News Intel Fixes 7nm, Meteor Lake and Granite Rapids Coming in 2023

[Admin]

Intel announced that it has solved its 7nm production issues and that Meteor Lake and Granite Rapids chips will come to market in 2023.

Intel Fixes 7nm, Meteor Lake and Granite Rapids Coming in 2023 : Read more

 

[hotaru.hino]

...the delayed production schedule will leave the company in direct competition with chips built on more advanced nodes from competing foundries.

If we measure how advanced nodes are by transistor density, Intel's 10nm is actually better than TSMC and Samsung's first generation 7nm (https://en.wikipedia.org/wiki/7_nm_process#7_nm_process_nodes_and_process_offerings)

Intel hopes to achieve ~237 MTr/mm^2 with their 7nm process (https://www.anandtech.com/show/13405/intel-10nm-cannon-lake-and-core-i3-8121u-deep-dive-review/3), which if they do, would blow away TSMC's current 5nm process at ~173MTr/mm^2. Of course two years will give TSMC to catch up, but still.

So I wonder if Intel was simply biting off more than they could chew.

 

[mattkiss]

In either case, that timeline is concerning in the face of AMD's continued execution with its EPYC data center chips – AMD's roadmaps outline its 5nm Genoa processors coming to market before the end of 2022.

I'm a little confused...the slides from the AMD roadmap article link show Zen 4 / 5nm cpus coming by the end of 2021, not 2022.

 

[rgd1101]

I'm a little confused...the slides from the AMD roadmap article link show Zen 4 / 5nm cpus coming by the end of 2021, not 2022.

which one? I see 2022
there multiple roadmap. some zen3, some zen4

 

[KeyserTomassi]

What about the 7nm Meteor Lake for the H-series mobile CPUs? Should we expect those in 2023 as well?

 

[mattkiss]

which one? I see 2022
there multiple roadmap. some zen3, some zen4

Timelines are from left to right. A "2022" at the end doesn't mean the end of 2022, it means the end of 2021/beginning of 2022. Just look at the second slide down titled "CPU Roadmap." The last date is 2021, but Zen 3 came out it 2020.

 

[rgd1101]

there still more zen3 / zen3+ coming. the APU, Zen 3+

and the road map is from AMD. might be an old one?

 

[Geef]

Your all wrong guys! 😉

You forgot to calculate the multiple years from Intel's first announcement date of 10nm to actual production. You gotta add at minimum a few more years to make it better! 😛

 

[escksu]

Yes, I have mentioned this numerous times to the AMD fans who keep thinking Intel 14nm is vastly inferior and keep laughing.....

There is NOTHING stopping Intel from using TSMC too!! This means any advantage AMD has will be gone once Intel uses TSMC....

 

[InvalidError]

There is NOTHING stopping Intel from using TSMC too!!

Yes there is: available wafer starts.

AMD, Apple and most other TSMC clients secured their wafer starts years ago and likely have priority dibs on most capacity that gets released by other clients. Intel as one of the newer clients should be near the end of queue for everything.

 

[dave.jeffers]

TSMCs goal is to TAKE high margin silicon production from Intel fabs, into its own. Intel MIGHT be making 7nm processors by 2023, at which time, AMD will be on 3nm with TSMC. TSMC has no interest in supporting Intel. TSMC has an interest in taking Intel's high-margin chip production, and they will use AMD as their weapon to accomplish this. Intel fabs could be re-tooled and managed with TSMC technology and know-how, but make no mistake-TSMCs goal is to bleed out Intel and take their high-margin chip production, and money. TSMC is not going to assist Intel in keeping Intel's margins and production intact. Their goal is to take production, market share, and money from Intel.

 

[spongiemaster]

Yes there is: available wafer starts.

AMD, Apple and most other TSMC clients secured their wafer starts years ago and likely have priority dibs on most capacity that gets released by other clients. Intel as one of the newer clients should be near the end of queue for everything.

I would think what a company is willing to pay for wafers would impact their slot in the queue more than loyalty. Apple doesn't get preferred treatment and first access to new nodes because TSMC likes them. Apple and AMD are reportedly in the same ballpark for wafer starts, yet Apple is a quarter of TSMC's revenue, while AMD is under 10%. If Intel is willing to pay Apple prices, I bet Intel will be leapfrogging some companies in the priority queue for open capacity. Intel likely isn't going to need massive amounts of wafers if they're only having TSMC produce their top of the line premium CPU's. Intel definitely has the cash and they have the name brand recognition to be able to charge more for whatever they sell to absorb some of those higher costs. Not what Apple is able to charge, but certainly more than AMD can.

 

[PapaCrazy]

Yes there is: available wafer starts.

AMD, Apple and most other TSMC clients secured their wafer starts years ago and likely have priority dibs on most capacity that gets released by other clients. Intel as one of the newer clients should be near the end of queue for everything.

