Intel 20A Node Cancelled for Foundry Customers, "Arrow Lake" Mainly Manufactured Externally

[AleksandarK]

Intel has announced the cancellation of its 20A node for Foundry customers, as well as shifting majority of Arrow Lake production to external foundries. The tech giant will instead focus its resources on the more advanced 18A node while relying on external partners for Arrow Lake production, likely tapping TSMC or Samsung for their 2 nm nodes. The decision follows Intel's successful release of the 18A Process Design Kit (PDK) 1.0 in July, which garnered positive feedback from the ecosystem, according to the company. Intel reports that the 18A node is already operational, booting operating systems and yielding well, keeping the company on track for a 2025 launch. This early success has enabled Intel to reallocate engineering resources from 20A to 18A sooner than anticipated. As a result, the "Arrow Lake processor family will be built primarily using external partners and packaged by Intel Foundry".

The 20A node, while now cancelled for Arrow Lake, has played a crucial role in Intel's journey towards 18A. It served as a testbed for new techniques, materials, and transistor architectures essential for advancing Moore's Law. The 20A node successfully integrated both RibbonFET gate-all-around transistor architecture and PowerVia backside power delivery for the first time, providing valuable insights that directly informed the development of 18A. Intel's decision to focus on 18A is also driven by economic factors. With the current 18A defect density already at D0 <0.40, the company sees an opportunity to optimize its engineering investments by transitioning now. However, challenges remain, as evidenced by recent reports of Broadcom's disappointment in the 18A node. Despite these hurdles, Intel remains optimistic about the future of its foundry services and the potential of its advanced manufacturing processes. The coming months will be crucial as the company works to demonstrate the capabilities of its 18A node and secure more partners for its foundry business.



View at TechPowerUp Main Site | Source

 

[Hyderz]

i think that is a wise decision... the 18A nodes will have better power efficiency i hope....

 

[londiste]

Interesting. What the hell is Arrow Lake manufactured on then? TSMC N3? N4?
Depends on what they mean "primarily" of course. Until now they have said cores chiplet is Intel and 20A.

 

[agent_x007]

i think that is a wise decision... the 18A nodes will have better power efficiency i hope....

"Better power efficiency" doesn't mean newer node CPUs won't be 250W+ TDP reactors. It's up to Intel to make a spec for CPUs after all.
Newer node will be better, but if things go as they are now - you will just see a bit higher performance with the same "maxed out" voltages and power usage.
Why ?
Because it sells (and most don't care about power usage or to run CPUs in their perf/power sweet spot).

 

[kondamin]

Wel it was rumoured to happenand I didn’t believe it.
serious blow In my confidence they would be ok, they are worse than Samsung at the moment

 

[Minus Infinity]

Interesting. What the hell is Arrow Lake manufactured on then? TSMC N3? N4?
Depends on what they mean "primarily" of course. Until now they have said cores chiplet is Intel and 20A.

N3B IIRC.

 

[john_]

It wasn't making much sense to be developing two almost identical nodes at the same time, expect if Intel wasn't sure that 18A will succeed and had 20A as a back up plan. Probably 18A is doing well enough for them to not need to keep throwing money at 20A node.
That's the good scenario.

The bad scenario takes in consideration the fact, Broadcom not being enthusiastic about Intel's manufacturing in it's current phase and also seeing Intel "relying on external partners for Arrow Lake production" instead of moving it to it's own 18A manufacturing node. That probably means that Intel, not just Broadcom, is still unsure about 18A quality or scheduling. Until Intel comes out with a mass producing product at 18A, their manufacturing will continue being a huge question mark, years behind TSMC and even Samsung.

 

[FiRe]

this smells like "20A has oxidation issues too, we're outsourcing this and then working on/fixing 18A"

 

[londiste]

this smells like "20A has oxidation issues too, we're outsourcing this and then working on/fixing 18A"

Why oxidation issues? This would not make sense. Some issues sure, probably resulting in yields issues. It takes time to resolve issues and 18A being incoming - and apparently on track - it is more beneficial to focus on getting that out sooner.

