Friday, October 24th 2025

Intel Reassures 18A Yields Are Okay, Continues 14A Development

Intel has just reported its Q3 earnings, delivering stronger than expected results and major revival of operational income. During the conference call after the earnings were published, Intel reassured its investors that the company is working with good, predictable 18A yields. In addition to 18A, the company confirmed that 14A node development is shaping up to be a leading-edge node co-designed with customers. CFO David Zinsner confirmed "We are making steady progress on Intel 18A. We are on track to bring Panther Lake to market this year. Intel 18A yields are progressing at a predictable rate, and Fab 52 in Arizona, which is dedicated to high-volume manufacturing, is now fully operational. In addition, we are advancing our work on Intel 18A, and we continue to hit our PDK milestones. Our Intel 18A family is the foundation for at least the next three generations of client and server products."

Adding to that, the CFO commented on 18A yields: "I wouldn't say Intel 18A yields are in a bad place. They're where we want them to be at this point. We had a goal for the end of the year, and they're going to hit that goal. To be fully accretive in terms of the cost structure of Intel 18A, we need the yields to be better. That's like every process. That's what happens. It's going to take all of next year, I think, to really get to a place where that's the case." For Intel, 18A yields are now at a very low defect rate where manufacturing even some of the bigger dies is not a problem or financial burden. However, as every node, it matures over time where defect rates are constantly reduced to increase operating margin and reduce waste dies. Hence, Intel is still investing into the 18A refinement, and the node will stick for a very long time.
On the topic of 14A node, CEO Lip-Bu Tan confirmed: "On the 14A, clearly we're engaging with multiple customers in terms of milestone basis. We also really driving some of the yield and performance reliability that I'm seeing improvement." CFO David Zinsner added that "I would tell you on Intel 14A, we're off to a great start. If you look at Intel 14A in terms of its maturity relative to Intel 18A at that same point of maturity, we're better in terms of performance and yield. We're off to an even better start on Intel 14A." At this point in development, 14A is reaching better milestones than 18A. Thanks to the use of High-NA EUV tools from ASML, Intel has reported processing over 30,000 14A wafers in a single quarter using High-NA EUV exposure. The company achieving simplified manufacturing by reducing the required steps for a particular layer from 40 to fewer than 10, which has led to faster cycle times.
Source: Q3 Earnings Call Transcript
Add your own comment

40 Comments on Intel Reassures 18A Yields Are Okay, Continues 14A Development

#1
Space Lynx
Astronaut
x3d cache only for me, so unless you can match that Intel, i just don't give a fk ye

when intel cpu can beat amd in factorio, i will consider intel. that's my test representative
Posted on Reply
#2
dgianstefani
TPU Proofreader
Space Lynxx3d cache only for me, so unless you can match that Intel, i just don't give a fk ye

when intel cpu can beat amd in factorio, i will consider intel. that's my test representative
Intel CPUs do win in factorio, you simply have to be late game, when it doesn't fit within the cache. It's funny that you picked that specific scenario where the cache does eventually get overwhelmed.
Posted on Reply
#3
TheGeekn°72
dgianstefaniIntel CPUs do win in factorio, you simply have to be late game, when it doesn't fit within the cache. It's funny that you picked that specific scenario where the cache does eventually get overwhelmed.
I'm guessing that's where the faster RAM comes in handy ? or is intel being lifted up by something else in that use case ?
Posted on Reply
#4
dgianstefani
TPU Proofreader
TheGeekn°72I'm guessing that's where the faster RAM comes in handy ? or is intel being lifted up by something else in that use case ?
Not really sure why tbh, perhaps faster RAM, stronger core etc. The 3D-VCache is great when it works but it's not 100% for all games all of the time, at which point the chips regress a lot.
Posted on Reply
#5
JimmyDoogs
AMD and Intel have kind of swapped sides this gen. AMD is the better performing CPU but Intel is the better price to performance. I like that back and forth between the two concepts. IMO there's really no reason to shame either chip maker. They're both have their ups. I work on the Twinscan EXE 5200 and I can't wait to see what the 18A chips can do.
Posted on Reply
#6
john_
Intel's architecture is different than AMD's and might not get the advantages from X3D cache that AMD is enjoying. AMD's CPUs where seeing significant performance increases with more cache, even from the Athlon XP era (Barton core). On the other hand performance in games with AMD CPUs with less than the typical cache is hugely negatively affected. My R5 5500 looks like an R5 3600 in games because the 5500 while being Zen 3 comes with half the cache of the Zen 2 3600. 16MBs of typical cache equals the jump from Zen 2 to Zen 3, which is significant.

