7nm EUV in 2019

[Ajay]

Intel plays with smoke and mirrors much more since they sell finished products, TSMC offers a service and have to be more transparent to customers.

The 76% yield rate is on 256Mbit SRAM modules.

So, that's not exactly a high performance SoC like an an Apple A11 or high TDP like a GPU. That's a very simple test sled (next will be a logic test with a simple ARM core). If TSMC has a high volume, high yield 7nm (GAA?) SoC/GPU using EUV in early 2H18, I'll be impressed. I hope they pull it off, honestly, but we'll see.

 

[Lodix]

So, that's not exactly a high performance SoC like an an Apple A11 or high TDP like a GPU. That's a very simple test sled (next will be a logic test with a simple ARM core). If TSMC has a high volume, high yield 7nm (GAA?) SoC/GPU using EUV in early 2H18, I'll be impressed. I hope they pull it off, honestly, but we'll see.

They have alredy tested a Cortex A72 exceeding 4GHz+ on their HP version of 7nm. The EUV will be used in a improved version in 2019. The company rumoured to use GAA is Samsung ( but their latest press release says they will have 10nmLPP for next year and 10nmLPU for 2019 ).

http://wccftech.com/tsmc-7nm-proces-sram-details/

https://news.samsung.com/global/sam...-finfet-process-technology-production-ramp-up

 

[Ajay]

Well, we will see next year - and see how it compares to Intel's 10nm process.

 

[Gideon]

Samsung EUV 7nm finished 6 months ahead of schedule? will be used for SD 855

 

[jpiniero]

Samsung EUV 7nm finished 6 months ahead of schedule? will be used for SD 855

I thought Qualcomm was using TSMC.

 

[oak8292]

I thought Qualcomm was using TSMC.

Qualcomm does not have exclusive relationships. They will use any capacity available and have products at both Samsung and TSMC.

Here is an announcement from Samsung about the cooperation of Qualcomm on 7nm for baseband processors.

https://news.samsung.com/global/sam...foundry-cooperation-on-euv-process-technology

I don't think there was anything official about Qualcomm moving to TSMC for 7nm capacity. I believe this was a rumor.

 

[iBoMbY]

I didn't understand what the problem with EUV was, until reading this article: Why EUV Is So Difficult

Now that I do, I won't hold my breath until someone actually manages to fully integrate it in a production process.

 

[Ajay]

Damn, bold move by Samsung and Qualcomm!

 

[Andrei.]

Samsung EUV 7nm finished 6 months ahead of schedule? will be used for SD 855

Very unlikely to be used for the S855, they're referring to a standalone 5G modem. Qualcomm would have had to be on board for a long time for the 855 to be Samsung, and everything points out to it being TSMC.

 

[Lodix]

Very unlikely to be used for the S855, they're referring to a standalone 5G modem. Qualcomm would have had to be on board for a long time for the 855 to be Samsung, and everything points out to it being TSMC.

Do you think Samsung has really completed 7nm and will be ready for production for the S10 ? I have my doubts about this untill I see Samsung officially speaking.

Looking at this panorama I would be pissed if Samsung doesn't launch the Snapdragon version in Europe for the S10 instead of the Exynos ( 7nm vs 8LPP, more efficient stock cores and GPU ).

 

[oak8292]

I didn't understand what the problem with EUV was, until reading this article: Why EUV Is So Difficult

Now that I do, I won't hold my breath until someone actually manages to fully integrate it in a production process.

That is a fairly old article and some of the information is not right. TSMC is planning on adding EUV to the second generation of 7nm. However, in a fairly limited role. Samsung is pretty committed to EUV as they are still doing 2D layout and LELELE. This is a pretty big gamble and may not end well if they don't get EUV reliability or yields. Here is a more recent article on the 'stochastic' issue from the same magazine.

https://semiengineering.com/euvs-new-problem-areas/

In the 2016 article there was a 'gentleman' scientist, Chris Mack talking about stochastic and he has a blog that reports from the annual SPIE.

http://life.lithoguru.com

 

[iBoMbY]

The important part is this:

The droplet generator is a small vessel. In operation, tin is loaded into the droplet generator and then heated. At that point, a train of tiny tin droplets flow out from the droplet generator, through a filter and into the vacuum chamber in the source. The droplets are 25 microns in diameter and are falling at a rate of 50,000 times a second.

In the vessel, there is a camera. A droplet passes a certain position in the chamber. Then, the camera tells the seed laser in the sub-fab to fire a laser pulse into the main vacuum chamber. This is called the pre-pulse.

Then comes the really hard part. The pre-pulse laser hits the spherical tin droplet and turns it into a pancake-like shape. Then the laser unit fires again, representing the main pulse. The main pulse hits the pancake-like tin droplet and vaporizes it. “We are trying to hit each droplet twice with the pre-pulse and main-pulse at 50,000 times a second,” ASML’s Lercel said.

At that point, the tin vapor becomes plasma. The plasma, in turn, emits EUV light at 13.5nm wavelengths.

And that's only the light source ...

 

[NostaSeronx]

Overlay improvements using a novel high-order distortion correction system for NIL high-volume manufacturing
https://www.spiedigitallibrary.org/...0.1117/12.2299644.short?SSO=1&tab=ArticleLink

Throughputs of up to 90 wafers per hour were achieved by applying a multi-field dispense method.

For EUVL -> 250W power source = 126 wafers per hour.
So, J-FIL is catching up. So, if Canon goes from four clusters to six clusters, then ya.. J-FIL will be more effective than EUVL.

 

[Andrei.]

S10 ? I have my doubts about this untill I see Samsung officially speaking.

It is official: https://www.samsungsemiblog.com/foundry/samsung-tees-up-the-worlds-first-commercial-euv-chips/

By the second half of this year we expect all the pieces to be in place to start mass-producing the world’s first EUV chips on our new S3 manufacturing line in Hwaseong, Korea.

Many people are confusing the timeline for EUV because of the new fab;

Meanwhile, to provide additional capacity and an alternate manufacturing source, we recently broke ground for a second EUV line in Hwaseong. It should be ready to start production in 2020 at an estimated cost of $6 billion.

There are two timelines here: 2H18 for EUV on the S3 line and 2020 for the new expansion which just recently started to get built.

This is a pretty big gamble and may not end well if they don't get EUV reliability or yields.

Samsung executed their timelines perfectly since 14nm and they won very big on that "gamble". Ever since they had announced EUV some time ago they've said 2H18 and that's approaching very fast now. People said Samsung can't do it in time because the pellicles aren't ready but Samsung is using an in-house pellicle, not ASML's.

In my opinion we'll see the next Exynos on 7LPP with a 8LPP backup project, and Qualcomm is going TSMC 7nm for the 855 for this generation because they didn't want to risk it / Samsung wouldn't have had the capacity.

 

[Lodix]

It is official: https://www.samsungsemiblog.com/foundry/samsung-tees-up-the-worlds-first-commercial-euv-chips/


Many people are confusing the timeline for EUV because of the new fab;

There are two timelines here: 2H18 for EUV on the S3 line and 2020 for the new expansion which just recently started to get built.


Samsung executed their timelines perfectly since 14nm and they won very big on that "gamble". Ever since they had announced EUV some time ago they've said 2H18 and that's approaching very fast now. People said Samsung can't do it in time because the pellicles aren't ready but Samsung is using an in-house pellicle, not ASML's.

In my opinion we'll see the next Exynos on 7LPP with a 8LPP backup project, and Qualcomm is going TSMC 7nm for the 855 for this generation because they didn't want to risk it / Samsung wouldn't have had the capacity.

But Samsung's meaning of "mass production" is just risk production in 2H2018... I still don't think it will be ready for Q1 launch of the S10.

Also, do you think UFS 3.0 or LPDDR5 will be ready for the Note 9 or Huawei Mate 20 ??

 

[Andrei.]

But Samsung's meaning of "mass production" is just risk production in 2H2018... I still don't think it will be ready for Q1 launch of the S10.

Samsung's meaning of mass production has always been mass production.

 

[witeken]

Samsung's meaning of mass production has always been mass production.

Samsung has said they will enter 2nd gen 10nm and 8nm production in 2018, and 7nm risk production. To me that implies 7nm in H2'19.

 

[Lodix]

Samsung's meaning of mass production has always been mass production.

I swear remembering this happening before, but never the less if they achieve this it will be a great feat. But aren't their PPA numbers a bit "low" for a EUV node ? At least compared to TSMC's 7FF.

 

[Andrei.]

Samsung has said they will enter 2nd gen 10nm and 8nm production in 2018, and 7nm risk production. To me that implies 7nm in H2'19.

I just quoted them saying 7nm mass production in 2H18.

https://news.samsung.com/global/sam...2nd-generation-10nm-finfet-process-technology
November 29th 2017 -> Snapdragon 845 / Exynos 9810 Galaxy S9 1Q18

https://news.samsung.com/global/sam...m-on-chip-with-10-nanometer-finfet-technology
October 17th 2016 -> Snapdragon 835 / Exynos 8895 1Q17

https://news.samsung.com/global/sam...-14-nanometer-finfet-logic-process-technology
January 14th 2016 -> Snapdragon 820 / Exynos 8890 1Q16 (They actually had been producing since November/December, PR lagged behind here)

http://www.samsung.com/semiconducto...rst-14nm-finfet-mobile-application-processor/
February 16th 2015 -> Exynos 7420 1Q15 (Again pretty late announcement given that the phones were showcased 2 weeks later)

So again

By the second half of this year we expect all the pieces to be in place to start mass-producing

By that statement I fully expect the next Exynos in 1Q19 to be very likely 7LPP.