 

[usiname]

At this point they can't make 5 nodes in 4 years even with just adding "+" every year, because they don't have one working node to begin with

It wasn't making much sense to be developing two almost identical nodes at the same time, expect if Intel wasn't sure that 18A will succeed and had 20A as a back up plan. Probably 18A is doing well enough for them to not need to keep throwing money at 20A node.
That's the good scenario.

They replace the "+" with new name and now they have one "new" node every year, well, this was the initial plan...

 

[Vayra86]

It wasn't making much sense to be developing two almost identical nodes at the same time, expect if Intel wasn't sure that 18A will succeed and had 20A as a back up plan. Probably 18A is doing well enough for them to not need to keep throwing money at 20A node.
That's the good scenario.

The bad scenario takes in consideration the fact, Broadcom not being enthusiastic about Intel's manufacturing in it's current phase and also seeing Intel "relying on external partners for Arrow Lake production" instead of moving it to it's own 18A manufacturing node. That probably means that Intel, not just Broadcom, is still unsure about 18A quality or scheduling. Until Intel comes out with a mass producing product at 18A, their manufacturing will continue being a huge question mark, years behind TSMC and even Samsung.

Yeah.
I'm sure there are some differences in perspective from different partners in the food chain here.

Interesting Then again the writing was on the wall, when Intel roadmaps something, its really just an educated guess that everyone knows they won't deliver on.

 

[Crackong]

The winner is.......

 

[Dr. Dro]

this smells like "20A has oxidation issues too, we're outsourcing this and then working on/fixing 18A"

No. Fab contamination slip ups occur every now and then, and Intel is not alone here. TSMC had to trash entire production batches due to contaminants in the past - VERY expensive ordeals. Realistically speaking the amount of production unit samples affected by the via oxidation issue should be relatively insignificant and not worthy of consideration, if you purchase a 12th gen Alder Lake chip or any Raptor Lake chip marketed as "14th gen" you're immune, and 99%+ of the original "13th gen" branded Raptor Lake runs should be OK with this as well. At this point in time it is extremely unlikely you will come across such a processor unless it's been warehoused for years and honestly, by just getting a 14900K instead of a 13900K it's a sure fire way to avoid the problem altogether, even if the risk is absolutely minimal.

The fact of the matter remains that no processor that was potentially affected should have ever shipped to a customer, even, if not particularly considering that Intel was aware that they were functional and could "potentially" become a problem down the line. In my opinion, the company should've relabeled all known potentially affected samples as "Intel Confidential" qualification samples with the ES bit set and handed that batch out to just about anyone in the tech field in an unusually generous grant program, really, just give them out like candy instead, at least they could make the disaster into a ton of good will with the community at large, waive any liability and avoid a lot of heat and negative press. Even if most ended up on Aliexpress down the road, it wouldn't have been as bad of a headache to deal with, since it'd all be "under the rug" and "on the down low", plus if anyone said anything they could wash their hands claiming they were ES chips not intended to be marketed or sold to end users and "any issues never affected production units, <insert boilerplate standard statement about ES chips>" yada yada.

You have to purchase a late-2022 early production run chip manufactured at a very specific plant where the problem occurred at the time to be affected by it, early 2023 chips should already be immune. My i9-13900KS is mid-2023 production unit and it's been stable throughout this entire ordeal, although I have run it with a minor undervolt from day one.

 

[Smartcom5]

this smells like "20A has oxidation issues too, we're outsourcing this and then working on/fixing 18A"

More like being code for “We pretend that the market buys our lame excuses of scrapping Intel 20Å over never-ending yield-issues in favour of Intel 18Å and Arrow Lake now coming fully from TSMC (like we internally had already planned since the beginning, despite deliberately claming publicly it's a internal Intel 20Å-product), and with that hope that it may buy us some time, before we have to acknowledge the inevitable split anyway”.

They never planned Arrow Lake for being on their own internal Intel 20Å (at least not internally and in secret), but it was already designed to be exclusively TSMC-sourced.

