random number that matches Intel's process node? I.e. they themselves are not even trying to justify it by any logic or metric at all?

 

[witeken]

I've argued with witeken on this before and I think he's totally wrong about III-V and Germanium being 10nm features. Check out this article:

http://www.eetimes.com/document.asp?doc_id=1263255

10nm is called "effectively done" and these two technologies are discussed as post-10nm strategies. Of course his interpretation is different from mine, I guess draw your own conclusions.

10nm was already effectively done since IDF12, but so what? That's what it means to be 4 years ahead. They're discussed as such because Intel totally wanted to give away its secret roadmap 2 years before they actually intend to do that (irony)!

Intel deeply cares about high transistor performance, certainly because 10nm was developed under P. Otellini's tenure, and since they've been researching III-V for a decade, I will take any other claim with a grain of salt until I see the evidence.

 

[witeken]

Nope. It's not going to be "out" until 2017. Samsung's 14 nm is in production, though.

Are you still confusing TSMC 16 and 10?

 

[III-V]

Are you still confusing TSMC 16 and 10?

Probably confusing it again, yeah. Ugh.

So you're saying that the way TSMC and Samsung spins this to the market is that they just selected any random number that matches Intel's process node? I.e. they themselves are not even trying to justify it by any logic or metric at all?

It's been that way for a little over a decade (since ~130 nm, IIRC). Everybody does it.

 

[witeken]

So you're saying that the way TSMC and Samsung spins this to the market is that they just selected any random number that matches Intel's process node? I.e. they themselves are not even trying to justify it by any logic or metric at all?

TSMC and SS move down a node without density scaling because they've added the precious, magical, 3D Fin Field Effect Transistor, hurray!

But then why didn't they do the same when they introduced HKMG, which is just as revolutionary? They should be at 14nm now with FinFET pushing the number down to 10millimicrons somewhere this year!

The only sensible explanation is that the marketing got pressed by Intel almighty manufacturing prowess. Think about it: the PR knew that Intel would be on 10 for a long time and on the verge of 7nm when TSMC/Samsung would only then move away from 20nm (+FinFET flavour) to their second generation FinFET 14nm (now called 10nm). Intel would/could openly humiliate all other semifailconductor companies for how many nodes they were behind!

Hope that makes sense to you.

In any case, for more knowledgeable people about this topic, semantics won't change reality, for the less knowledgeable, benchmarks also won't be affected by names.

 

[dahorns]

So you're saying that the way TSMC and Samsung spins this to the market is that they just selected any random number that matches Intel's process node? I.e. they themselves are not even trying to justify it by any logic or metric at all?

Maybe? It is a marketing term after all. Absent a definitive statement from Samsung or TSMC, there is no real way to know how they arrived at their naming conventions. But the available evidence points to naming driven more by marketing convenience than driven by process performance/density. Hell, if I remember correctly, the processes (at least for TSMC) were renamed midway through development.

And really, the answer isn't much different from Intel's side of things either. I think Intel has at least provided some justification for its naming convention (i.e., roughly basing it off of the performance/density metrics required to maintain Moore's law). But ultimately the name itself is just a marketing tool.

Finally, why should the name matter? Don't we know enough of the density/performance characteristics of each process to know which one is generally "better"? Or at least better in a particular category? The name is just a distraction at this point.