TSMC 7nm info

[name99]

So Glofo is basically a second rate company making a second rate process.

Or, more reasonably, GloFo is optimizing for different things than TSMC is optimizing for. GloFo's big areas are supporting a certain type of server chip (ie IBM), and supporting RF. Given finite resources, doing that means they can't support leading-edge mobile as well as a company (TSMC) that has prioritized it above everything else...
That doesn't mean they're second rate at either their RF stuff, or their IBM support stuff. It's also possible (jury is still out) that their 22nm SOI stuff may be a very sweet spot for a wide range of wearables and IoT chips that want low-power (and perhaps even on-board wireless), but that can't afford the commitment required by the next move down, to FinFETs.

The one thing GloFo IS second at is second largest foundry in the world. That suggests they're not totally incompetent.

 

[name99]

I think by now it's clear, the node name each foundry uses is decided by marketing not engineering. With that said, what's the performance characteristics and implications. Consider this, the actual processor cores take very little space on the SoC even using 16FFC. The elephant in the room is the logic board and replacing the substrate with wafer level integration is far more of a story than going to 7nm. So, my only interest in 10nm and 7nm would be the the performance/power ratio improvements at each node, not the actual density.

I'd like to see a chart comparing that between each foundry's node.

I don't think this ("the node name each foundry uses is decided by marketing not engineering") is completely true. I have suggested, for example, that fin width can be considered as *somewhat* tracking the node number. I've reason to believe this is correct, but it's hard to be sure.
Intel 14nm fins are 8nm wide. I have never seen numbers for the width of anyone else's fins.

I suspect that the reason companies are so happy to release the sorts of numbers that are being quoted in this thread is precisely that they are not especially useful. The numbers that likely DO matter the most (among which, I suspect, are fin width) are the numbers that we never see quoted. (Or are quoted for the research version, not the manufacturing version, or are quoted three years after the manufacturing process was introduced.)

But I do agree with the larger point --- there are few purposes these days for which density is vastly interesting. Tweaks to the process (the upcoming tweaks appear to SiGe and III-V modifications to the transistors, and maybe air gaps for TSMC --- Intel already has them, and I assume IBM will demand them), and packaging seem more significant issues to track for the purposes of interest to most of us.

 

[raghu78]

GloFo's 7nm looks like it's more like TSMC's 10nm. Basically, they named it 7nm to make it look like they're on the same playing field as TSMC when they're really just posers who have yet to develop (successfully) a post-28nm node.

You were way off mark there. GF 7LP is actually slightly more denser than TSMC N7.

https://www.semiwiki.com/forum/cont...alfoundries-discloses-7nm-process-detail.html

GF 7LP has higher performance than TSMC N7 HPC and is the only foundry process which is optimized for 5 Ghz operation. Both GF 7LP and TSMC N7 beat Intel 10nm for low power mobile process due to the availability of a 6T library. For high performance Intel 10nm is denser by roughly 13% assuming same track count.High performance GF 7LP uses 9T library . Intel 10nm standard library is 7.56T but I am not sure if their highest performance CPUs use 7.56T or 9T.

 

[wingman04]

How much smaller can they go below 7nm?

 

[raghu78]

GloFo's 7nm looks like it's more like TSMC's 10nm. Basically, they named it 7nm to make it look like they're on the same playing field as TSMC when they're really just posers who have yet to develop (successfully) a post-28nm node.

Just wanted to laugh out loud on this prediction. lol.

https://www.semiwiki.com/forum/content/7191-iedm-2017-intel-versus-globalfoundries-leading-edge.html

GF 7nm is shaping up to be one of the best foundry 7nm nodes. Very competitive on technical specs - Power, Perf, Area, SRAM cell size. Now what remains is for GF to deliver to execute a steady yield learning curve and ramp to HVM by early 2019 and then it will be settled once and for all that GF is truly a world class leading edge foundry.

 

[goldstone77]

Just wanted to laugh out loud on this prediction. lol.

https://www.semiwiki.com/forum/content/7191-iedm-2017-intel-versus-globalfoundries-leading-edge.html

GF 7nm is shaping up to be one of the best foundry 7nm nodes. Very competitive on technical specs - Power, Perf, Area, SRAM cell size. Now what remains is for GF to deliver to execute a steady yield learning curve and ramp to HVM by early 2019 and then it will be settled once and for all that GF is truly a world class leading edge foundry.

Honestly, I don't see what's wrong with that statement if the information being provided by IEDM is accurate. If they pull this off that changes my opinion of them, and gives me more confidence in their reliability moving forward. They have everything riding on having good execution at 7nm. It's a make it or break it situation in my mind. They simply can't afford to fail at 7nm. They still have a number of years to go on the 10 year agreement with IBM, but AMD is free in 2020.

 

[itsmydamnation]

Honestly, I don't see what's wrong with that statement if the information being provided by IEDM is accurate. If they pull this off that changes my opinion of them, and gives me more confidence in their reliability moving forward. They have everything riding on having good execution at 7nm. It's a make it or break it situation in my mind. They simply can't afford to fail at 7nm. They still have a number of years to go on the 10 year agreement with IBM, but AMD is free in 2020.

I'm sure I've seen comments sayings that on 7nm SOC they exceeded there performance targets, god knows where i saw the quote but.....

So we will just have to wait and see.

actually quick google found this :

https://www.eetimes.com/document.asp?doc_id=1331887
I dont think GF 7nm will be the best process, but i dont think it will be very far behind, certainly a lot closer then 14nm LLP compared to in 14nm and TMSC 16nm........