No. Fab contamination slip ups occur every now and then, and Intel is not alone here. TSMC had to trash entire production batches due to contaminants in the past - VERY expensive ordeals.

Except that the very difference in regards to Intel here is, on others like Samsung and TSMC, it's rather imminently communicated quickly, publicly and fully transparently, instead of being sit out and kept shut about for two years straight …

Realistically speaking the amount of production unit samples affected by the via oxidation issue should be relatively insignificant and not worthy of consideration, if you purchase a 12th gen Alder Lake chip or any Raptor Lake chip marketed as "14th gen" you're immune, and 99%+ of the original "13th gen" branded Raptor Lake runs should be OK with this as well. At this point in time it is extremely unlikely you will come across such a processor unless it's been warehoused for years and honestly, by just getting a 14900K instead of a 13900K it's a sure fire way to avoid the problem altogether, even if the risk is absolutely minimal.

Right … And the via-oxidation-issues are proven to totally not help increasing the impact of the other voltage-related disasters and speed up electro-migration.

The fact of the matter remains that no processor that was potentially affected should have ever shipped to a customer, even, if not particularly considering that Intel was aware that they were functional and could "potentially" become a problem down the line.

At least we can agree on that, chap! Fair enough.

In my opinion, the company should've relabeled all known potentially affected samples as "Intel Confidential" qualification samples with the ES bit set and handed that batch out to just about anyone in the tech field in an unusually generous grant program, really, just give them out like candy instead, at least they could make the disaster into a ton of good will with the community at large, waive any liability and avoid a lot of heat and negative press. Even if most ended up on Aliexpress down the road, it wouldn't have been as bad of a headache to deal with, since it'd all be "under the rug" and "on the down low", plus if anyone said anything they could wash their hands claiming they were ES chips not intended to be marketed or sold to end users and "any issues never affected production units, <insert boilerplate standard statement about ES chips>" yada yada.

That would've required to have someone with a brain over there in the first place, to engage in such clever actions. You do know that we're talking about Intel here after all, right?

Intel … a degrading shop already way over the hill (after being the king of it in previous times for way too long), which apparently loves to constantly shoot themself in the foot and immediately after, trigger-happily looks down the barrel on purpose again, for finding out how the previous blow-up and fall-out afterwards could happen.

 

[Dr. Dro]

More like being code for “We pretend that the market buys our lame excuses of scrapping Intel 20Å over never-ending yield-issues in favour of Intel 18Å and Arrow Lake now coming fully from TSMC (like we internally had already planned since the beginning, despite deliberately claming publicly it's a internal Intel 20Å-product), and with that hope that it may buy us some time, before we have to acknowledge the inevitable split anyway”.

They never planned Arrow Lake for being on their own internal Intel 20Å (at least not internally and in secret), but it was already designed to be exclusively TSMC-sourced.


Except that the very difference in regards to Intel here is, on others like Samsung and TSMC, it's rather imminently communicated quickly, publicly and fully transparently, instead of being sit out and kept shut about for two years straight …


Right … And the via-oxidation-issues are proven to totally not help increasing the impact of the other voltage-related disasters and speed up electro-migration.


At least we can agree on that, chap! Fair enough.

That would've required to have someone with a brain over there in the first place, to engage in such clever actions. You do know that we're talking about Intel here after all, right?

Intel … a degrading shop already way over the hill (after being the king of it in previous times for way too long), which apparently loves to constantly shoot themself in the foot and immediately after, trigger-happily looks down the barrel on purpose again, for finding out how the previous blow-up and fall-out afterwards could happen.

We're pretty much on the same page

 

[TumbleGeorge]

Lithographic nodes, have reached a size beyond which reduction will primarily lead to an increase in manufacturing defects. The right way is to stop the race to scale down and work on architectural improvements to the compute lines. Gateway width, more gateways, more and smart gates, more efficient predictors. Bigger and smart caches.

 

[R0H1T]

The right way is to stop the race to scale down and work on architectural improvements to the compute lines. Gateway width, more gateways, more and smart gates, more efficient predictors. Bigger and smart caches.

