News Intel announces cancellation of 20A process node for Arrow Lake, goes with external nodes instead, likely TSMC [Updated]

[Admin]

Intel announced today that it no longer plans to use its own 20A process node with its upcoming Arrow Lake processors. Instead, it will use external nodes for all of the compute tiles, and then package the chips together.

Intel announces cancellation of 20A process node for Arrow Lake, goes with external nodes instead, likely TSMC [Updated] : Read more

 

[ottonis]

So, why did Intel even bother developing and spending so much money on the 20A process node, when they have already come so far with the more advanced 18A process? Just a couple of days ago it was reported that the 18A process node is doing remarkably well and has low defect rates, considering its development stage.

 

[Murissokah]

Disappointing to see Intel abandon a process amid so many bad news, but it's probably for the best. Using a proven TSMC process will likely yield a better product in short term and allow them to focus on 18A. Considering Gelsinger himself said he "bet the whole company on 18A", they really should focus on that.

 

[bolweval]

So, why did Intel even bother developing and spending so much money on the 20A process node, when they have already come so far with the more advanced 18A process? Just a couple of days ago it was reported that the 18A process node is doing remarkably well and has low defect rates, considering its development stage.

From the article.

"However, Intel's 20A served as a vehicle for several new advances, like  RibbonFet Gate-All-Around (GAA) technology, which is Intel’s first new transistor design since FinFET arrived in 2011. It also marked the debut of the company's  PowerVia backside power delivery tech, which routes power for the transistors through the backside of the processor die.

Intel says the learnings it gained from its 20A node have contributed to the success of its 18A node, which makes sense given that 18A is a tighter refinement of the technologies invented for 20A."

 

[ottonis]

Thank you. While Intel's argument makes sense to a certain degree, developing an entire process node the sole purpose of which is to serve as a platform for learning and developing is probably not the most economical way of doing things. So, I sense that Intel management is using this - at least in part - as an excuse for the platform failure.

 

[NinoPino]

From the article.

"However, Intel's 20A served as a vehicle for several new advances, like  RibbonFet Gate-All-Around (GAA) technology, which is Intel’s first new transistor design since FinFET arrived in 2011. It also marked the debut of the company's  PowerVia backside power delivery tech, which routes power for the transistors through the backside of the processor die.

Intel says the learnings it gained from its 20A node have contributed to the success of its 18A node, which makes sense given that 18A is a tighter refinement of the technologies invented for 20A."

Nobody spend billions on a process node only to use the knowhow on another process node. If they spent so much money on 20A is because they wanted to use it to produce a big amount of chips. Also the article's consideration that they abandon the 20A because there are no products to justify the expense is a bit wobby. If this is the real motivation to give up on 20A, than this imply a very bad management that did a wrong business plan.

 

[Pierce2623]

From the article.

"However, Intel's 20A served as a vehicle for several new advances, like  RibbonFet Gate-All-Around (GAA) technology, which is Intel’s first new transistor design since FinFET arrived in 2011. It also marked the debut of the company's  PowerVia backside power delivery tech, which routes power for the transistors through the backside of the processor die.

Intel says the learnings it gained from its 20A node have contributed to the success of its 18A node, which makes sense given that 18A is a tighter refinement of the technologies invented for 20A."

So they couldn’t design those without a totally wasted node? 18a and 20a are nearly identical and 18a is just slightly denser. This is horrible news.

 

[thisisaname]

To me it smells of 20a is broken in some way and 18a is not due until 2025 (so they have some wriggle room)and the last thing Intel's share price needs right now is something else to go wrong, so they are pushing the next mode into the long grass and hoping it is fixed by then. 20a is dead long live 18a, please just think of 18a (and do not remember 10nm)
From nearly $50 to just under $29 in less than 9 months and now there is talk of it being removed from the Dow Jones, Intel stock is deeply in trouble.

​

 

[Taslios]

Has intel had a genuinely successful node since 14+++++++++++++++++? I mean all the +++++s only existed because 10NM.. now rebranded as intel7 was so vastly delayed and cost prohibitive.

Intel 4 is supposedly good, but little to nothing is being produced on it.

At this Rate intel fabs are not going to win many customers and are rapidly becoming an anchor to the rest of the company.. albeit an anchor that I suspect the US gov will be reluctant to let fail.

 

[gg83]

I think the only way Intel will get the lead is by going all-in on future nodes. Using TSMC makes sense to me.

 

[Marlin1975]

Has intel has a genuinely successful node since 14+++++++++++++++++? I mean all the +++++s only existed because 10NM.. now rebranded as intel7 was so vastly delayed and cost prohibitive.

Intel 4 is supposedly good, but little to nothing is being produced on it.