I'm not looking forward to a future where Intel chips are even less plentiful than AMD chips... and production of a vast ratio of world-wide silicon in so few hands.

 

[InvalidError]

I would think what a company is willing to pay for wafers would impact their slot in the queue more than loyalty.

No amount of money from Intel or anyone else can reduce AMD's or anyone else's allocation below whatever amount they are committed to. Since TSMC's 5-7nm fabs are booked over a year in advance and most of TSMC's customers are desperate for more starts though, Intel would have little means of securing more wafer starts than whatever they may have signed up other than bidding on whatever few cancellations of meaningful scale may happen.

I'm not looking forward to a future where Intel chips are even less plentiful than AMD chips... and production of a vast ratio of world-wide silicon in so few hands.

Well, Intel's 7nm is allegedly back on track and will likely be close to TSMC's 5nm if true. In that case, I wouldn't worry too much about Intel relying on TSMC for its flagship products longer than absolutely necessary.

If this plays out like 10nm where Intel says 10nm is on track only to announce that it ran into a brick wall for a year, rinse and repeat for three years and still have a sub-par process that needs two more years to become decent, then yes, that could be the end of Intel's fab adventures.

 

[Co BIY]

Yes there is: available wafer starts.

AMD, Apple and most other TSMC clients secured their wafer starts years ago and likely have priority dibs on most capacity that gets released by other clients. Intel as one of the newer clients should be near the end of queue for everything.

Intel is actually a long time TSMC customer and accounts for almost as much revenue as AMD already. Pretty sure this announcement by intel will not be a surprise to anyone at TSMC.

All of these companies are going to be maximizing the profits of their companies and sometimes with a pretty short term outlook. (Future profits are heavily discounted in planning for many good reasons - "a bird in hand" ) They will do what makes them the most money now. For TSMC intel is a great client add. The beauty of the foundry model is that they always win the no matter who designs the technology for them to build.

 

[spongiemaster]

No amount of money from Intel or anyone else can reduce AMD's or anyone else's allocation below whatever amount they are committed to. Since TSMC's 5-7nm fabs are booked over a year in advance and most of TSMC's customers are desperate for more starts though, Intel would have little means of securing more wafer starts than whatever they may have signed up other than bidding on whatever few cancellations of meaningful scale may happen.

Uh, yea. That's why I said open capacity. TSMC is reportedly booked through the 2nd half of 2021 for 5nm and 7nm. From the article:

"the company will also release other lines of CPUs in 2023 that will use CPU cores with an as-yet-unspecified process node from TSMC. "

Intel is definitely not using TSMC's 7nm in 2023. It's likely that Intel will be bidding for TSMC's 3nm if it is on schedule. They have the funds to outbid every one, minus Apple, if they wanted to. Otherwise a more refined 5nm could be a possibility.

 

[Co BIY]

Well, Intel's 7nm is allegedly back on track and will likely be close to TSMC's 5nm if true. In that case, I wouldn't worry too much about Intel relying on TSMC for its flagship products longer than absolutely necessary.

It is likely that at some point the most "cutting -edge high-performance" production node or method will not be the most profitable. Because of the steady progression of Moore's Law the two have so far tracked together. Performance and performance/$ have paced each other.

What if it becomes like the production of High-end supercars where the bang for the buck is only worth it for a small number of purchasers or use cases. Then it makes perfect sense to cede "process leadership" for profitability.

Even better, rebrand a Threadripper as an "i11", package it in a Buckyball shaped case with blue LEDs, incentivize the OEMs, talk about the x86 ecosystem support from the undisputed king of silicon and market it for 3 times what it will sell for as an AMD product.

If this plays out like 10nm where Intel says 10nm is on track only to announce that it ran into a brick wall for a year, rinse and repeat for three years and still have a sub-par process that needs two more years to become decent, then yes, that could be the end of Intel's fab adventures.

If you mean they could flounder their way to multiple back-to-back years of record profitability, again. I think that you might be right.

 

[zodiacfml]

meh. it looks like we're going to see the same narrative of AMD vs Intel for several years until Intel has outsourced most of their products to TSMC

 

[everettfsargent]

If we measure how advanced nodes are by transistor density, Intel's 10nm is actually better than TSMC and Samsung's first generation 7nm (https://en.wikipedia.org/wiki/7_nm_process#7_nm_process_nodes_and_process_offerings)

Intel hopes to achieve ~237 MTr/mm^2 with their 7nm process (https://www.anandtech.com/show/13405/intel-10nm-cannon-lake-and-core-i3-8121u-deep-dive-review/3), which if they do, would blow away TSMC's current 5nm process at ~173MTr/mm^2. Of course two years will give TSMC to catch up, but still.

So I wonder if Intel was simply biting off more than they could chew.