That's not going to happen the same way you'll always have better "phones" then TV/monitors/refrigerators/AC et al. In at least half of those categories you don't really "need" to upgrade!

 

[Eternit]

It is the same as with 10nm. 20A and 18A is the same node just with improvements. So Arrow Lake 20A compute tile will be like Cannonlake. They will make enough of them to tell the shareholders they haven't lied about 5 nodes in 4 years. I'm wondering when Pat will tell he had an affair and needs to resign.

 

[R0H1T]

Not sure why Pat needs to be sacrificed at the altar to burn Intel, they're doing a fine job since at least Otellini. The rot IMO is close to 2 decades old now!

 

[Eternit]

Not sure why Pat needs to be sacrificed at the altar to burn Intel, they're doing a fine job since at least Otellini. The rot IMO is close to 2 decades old now!

It is not his fault, but shareholders will demand someones head and CEO is natural choice. He hasn't caused these problems, but he was hired to fix Intel and he hasn't.

 

[N/A]

18A is also canceled for 14A that simplifies the masks to lower the cost and as soon as that boots in the lab skip to 10A.

 

[phanbuey]

It is not his fault, but shareholders will demand someones head and CEO is natural choice. He hasn't caused these problems, but he was hired to fix Intel and he hasn't.

The worst decision right now would be to get rid of Pat IMO. A change in direction now could actually be fatal.

 

[Eternit]

The worst decision right now would be to get rid of Pat IMO. A change in direction now could actually be fatal.

I think they are already in critical state. They have fabs set to 4, 3, 20A and 18A manufacturing very little while they spent a lot of money on them. If 18A will be another flop their manufacturing branch will be useless and I doubt shareholders will accept investments in another nodes. 5 years ago they said their 10nm was poor while 7nm was great so they wanted to speed up switching from 10nm to 7nm, but a year later 7nm was delayed again and 10nm was renamed to Intel 7. I doubt spinning out their fabs is viable. When AMD did that, their fabs were in a better shape than Intel's are now. Also government gave them money for their factories upgrades and they wasted them. Will the government approve selling factories to TSMC or Samsung? I doubt it. Can they find a buyer in the US? So they must keep on financing unprofitable branch of their business.

 

[usiname]

I think they are already in critical state. They have fabs set to 4, 3, 20A and 18A manufacturing very little while they spent a lot of money on them. If 18A will be another flop their manufacturing branch will be useless and I doubt shareholders will accept investments in another nodes. 5 years ago they said their 10nm was poor while 7nm was great so they wanted to speed up switching from 10nm to 7nm, but a year later 7nm was delayed again and 10nm was renamed to Intel 7. I doubt spinning out their fabs is viable. When AMD did that, their fabs were in a better shape than Intel's are now. Also government gave them money for their factories upgrades and they wasted them. Will the government approve selling factories to TSMC or Samsung? I doubt it. Can they find a buyer in the US? So they must keep on financing unprofitable branch of their business.

The think is that their 7nm (Intel 4) is garbage, it has 120 MTr/mm2, for comparison, their 10nm (intel 7) is 100 MTr/mm2. No wonder it's marginal better than their polished 10nm. Now lets see the TSMC 3nm, that will be used for Arrow and Lunar Lake - 200-220 MTr/mm2 and their mythical 18a won't be more than 160 MTr/mm2, 180 at most and it will be ready for 2026. They will release Arrow Lake in the end of 2024, refresh it with the same TSMC node in 2025 (Arrow Lake refresh) and when they are ready to release whatever is real upgrade over Arrow Lake, it will be in 2026 and how you see them to downgrade from 220 MTr/mm2 to their miserable 160 MTr/mm2? They won't, they will go again for the next TSMC node. Their fabs are so behind, that its no longer cost effective to build their own CPUs here, because this will hurt the performance and the competitiveness

 

[rv8000]

They just stack all the cash the US has been pumping into them and burn it Joker style? I don’t think I’ve heard a single positive thing about Intel’s foundries for quite some time now.