Now I don't know about old Intel CPUs, but the following video might be offering an explanation of why Intel haven't done X3D cache already with it's modern CPUs. The hybrid design might be adding more complexity, but Intel isn't a company that can't build hardware schejulers(or whatever) in it's chips to bypass these kind of difficulties.


In many cases we expect a feature to be equally good in every case of CPUs. So, when AMD introduced the 5800X3D everyone was asking X3D on the 16 core models. AMD did that, only for us to see that an 9950X3D is slower in most cases against an 9800X3D in games. It might be faster in 5 years when 8 P cores will not be enough. People then started asking AMD to integrate X3D in both CCDs. Well we have rumors about an 9950X3D2. We will see then if adding cache in both CCDs will do the trick or end up getting zero extra performance from that second cache, or even get a regression of performance if the CPU is trying to find in which X3D cache is the data it is looking for and which cores to use(I am not an engineear that last part might be a stupid guess - or all the post, who knows?).
dgianstefaniThe 3D-VCache is great when it works but it's not 100% for all games all of the time, at which point the chips regress a lot.
On the other hand, if Intel chooses to follow AMD by introducing a type of X3D cache, we can expect more in the beginning and almost all AAA games latter, to support X3D cache, meaning AMD's X3D cache CPUs will be seeing performance increases in almost all AAA games.
Posted on Reply
#7
dgianstefani
TPU Proofreader
john_On the other hand, if Intel chooses to follow AMD by introducing a type of X3D cache, we can expect more in the beginning and almost all AAA games latter, to support X3D cache, meaning AMD's X3D cache CPUs will be seeing performance increases in almost all AAA games.
AMD actually followed Intel's 2014 Broadwell, which introduced the large cache CPU. I'm not sure cross-vendor large cache CPUs would change the fundamental limitations of that design choice. Games that don't benefit from or fit within the ~100 MB cache still won't benefit from or fit in it.
Posted on Reply
#8
john_
dgianstefaniAMD actually followed Intel's 2014 Broadwell, which introduced the large cache CPU. I'm not sure cross-vendor large cache CPUs would change the fundamental limitations of that design choice. Games that don't benefit from or fit within the ~100 MB cache still won't benefit from or fit in it.
I think that back then with Broadwell, the extra cache was mostly to help the iGPU. Intel's "answer" to AMD's APUs. Don't remember how much it was helping the CPU.