At this Rate intel fabs are not going to win many customers and are rapidly becoming an anchor to the rest of the company.. albeit an anchor that I suspect the US gov will be reluctant to let fail.

Yea this news and Broadcom saying they are disappointed in Intels 18a node is not a good sign for intel.

Seems like everyday its just another year... just another year... and they will have it all worked out.

 

[why_wolf]

Well hopefully whatever was critically bad with 20a was fixed in 18a. Because otherwise we are looking at a TSMC monopoly and bottleneck on advanced chip production. So get ready for more price increases.

 

[epobirs]

So, why did Intel even bother developing and spending so much money on the 20A process node, when they have already come so far with the more advanced 18A process? Just a couple of days ago it was reported that the 18A process node is doing remarkably well and has low defect rates, considering its development stage.

Because there is no way to know how things are going to progress until they are already underway. Intel has been doing this for decades, working on the next two or three nodes while the next one is nearing mass production of products. The investment between a node being developed and building a site to perform mass production of that node is mind-boggling. In the $BILLIONS nowadays. In the past, they always knew they had a better node well into development but that the one set for introduction would have a year or more as the market leader. That used to be the norm for Intel. It wasn't sustainable forever. Eventually costs for every step of the R&D process became such as to make every milestone a major decision point. At the same time, after Intel's long stumble at 10nm that ended their longstanding process node lead, revenues were not as reliable as they'd been. Spending the massive amount needed to bring 20A to mass production for it to have a very short life span as their lead foundry offering is a decision that had to be weighed against skipping in favor of the refined version.

Imagine if the 6NP node at TSMC was expected to be available inside less than a year of their NP process node? They would have been in the same position then as Intel is now. 6NP is primarily a refinement of 7NP, and mainly attractive to customers with existing 7NP products, like the Sony PS5. There are some major technical and financial reasons Sony won't put the PS5 one of the major newer nodes, like 5NP. 6NP gave them just enough improved value at an acceptable cost to do the PS5 slim and the all but announced PS5 Pro. In real life, 6NP became available years after 7NP had a highly successful run as the market leader and even after 5NP and denser nodes had entered production. It was useful for a certain subset of customers but not all.

 

[technologyiskiller]

Nobody spend billions on a process node only to use the knowhow on another process node. If they spent so much money on 20A is because they wanted to use it to produce a big amount of chips. Also the article's consideration that they abandon the 20A because there are no products to justify the expense is a bit wobby. If this is the real motivation to give up on 20A, than this imply a very bad management that did a wrong business plan.

To me it smells of 20a is broken in some way and 18a is not due until 2025 (so they have some wriggle room)and the last thing Intel's share price needs right now is something else to go wrong, so they are pushing the next mode into the long grass and hoping it is fixed by then. 20a is dead long live 18a, please just think of 18a (and do not remember 10nm)
From nearly $50 to just under $29 in less than 9 months and now there is talk of it being removed from the Dow Jones, Intel stock is deeply in trouble.

​

Do not try to to find negativity everywhere. From the beginning Intel said that 20A was never for mass production. They had multiple new technologies which were coming in like BSPDN and GAA so putting all of that in just one node is very risky. So 20A was used mostly as a stepping stone for 18A.

There is no point in commenting if you have no idea how developing a new process node works.

 

[epobirs]

So they couldn’t design those without a totally wasted node? 18a and 20a are nearly identical and 18a is just slightly denser. This is horrible news.

A bit of a contradiction there. If 20A enabled 18A, which is more a refinement of 20A, then the sunk costs are kept within reason compared to the need to build separate facilities if one commits to both 20A and 18A, with 20A having only a brief period in which it is the lead offering. That final step, building the facility for mass production, is EXPENSIVE. Putting off that cost to go directly to 18A makes sense from the financial perspective. They might have done differently in a previous era when the money was less tight but even then the analysis could still have favored skipping the short-lived node and the cost it would require.

This isn't unprecedented in product development and manufacturing. The point between prototype and committing to mass production has always been where a company pauses to assess before deciding how to proceed.

 

[technologyiskiller]

A bit of a contradiction there. If 20A enabled 18A, which is more a refinement of 20A, then the sunk costs are kept within reason compared to the need to build separate facilities if one commits to both 20A and 18A, with 20A having only a brief period in which it is the lead offering. That final step, building the facility for mass production, is EXPENSIVE. Putting off that cost to go directly to 18A makes sense from the financial perspective. They might have done differently in a previous era when the money was less tight but even then the analysis could still have favored skipping the short-lived node and the cost it would require.

This isn't unprecedented in product development and manufacturing. The point between prototype and committing to mass production has always been where a company pauses to assess before deciding how to proceed.