 

[goldstone77]

I'm sure I've seen comments sayings that on 7nm SOC they exceeded there performance targets, god knows where i saw the quote but.....

So we will just have to wait and see.

actually quick google found this :

https://www.eetimes.com/document.asp?doc_id=1331887
I dont think GF 7nm will be the best process, but i dont think it will be very far behind, certainly a lot closer then 14nm LLP compared to in 14nm and TMSC 16nm........

Intel 10nm and GF 7nm at IEDM 2017
Discussion in 'CPUs and Overclocking' started by raghu78, Oct 19, 2017.

This thread is filled with updated information.

And the information provided by Scott Jones article makes it look like Intel's 10nm and GlobalFoundries 7nm are very close in terms of density.

IEDM 2017 - Intel Versus GLOBALFOUNDRIES at the Leading Edge
by
Scotten Jones
Published on 12-17-2017 08:00 AM

If this information is accurate(coming from IEDM this year), it seems fairly compelling.

 

[raghu78]

Honestly, I don't see what's wrong with that statement if the information being provided by IEDM is accurate.

Actually i was laughing at an earlier made statement by a forum poster which i quoted.

If they pull this off that changes my opinion of them, and gives me more confidence in their reliability moving forward. They have everything riding on having good execution at 7nm. It's a make it or break it situation in my mind. They simply can't afford to fail at 7nm. They still have a number of years to go on the 10 year agreement with IBM, but AMD is free in 2020.

Oh I definitely think GF will execute and deliver a competitive leading edge foundry node. btw the WSA runs till 2024. The current amendment runs through till 2020.

https://www.anandtech.com/show/10631/amd-amends-globalfoundries-wafer-supply-agreement-through-2020

"The overall WSA agreement is set to run through 2024. However throughout the lifetime of the agreement, AMD and GlobalFoundries have amended it several times to account for changing market conditions, AMD’s manufacturing needs, and GlobalFoundries own manufacturing plans. Traditionally this has been a near-yearly event – the WSA has been amended 5 times before today – however for this 6th amendment, AMD and GlobalFoundries have inked a much longer 5-year amendment that will take them through 2020"

I'm sure I've seen comments sayings that on 7nm SOC they exceeded there performance targets, god knows where i saw the quote but.....

So we will just have to wait and see.

actually quick google found this :

https://www.eetimes.com/document.asp?doc_id=1331887
I dont think GF 7nm will be the best process, but i dont think it will be very far behind, certainly a lot closer then 14nm LLP compared to in 14nm and TMSC 16nm........

https://www.globalfoundries.com/new...ver-leading-performance-7nm-finfet-technology

In September 2016, GF announced plans to develop its own 7nm FinFET technology leveraging the company’s unmatched heritage of manufacturing high-performance chips. Thanks to additional improvements at both the transistor and process levels, the 7LP technology is exceeding initial performance targets and expected to deliver greater than 40 percent more processing power and twice the area scaling than the previous 14nm FinFET technology. The technology is now ready for customer designs at the company’s leading-edge Fab 8 facility in Saratoga County, N.Y.

 

[Herr Kutz]

...
https://www.globalfoundries.com/new...ver-leading-performance-7nm-finfet-technology

In September 2016, GF announced plans to develop its own 7nm FinFET technology leveraging the company’s unmatched heritage of manufacturing high-performance chips. Thanks to additional improvements at both the transistor and process levels, the 7LP technology is exceeding initial performance targets and expected to deliver greater than 40 percent more processing power and twice the area scaling than the previous 14nm FinFET technology. The technology is now ready for customer designs at the company’s leading-edge Fab 8 facility in Saratoga County, N.Y.

What exactly does this mean? Are they saying their heritage is unmatched, or their manufacturing of high performance chips is unmatched?

 

[goldstone77]

What exactly does this mean? Are they saying their heritage is unmatched, or their manufacturing of high performance chips is unmatched?

Now, that's laughable. But from the looks of it their design looks similar the Intel's.

2.1 Fin shape
when Intel first introduced FinFETs at 22nm, the bottom of the fin was significantly wider than the top. At the time there was a lot written about the impact of this shape on performance. An ideal fin is rectangular with some rounding of the upper corners to prevent hot spots. If the width of the fin varies from top to bottom the different widths will result in different electrical behavior. I have heard that if you measured Intel's early 22nm fins they actually looked like two transistors. Figure 1 compares Intel's 10nm fin on the left to GF's 7nm fin on the right. Fins are much more rectangular today than the were back in 2011 when Intel introduced their 22nm process. The GF process appears to have more rectangular fins than the Intel process.

Figure 1. Intel fin shape (left) and GF fin shape (right).

 

[raghu78]

What exactly does this mean? Are they saying their heritage is unmatched, or their manufacturing of high performance chips is unmatched?

Yeah that statement is hilarious given that GF could not even design their own 14nm node and had to license. IBM which was acquired by GF delivered high performance processes but those were not effective and could never yield as well as Intel's past nodes like 45,32,22nm.

Now, that's laughable. But from the looks of it their design looks similar the Intel's.

Figure 1. Intel fin shape (left) and GF fin shape (right).

GF 7nm fin shape is more rectangular than Intel 10nm fin shape. Bodes well for performance.

 

[Herr Kutz]

Yeah, that's why I'm thinking they threw heritage in there because it makes the statement meaningless.

 

[jpiniero]

http://www.digitimes.com/news/a20180104PD206.html

New confirmation that Apple is using TSMC 7FF for the 2018 iPhones. Also says they are going to start building the fab for 5 nm in 2018 with trial production in 1H 19.