In any case both companies copy each other. Intel tried to make a rushed copy of AMD's Athlon X2 by gluing two cores, and we know that it worked, then moved the IMC on the CPU like AMD(AMD had copied it from? don't remember), AMD on the other hand copied Pentium 4 with Bulldozer and almost gone bankrupt, then use it's own glue to build the first Zen, understanding that trying to build the perfect chip in one go could end up with lost time, lost money and a nice feature called "TLB bug"(hello Barcelona). Then Intel copied ARM by building the hybrid CPUs to manage to match or exceed the number of advertised cores. Now who knows, they might integrate more cache, but in my opinion I expect just faster cores, higher IPC, not an X3D cache feature. Intel always tries to get higher IPC.
Posted on Reply
#9
dgianstefani
TPU Proofreader
john_Now who knows, they might integrate more cache, but in my opinion I expect just faster cores, higher IPC, not an X3D cache feature. Intel always tries to get higher IPC.
I suspect that's the best way forwards, a faster core is pretty much always useful, a big L3 cache is sometimes useful, and expensive to incorporate. Look at the aging Raptor Lake, it's still faster than non-X3D Zen 5. So having big cache variants for a premium, for those who want/need them is the way forwards. Mind you with higher and higher core count CPUs still on dual channel, large cache can be good for reducing reliance on calls to memory.
Posted on Reply
#10
watzupken
“Adding to that, the CFO commented on 18A yields: "I wouldn't say Intel 18A yields are in a bad place. They're where we want them to be at this point. We had a goal for the end of the year, and they're going to hit that goal.”

I guess they expected to be questioned on the yields, but the above response doesn’t provide any assurance or confidence in my opinion. The beat around the bush response seems to imply it’s not good. “Where we want them to be” meaning what? 20, 30%? It’s where they expect, but is it good such that they attract more customers and start making money?
Posted on Reply
#11
Tek-Check
So... what are the yields now?
The narrative about yields by Intel folks sounds too generic, without much substance. Worryingly, the article reads that it would take them until the end of 2026 to get yields high enough in order to balance financial books on a positive side.
Posted on Reply
#12
Outback Bronze
dgianstefaniLook at the aging Raptor Lake, it's still faster than non-X3D Zen 5
Sometimes even faster than X3D in certain instances. Found an interesting video here:

Posted on Reply
#13
TechBuyingHavoc
Tek-CheckSo... what are the yields now?
The narrative about yields by Intel folks sounds too generic, without much substance. Worryingly, the article reads that it would take them until the end of 2026 to get yields high enough in order to balance financial books on a positive side.
Talk is cheap and Intel loves to talk a big game and under-deliver for a full decade now.. I will believe Intel when they finally release a kickass product.
Posted on Reply
#14
Tek-Check
Besides, the American public and taxpayers now own 10% of Intel, so the level of scrutiny and demands of transparency will only increase. They will have to start releasing more specific hard data into public domain.
Posted on Reply
#15
john_
dgianstefaniI suspect that's the best way forwards, a faster core is pretty much always useful, a big L3 cache is sometimes useful, and expensive to incorporate. Look at the aging Raptor Lake, it's still faster than non-X3D Zen 5. So having big cache variants for a premium, for those who want/need them is the way forwards. Mind you with higher and higher core count CPUs still on dual channel, large cache can be good for reducing reliance on calls to memory.
Intel was always playing the IPC game. Especially in that era when AMD was half dead and 4 cores with HT where considered more than enough. Even when AMD started winning with Zen, Intel kept insisting on IPC. Better cores, faster cores, more wattage, new instructions, anything to gain IPC from the core itself, not from cache or anything else. They already had and probably still have the best IMC anyway.
I wonder if both AMD and Intel are reverse engineering Apples' M cores, trying to find the good ideas in those ARM cores that can be copied in X86 cores.
Posted on Reply
#16
TechBuyingHavoc
Tek-CheckBesides, the American public and taxpayers now own 10% of Intel, so the level of scrutiny and demands of transparency will only increase. They will have to start releasing more specific hard data into public domain.
The public and taxpayers may own Intel but the government is who manages this investment. Just how shareholders may own a company but the actual administration and decision making is taken by executives and upper management, the same is true here.

