[GSMArena] Samsung demonstrates 10nm FinFET for mobile - ahead of Intel

[erunion]

This is getting very disgusting.

Its just marketing. No need to get worked up.

 

[witeken]

Its just marketing. No need to get worked up.

It distorts people's view of the semiconductor industry.

 

[Ajay]

Heck, I can't even tell from the article what device Samsung detailed - could have just been a single Arm 7 core for all I know - which wouldn't be that big a deal.

I think that the main point is that Samsung's Fab development has been reinvigorated, courtesy of the poached top level Fab team from TSMC. There seems to be a renewed focus and, likely, increased funding - which will lead to Samsung becoming a strong number 2 to Intel in process development. Interesting times.

 

[el etro]

Samsung presented the details of its 10nm process at the VLSI symposium and it really is more comparable to Intel 14nm than 10nm. 48nm minimum metal, 64nm gate pitch, 0.052um^2 SRAM cell.

Think of it as a slightly improved Intel 14nm

Yes, is this. That's why the other foundries are ramping up so fast the 10nm processes.

I don't think we will see any foundry catch the one node lead Intel haves.

 

[witeken]

Yes, is this. That's why the other foundries are ramping up so fast the 10nm processes.

I don't think we will see any foundry catch the one node lead Intel haves.

Of course not, they're chasing a moving target, so if it happens, and I don't see why it should, it will be a gradual catch-up, not "Intel had a 1 node lead with 22nm for 3 years, but now suddenly TSMC and Samsung are on par with Intel or just slightly behind at 14nm". Violates all common sense.

 

[Vesku]

Makes me wonder if Samsung managed to poach Intel talent or secrets like they did TSMC.

 

[jpiniero]

Makes me wonder if Samsung managed to poach Intel talent or secrets like they did TSMC.

Now that would not be surprising.

 

[Khato]

Heck, I can't even tell from the article what device Samsung detailed - could have just been a single Arm 7 core for all I know - which wouldn't be that big a deal.

Likewise, none of the articles I found had any actual details... In fact, the original source article sounds more like it's just a PR announcement for their 10nm process, not an actual device. (But that's hard to tell given the shortcomings of auto-translate.)

 

[ElFenix]

Heck, I can't even tell from the article what device Samsung detailed - could have just been a single Arm 7 core for all I know - which wouldn't be that big a deal.

unless there's something specific to the contrary, it's an SRAM. those are always first.

 

[scannall]

It distorts people's view of the semiconductor industry.

The only people paying attention are mostly those that know better anyway. John Q. Public really doesn't understand or care about 10nm or 14nm or gate pitch. Cat videos and arguing with strangers is more important.

 

[jdubs03]

The delay for 14nm definitely hurt the timeframe delay between Intel and Samsung/TSMC. But 10nm at Intel seems to be on the normal schedule, meaning the delay at 14nm won't impact the roll out of 10nm.

So with the information gleaned from Arachno, Samsung will still have the same lag to Intel (projecting a CNL release in Q4-2016/Exynos 9? Q1-2017) as they do now (meaning that Samsung still has accelerated their process development compared to their previous pace, but alas ~2.25 years + a finFET generation behind).

But in terms of marketing I would have to agree with the argument that Samsung calling their next process "10nm" is disingenuous. People unlike us will see that and think that they might usurp Intel as process leader, or at least be on the same level.

But like frozen said, it does come down to perf/watt; and regarding this uArch and other design choices makes a big difference as we all know. I would hazard a guess that by this time next year we will have a clearer picture of where the pieces are going to fall in the design/process world compared to now.

@scannall - youtube: nyan cat, and the view count says it all... we humans are doomed!

 

[DrMrLordX]

It distorts people's view of the semiconductor industry.

scannall beat me to it . . .

Makes me wonder if Samsung managed to poach Intel talent or secrets like they did TSMC.

Ha ha, poor Samsung. They have become the Darloks of the semiconductor industry.

The only people paying attention are mostly those that know better anyway. John Q. Public really doesn't understand or care about 10nm or 14nm or gate pitch. Cat videos and arguing with strangers is more important.

Pretty much. There's probably some investor types that eat up these press releases. ZOMG Samsung demos 10nm process before Intel! News at 11! Seriously though, anyone who can not cut through the bs deserves what they get.

If Samsung's "10nm" process is close to Intel's 14nm process, that does make the game a little bit more interesting. Now let's see what kind of products can be made from said "10nm" process . . .

 

[Vesku]

Pretty much. There's probably some investor types that eat up these press releases. ZOMG Samsung demos 10nm process before Intel! News at 11! Seriously though, anyone who can not cut through the bs deserves what they get.

That's my opinion it's tailored for upper management both to impress them and to be used as material by salesman or project managers. "We plan to stay with X look at this strong roadmap they released." or "Move your business to us if you stay with Y you'll be left behind."

 

[Khato]

unless there's something specific to the contrary, it's an SRAM. those are always first.

But is there even that? The article provided in the OP states that, "Samsung scooped Intel and demoed the first 10nm FinFET chip in the world." yet neither of its source articles say anything to that effect. Nor do any of the other independent original articles I've found. Everything else points to this being nothing more than a PR 'official' announcement of Samsung's pending 10nm process during their keynote at ISSCC.

 

[Idontcare]

Seriously though, anyone who can not cut through the bs deserves what they get.

+1

That's the nuts and bolts of it.

If Samsung's "10nm" process is close to Intel's 14nm process, that does make the game a little bit more interesting.

This is quite literally all that is relevant about the entire topic matter IMHO.

If there is any reality to Samsung's efforts to close the gap with Intel then it will make the "game" (the jockeying for fabless customers) a little more interesting for the fabless customers, and quite possibly for we end-users as well.

It doesn't say anything about the xtor cost delta though, which is something many fabless customers have found to be problematic with 20nm, let alone 16/14nm, to date.

Building the world's first, but entirely cost prohibitive, 10nm node is not exactly what the fabless community is looking for a foundry to provide.

 

[krumme]

Lets see how "14/10"nm devices perform and how many units is sold at what profit. It cuts through all the pr crap.

 

[witeken]

The only people paying attention are mostly those that know better anyway. John Q. Public really doesn't understand or care about 10nm or 14nm or gate pitch. Cat videos and arguing with strangers is more important.

That is not true. Most people who follow tech know that chips are produced on a certain manufacturing node. Most of those people think that a 10nm node will really have 10nm transistors, and that all 14nm nodes are created equal.

They just see the node name, know that lower is better (an advancement on ML), but nothing more. They don't know about feature sizes, transistors or HKMG.

 

[jdubs03]

That is not true. Most people who follow tech know that chips are produced on a certain manufacturing node. Most of those people think that a 10nm node will really have 10nm transistors, and that all 14nm nodes are created equal.

They just see the node name, know that lower is better (an advancement on ML), but nothing more. They don't know about feature sizes, transistors or HKMG.

Most people who follow tech != most people, or the general population. While you're quite right that a subset of tech followers aren't as interested as we are, that doesn't mean that the general population will even care what manufacturing node is associated with what company...

 

[witeken]

Most people who follow tech != most people, or the general population. While you're quite right that a subset of tech followers aren't as interested as we are, that doesn't mean that the general population will even care what manufacturing node is associated with what company...

No one was arguing that.

 

[jdubs03]

No one was arguing that.

Ok, I may have misinterpreted what you were saying.

 

[ElFenix]

That is not true. Most people who follow tech know that chips are produced on a certain manufacturing node. Most of those people think that a 10nm node will really have 10nm transistors, and that all 14nm nodes are created equal.

They just see the node name, know that lower is better (an advancement on ML), but nothing more. They don't know about feature sizes, transistors or HKMG.

Thankfully, none of those people are the customers of fans.

 

[DrMrLordX]

Building the world's first, but entirely cost prohibitive, 10nm node is not exactly what the fabless community is looking for a foundry to provide.

Once all the fluff press releases are done, we'll see how much Samsung wants to talk about their 10nm process. They'll get really quiet about it if they have problems convincing fabless design firms to use their 10nm process node. And that could be awhile.

Are any of the non-Intel fabs actually producing products that are being sold on the open market using any process labelled 16nm, 14nm, or anything else sub-20nm?

 

[Idontcare]

Are any of the non-Intel fabs actually producing products that are being sold on the open market using any process labelled 16nm, 14nm, or anything else sub-20nm?

Nope. Samsung has been talking about their 14nm being in production but to date there are no devices to be bought that contain said chip.

 

[Fjodor2001]

Nope. Samsung has been talking about their 14nm being in production but to date there are no devices to be bought that contain said chip.

Maybe we need to wait 4 more days until MWC where Samsung reveals Galaxy S6 using 14 nm. :sneaky:

 

[Fjodor2001]

Lets see how "14/10"nm devices perform and how many units is sold at what profit. It cuts through all the pr crap.

One could hope. But I guess the losing camp will then start to argue that the chip designs are not equal, and thus not comparable. E.g. one might have a larger iGPU, one might be designed for high clocked, another for high density, etc.

Is there ever any way to settle that dispute? I guess we'd need the same chip designs on two difference process techs, which I likely will never happen.