Maybe they'll have a 8nm refresh mid-year this year for the Note 9 and that's where those claims come from, but I really doubt this.

 

[ksec]

Well...

You cant really define mass production as mass production for Apple is different to say mass production to any one else. Similar to TSMC has 10nm mass production for Huawei shipped before iPad Pro and iPhone X launch. They are still mass production and not risk production, just smaller qualities. Samsung's mass production is similar as well.

 

[oak8292]

Something will have to give. Either the 855 is on TSMC 7nm or Apple A12 is on 7nm but I really doubt that both are on 7nm. I think Apple probably has most of the TSMC 7nm capacity and the rumors for Qualcomm at TSMC are for the server processors. TSMC has a high performance process and this will be used by Qualcomm. Pure speculation. However, I really doubt that TSMC has adequate 7nm capacity for both Qualcomm 855 and Apple A12.

 

[Andrei.]

You cant really define mass production as mass production for Apple is different to say mass production to any one else. They are still mass production and not risk production, just smaller qualities. Samsung's mass production is similar as well.

Mass productions is simply line production for commercial devices, be that millions, tens of millions, or whatever Apple sells.

Similar to TSMC has 10nm mass production for Huawei shipped before iPad Pro and iPhone X launch.

Actually Huawei only launched in October while Apple had already shipped millions by then. MediaTek with the X30 was first along with the A10X but that was a commercial failure.

 

[oak8292]

Samsung executed their timelines perfectly since 14nm and they won very big on that "gamble". Ever since they had announced EUV some time ago they've said 2H18 and that's approaching very fast now. People said Samsung can't do it in time because the pellicles aren't ready but Samsung is using an in-house pellicle, not ASML's.

It is always a little hard to tell if Samsung is executing well. The memory business revenues can hide all sorts of 'sins' in the foundry business. The yields on 10nm could have been horrible and it would be hard to tell. The node name in mass production is more important than the actual cost of a transistor.

With 7nm EUV they have committed to using the technology in a lot of steps and even with an in-house pellicle this does not solve all of the issues around EUV. Has the equipment reliability issue been resolved? Do they have their own resists for higher throughput? Who eats the cost on low yields? They give the appearance with their announcements that they are doing well. Hopefully they are.

 

[lopri]

I just quoted them saying 7nm mass production in 2H18.

https://news.samsung.com/global/sam...2nd-generation-10nm-finfet-process-technology
November 29th 2017 -> Snapdragon 845 / Exynos 9810 Galaxy S9 1Q18

https://news.samsung.com/global/sam...m-on-chip-with-10-nanometer-finfet-technology
October 17th 2016 -> Snapdragon 835 / Exynos 8895 1Q17

https://news.samsung.com/global/sam...-14-nanometer-finfet-logic-process-technology
January 14th 2016 -> Snapdragon 820 / Exynos 8890 1Q16 (They actually had been producing since November/December, PR lagged behind here)

http://www.samsung.com/semiconducto...rst-14nm-finfet-mobile-application-processor/
February 16th 2015 -> Exynos 7420 1Q15 (Again pretty late announcement given that the phones were showcased 2 weeks later)

So again
By that statement I fully expect the next Exynos in 1Q19 to be very likely 7LPP.

Maybe they'll have a 8nm refresh mid-year this year for the Note 9 and that's where those claims come from, but I really doubt this.

I just quickly glanced through the titles and two of them (related to 10nm) begin with "Samsung starts mass production," and the other two (related to 14nm) with "Samsung announces mass production." All of them are official press releases. On the contrary, the 7nm stuff is 1) a blog post, and 2) a prediction instead of an announcement after the fact, and 3) by necessity worded quite differently, i.e. "we expect all the pieces to be in place to start mass-producing.." It's a loaded expression with a degree of caution. I would not presume it in either direction.

 

[Andrei.]

1) a blog post, and 2) a prediction instead of an announcement after the fact, and 3) by necessity worded quite differently, i.e. "we expect all the pieces to be in place to start mass-producing.." It's a loaded expression with a degree of caution. I would not presume it in either direction.

It being a blog post is irrelevant, it's their official blog and it's posted by the senior director of marketing of Samsung semiconductor. They post official news / plans in blog format there. As for the cautionary wording, well that's obvious, nobody is going make more precise statements than "we expect to" more than half a year ahead with products 11 months ahead, the foundry might as well burn down in that time or something.