If the US administration does not care to release hard data (and why would they if it turns out to be negative on their actions), then they won't have to.
Posted on Reply
#17
Squared
john_I think that back then with Broadwell, the extra cache was mostly to help the iGPU. Intel's "answer" to AMD's APUs. Don't remember how much it was helping the CPU.
Broadwell's "Crystal Well" eDRAM cache was specifically included for the iGPU and it worked great for that purpose, but the CPU could also access it. Someone (I can't remember which site, possibly AnandTech in which case the article is gone now) retested Broadwell with a discrete GPU after V-cache came out and found that it was ahead of its time when playing modern games. That wasn't true in 2015; back then Skylake beat it in nearly all games when using a discrete GPU.
Posted on Reply
#18
W1zzard
Tek-Checkthat it would take them until the end of 2026 to get yields high enough in order to balance financial books on a positive side.
that's how I'm interpreting it, too. But we don't know what the comparison is, buy at TSMC? Make something else with the allocation? Make nothing? sure that would result in a bigger loss, having idle fabs and not learning anything by making production
Posted on Reply
#19
Squared
watzupken“Adding to that, the CFO commented on 18A yields: "I wouldn't say Intel 18A yields are in a bad place. They're where we want them to be at this point. We had a goal for the end of the year, and they're going to hit that goal.”

I guess they expected to be questioned on the yields, but the above response doesn’t provide any assurance or confidence in my opinion. The beat around the bush response seems to imply it’s not good. “Where we want them to be” meaning what? 20, 30%? It’s where they expect, but is it good such that they attract more customers and start making money?
For Intel, 18A yields are now at a very low defect rate where manufacturing even some of the bigger dies is not a problem or financial burden.
My take-away is that yields are holding 18A back and are probably the reason why 20A was cancelled. But Samsung's GAAFET nodes have reportedly had around 20-30% yields and have never produced anything more substantial than the W1000 smartwatch processor. Intel is claiming that 18A is ready for big chips today. 30% yields on a small chip might mean around 6% yields on a big chip. With Panther Lake and Clearwater Forest just around the corner yields have to be better than where Samsung was at this point. Also Diamond Rapids is coming next year with 48 "Cove" cores per die.

With Samsung and Intel both struggling to get GAAFETs fully operational, I'm curious how TSMC will handle the transition.
Posted on Reply
#20
EsaT
john_Intel's architecture is different than AMD's and might not get the advantages from X3D cache that AMD is enjoying. AMD's CPUs where seeing significant performance increases with more cache, even from the Athlon XP era (Barton core). On the other hand performance in games with AMD CPUs with less than the typical cache is hugely negatively affected. My R5 5500 looks like an R5 3600 in games because the 5500 while being Zen 3 comes with half the cache of the Zen 2 3600. 16MBs of typical cache equals the jump from Zen 2 to Zen 3, which is significant.
AMD's current chiplet design adds memory access latency more than marginally over monolithic chips.
And in their core games have lots of serial code making memory access latency important for gaming performance.
That's why amount of cache has more than marginal effect.

And hence why everyone is hoping for this to be true:
www.techpowerup.com/341445/amd-d2d-interconnect-in-zen-6-gets-sea-of-wires-upgrade
Posted on Reply
#21
nguyen
Outback BronzeSometimes even faster than X3D in certain instances. Found an interesting video here:

He just posted a new video saying the AMDips were caused by Afterburner OSD polling GPU power percent, everyone knew about this causing stutters months ago :wtf:
Posted on Reply
#22
Tek-Check
watzupken“Where we want them to be” meaning what? 20, 30%? It’s where they expect, but is it good such that they attract more customers and start making money?
Unknown. Elon went with Samsung's 2nm GAA with ~50% of yields. Samsung themselves are said to be dedicating only smaller amount of 2nm volume for Galaxy S26 Ultra. Majority of phone chips will be from Qualcomm on TSMC. How much? A closely kept secret. Those are the news that I am aware of at the moment.