If one vendor is strictly using 2D planar transistor technologies and the other vendor is using 3D stacking technologies (Foveros) that would be called ... wait for it ... any minute now ... almost there ... stacking the decks! An at least 26-month old guesstimate from Intel and not a single 7nm desktop CPU to be available for perhaps another 22 months (Q1 2023) if we are lucky. Your ratio is also off as we currently have ~100 Tr/nm^2 from both vendors (one calls it 7nm and the other calls it 10nm), (10/7)^2 = 100/49 = 2.04 while your low ball versus high ball numbers gives 2737^2/(2375^2) = .2.26. One vendor decides to spread the heat out over a larger area using a smaller process node chiplet while the other vendor is hell bent on a more or less monolithic die and then stacking that design into multiple layers. I think Intel hired Captain Peter "Wrongway" Peachfuzz.

 

[ingtar33]

Well, Intel's 7nm is allegedly back on track

when they quote a 2023 product to market date it's not on track. they clearly stated they eliminated the "major" problem, they didn't say the process was problem free. it's why they quoted 2+ years out.

 

[TerryLaze]

Your ratio is also off as we currently have ~100 Tr/nm^2 from both vendors (one calls it 7nm and the other calls it 10nm),

What we have is 100 for high leakage and 65 for high performance from TSMC ,and 100 from intel...period.
https://en.wikichip.org/wiki/7_nm_lithography_process#Std_Cells

Based on WikiChip's own analysis, the dense cells come at around 91.2 MTr/mm² while the less dense, high-performance cells, are calculated at around 65 MTr/mm².

 

[ginthegit]

If we measure how advanced nodes are by transistor density, Intel's 10nm is actually better than TSMC and Samsung's first generation 7nm (https://en.wikipedia.org/wiki/7_nm_process#7_nm_process_nodes_and_process_offerings)

Intel hopes to achieve ~237 MTr/mm^2 with their 7nm process (https://www.anandtech.com/show/13405/intel-10nm-cannon-lake-and-core-i3-8121u-deep-dive-review/3), which if they do, would blow away TSMC's current 5nm process at ~173MTr/mm^2. Of course two years will give TSMC to catch up, but still.

So I wonder if Intel was simply biting off more than they could chew.

It was all wind. Intels 10nm process was in theory better because of the placing of 10nm Transistors on top of each other, causing a 3 transistor density instead of 2 per unit area. The problem with this technique is that it now creates twice as much heat per unit transistor space, and the cooling issue cannot be solved due to Convection alone, because the heat of the lowest one has to travel through two transistors before it can get to the top one meaning that it can get damaged easily at the bottom. Not only that. Intel never figured out a method to reliably connect all the transistors individually (gate connection), and Cobalt was their last ditch effort to resolve that. But cobalt, like other crystalline type metals are difficult to shape at the Nano level and intels 10nm did in theory have the advantage, but could only ever be tapped out with very low success per yield. So bye bye 10nm.

I guarantee that the new 7nm node they are generating are normal 7nm designs based on the current TSMC tech and have no advantage.

Samsung however, has resolved Intels issue with its 5nm design and designed with success what Intel could not, and leaving Intel trailing even further behind.
So unless Intel can create a refinery for the rare earth materials not in China, intel is going to have to suck up its loss over the last 7 years and cry. And go back to conventional designs now, 2-5 years behind the competition.

 

[dalek1234]

Intel says 7nm in 2023. How many times has Intel said something is coming in year/quarter x but it never came? I lost count.

 

[Deleted member 431422]

Intel announced that it has solved its 7nm production issues and that Meteor Lake and Granite Rapids chips will come to market in 2023.

Intel Fixes 7nm, Meteor Lake and Granite Rapids Coming in 2023 : Read more

That's good to hear. I support AMD, but recently they behave similarly to Intel when it had leading position.
Competition from will keep them on their toes. I'm looking forward to Intel's APU's in particular. If their design is better and they can keep power consumpion closer to advertised TDP I'm switching to team blue.

 

[BogdanH]

Intel says 7nm in 2023. How many times has Intel said something is coming in year/quarter x but it never came?

It's a marketing speech which should also keep stock holders in good mood. At the same time it serves as a motivation for "moving forward" -is something most companies do.
For us (consumers) however, it's quite simple... in year 2023 we will buy whatever is better and hopefully cheaper. That is, doesn't really matter if it's Intel, AMD or some manufacture in Nepal.

I support AMD, but recently they behave similarly to Intel when it had leading position.

That's to be expected: AMD actually has quite better product and uses the opportunity to make good money out of it.. finally (for AMD). In short, right now AMD is the price setter -and will stay for at least next 4-5 years in my opinion. And after that? Don't ask me

The way I see news, AMD seems to have much clearer visions/goals for near future than Intel. To explain what I mean.. while Intel says "we need to.. must be done.. should be solved.. we plan to.." etc., AMD says "we will make it" -and because of that reason, I predict bright future for AMD in next years.