Yes absolutely. Yet most people have no idea how semiconductor manufacturing works and think it's just like software development.

 

[technologyiskiller]

One more thing, do understand that hardware design starts with the correct node in mind which means that ARL was always going to be mostly on external process. They couldn't just design for 20A and suddenly cancel and move on to an external process one fine day. So there is not a major surprise here but more like stating the obvious.

 

[epobirs]

One more thing, do understand that hardware design starts with the correct node in mind which means that ARL was always going to be mostly on external process. They couldn't just design for 20A and suddenly cancel and move on to an external process one fine day. So there is not a major surprise here but more like stating the obvious.

It appears now that the subset of ARL product originally intended to be produced on 20A were proof of life to potential foundry customers. A shipping product is a big improvement for credibility but unneeded if the node itself is not to be a product for sale.

 

[NinoPino]

Do not try to to find negativity everywhere.

It is highly negative for Intel, whatever you look at it. It is not necessary to "find" anything. It is a fact.

From the beginning Intel said that 20A was never for mass production.

"Never for mass production" is a nonsense for a highly advertised and long awaited process node. It is Intel that put 20A in his road map. What is the sense of "not for mass production" ? From a business pov it is wasted money at best. If they want to experiment with new technologies they do it silently and invest the minimum so if something goes wrong there will be a minimal damage.

They had multiple new technologies which were coming in like BSPDN and GAA so putting all of that in just one node is very risky. So 20A was used mostly as a stepping stone for 18A.

There is no point in commenting if you have no idea how developing a new process node works.

I can have no idea of silicon development but business, money and industrial works always in the same way.

 

[DS426]

So, why did Intel even bother developing and spending so much money on the 20A process node, when they have already come so far with the more advanced 18A process? Just a couple of days ago it was reported that the 18A process node is doing remarkably well and has low defect rates, considering its development stage.

A lack of executive and management discipline. Advanced nodes are already prohibitively expensive to develop and spin up fab capacity on, so it seems like they were 1) waving the carrot in front of investors and 2) wanting to have the bragging rights of being on par with TSMC and therefore creating an induction mechanism for pulling in new customer orders.

They *could* have just introduced PowerVia and GAA with 18A, but they went the faster and more expensive route. Judging by recent events, they miss-stepped, even if you can call 20A a "springboard."

 

[rtoaht]

This bodes well for 18A. The two nodes are very close. If 18A is doing well and launching shortly after 20A, then there is no reason to spend the money to ramp the 20A for a small volume of products.

 

[emv]

This bodes well for 18A. The two nodes are very close. If 18A is doing well and launching shortly after 20A, then there is no reason to spend the money to ramp the 20A for a small volume of products.

Intel is not moving Arrow lake to 18A. 18A wont ramp until late 2025. Intel chose TSMC over 20A. 20A and 18A capacity are the same tool set.

Wait till you see what arrow lake refresh looks like in 2025. Hint its not 18A either LOL

the problem is that 20A isnt s good as N3..... Its not as good as Intel 3 either, apparently

 

[phead128]

Intel outsourcing to TSMC is not a good look. Normally you would use 20A to showcase to potential customers that your advanced nodes are yielding well, sufficient volume, and cost-efficient. Instead, customers see that even Intel is not confident in it's own internal nodes, preferring TSMC instead. So why would any customer abandon TSMC unless the US gov't forces it at gunpoint?

 

[technologyiskiller]

It appears now that the subset of ARL product originally intended to be produced on 20A were proof of life to potential foundry customers. A shipping product is a big improvement for credibility but unneeded if the node itself is not to be a product for sale.

It isn't. It was a stop gap till 18A comes up. But it comes with a high cost.

 

[technologyiskiller]

It is highly negative for Intel, whatever you look at it. It is not necessary to "find" anything. It is a fact.


"Never for mass production" is a nonsense for a highly advertised and long awaited process node. It is Intel that put 20A in his road map. What is the sense of "not for mass production" ? From a business pov it is wasted money at best. If they want to experiment with new technologies they do it silently and invest the minimum so if something goes wrong there will be a minimal damage.


I can have no idea of silicon development but business, money and industrial works always in the same way.

No it doesn't. Its just the way the press these days has made it.

Creating a new process node which is sufficiently advanced is incredibly complex and something that requires very deep understanding of theoretical physics and chemistry (it's actually more complex than rocket science). That's the reason why there is only a handful companies in the world who does semiconductor manufacturing as opposed to design.

Most of the tech writers are incompetent and have absolutely zero idea about this (naturally because most of them have no idea about physics). So they just want to create a populist sensation by creating negative reports.