Microsoft made an order with Intel for undisclosed amount of chips. Anyone knows how much? Qualcomm and Nvidia clearly said they are not ready, as yet, to place any orders on 18A. It's much easier for Nvidia to get Intel's Xeons ready-made CPUs than etch their own state-of-the-art chips on 18A. It's a precarious situation, as TSMC has been pushed hard behind the scenes... to invest over $160 billion into US expansion, so vast majority of the cutting-edge chips will be de facto produced by them. Together, all semiconductor fabs in the US are estimated to reach ~20% of global chip output by 2030, for 'advanced' chips. How much of that would be cutting-edge advanced is anyone's guess. We'd need an industry insider to crunch those numbers for us and define chip categories more precisely.

If the yields on 18A are below 50% at the moment, it will be tough until the end of 2026, as the bigger 16 core Panther Lake compute tile will arrive first. Bigger die means more defective dies. I'd expect a slow launch throughout H1 2026, on more expensive side of things.
W1zzardthat's how I'm interpreting it, too. But we don't know what the comparison is, buy at TSMC? Make something else with the allocation? Make nothing? sure that would result in a bigger loss, having idle fabs and not learning anything by making production
True that, I agree. Now that the company is partially owned by the public, they will need to be more explicit and upfront with the progress of yields. Such information should be more readily available, despite potentially being uncomfortable and any business sensitivities surrounding transparency. Somebody who invests individually or via bigger brokers, or companies investing public money in stock markets, should be asking those questions in order to avoid unplesant surprises and tame any decisions based on vague marketing language. I would not be surprised if yields leak into public domain at some point via tech websites or similar. It's a strange situation because such information should not leak, but simply be available now that the corporate ownership status has changed. Interesting times ahead to recovery, including recent brutal job cuts.
www.tomshardware.com/pc-components/cpus/intel-has-cut-35-500-jobs-in-less-than-two-years-more-than-20-000-let-go-in-in-recent-months-as-lip-bu-tan-continues-drastic-recovery-journey
Posted on Reply
#23
anonuser57
john_I think that back then with Broadwell, the extra cache was mostly to help the iGPU. Intel's "answer" to AMD's APUs. Don't remember how much it was helping the CPU.

In any case both companies copy each other. Intel tried to make a rushed copy of AMD's Athlon X2 by gluing two cores, and we know that it worked, then moved the IMC on the CPU like AMD(AMD had copied it from? don't remember), AMD on the other hand copied Pentium 4 with Bulldozer and almost gone bankrupt, then use it's own glue to build the first Zen, understanding that trying to build the perfect chip in one go could end up with lost time, lost money and a nice feature called "TLB bug"(hello Barcelona). Then Intel copied ARM by building the hybrid CPUs to manage to match or exceed the number of advertised cores. Now who knows, they might integrate more cache, but in my opinion I expect just faster cores, higher IPC, not an X3D cache feature. Intel always tries to get higher IPC.
I think whether Intel starts using stacked cache is going to depend less on strategy and more on whether they can integrate it into their interposer die. I think currently the interposer dies don't have much directly under where the cores of the core die sits. So the area is just free transistors if they can make it useful.
Posted on Reply
#24
Tek-Check
TechBuyingHavocIf the US administration does not care to release hard data (and why would they if it turns out to be negative on their actions), then they won't have to.
True that. It's going to be interesting to see if anyone or any institution pushes for public access to such technical and sensitive information. Public ownership, even partial, also brings new ways of dealing with data. I don't know exactly how this is handled in the US.
Posted on Reply
#25
Geofrancis
my understanding is that Intel and AMD have very different caching strategies,

AMD's L3 cache is just a victim cache, it only holds data ejected from L2, so it's almost like a big L2 extension. Data is only pulled from ram when it's needed, so latency is more important than bandwidth. with X3D the L2 extension is enormous.

Intel's L3 cache is more predictive and tries to prefetch to stream in data before it's needed, This means intel can take advantage of the bandwidth because its always streaming in data in an attempt to have it there before its needed I belive this is why It's less effected by slower memory.


so it will be interesting how much of a difference a big cache would make on intel as it would be used in a totally different way.
Posted on Reply
Add your own comment
Nov 20th, 2026 17:05 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts