Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/tsmc-q3-2023-results-are-good.18988/page-2
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

TSMC Q3 2023 Results are Good!

I'm just surprised we are talking about Meteor lake, when we should be talking about this statement:
internal assessment shows that our N3P -- now, I'll repeat again, N3P technology, demonstrated comparable PPA to 18A, my competitors' technology, but with an earlier time to market, better technology maturity, and much better cost. In fact, let me repeat again, our 2-nanometer technology without backside power is more advanced than both N3P and 18A


It seems like TSMC think they can do N3P in 2024, ahead of Intel 18A, which become manufacturing-ready by the end of 2024, the first product of 18A will be most likely to come out in the first half of 2025. This sounds good to Intel's competitors, like AMD and Nvidia, that they can still have a good year in 2024. TSMC really has mastered FinFET! They managed to beat first-gen GAA from both Samsung and Intel. Maybe this is no surprise considering TSMC started N3 manufacturing in 2022, only a year and a half after Pat G come back to the realm.

But I think what comes next. Apparently, Intel is getting a lot better at execution. This means that if according to the annual cadence Pat promised after 18A and Intel's future roadmap with Intel 18A, Next, Next+, Next++. This probably means that TSMC and its customers will have to deal with 18A, and Intel Next (probably manufacturing ready in late 2025 but go mass production 2026 with some chiplet done on it), which will have gen 2 GAA and gen 2 Power Via that are going to be more mature as gen 1 come out with 20A. I think for Intel being the first on new technologies is quite important. They can get to there first, then improve and refine based on that. I think this will likely create more headaches than the statement CC Wei is making because getting the first step is the most complicated among anything. Intel will be able to get its first steps taken in 2024, while TSMC makes a half step with N2 in 2025. And the other half in 2026.
 
the first product of 18A will be most likely to come out in the first half of 2025.
Not a chance. Arrow lake come out at very end of 2024 and some SKUs will be using 20A. Panther Lake come out at very end of 2025 and it will use 18A. So, really it's 2026 for 18A
 
Not a chance. Arrow lake come out at very end of 2024 and some SKUs will be using 20A. Panther Lake come out at very end of 2025 and it will use 18A. So, really it's 2026 for 18A
well, no. you forgot clearwater forest, apparently there are going to be four 18A products that're going to come out, and Clearwater Forest is one of them. as well as Falcon Shore. and Ericsson Soc. so one client-side, one datacenter (cloud), one AI/GPU, and custom SoC for network. And you can bet 18A is much different from Intel 4 (the one we see in Meteor Lake) why? because Intel 4 is a completely new process, 1st gen EUV for Intel, first mass-produced 3D capabilities., which is why I think they deliver Meteor Lake at end of 2023. Panther Lake is different, 18A and 20A are same class process node. The level of complexity they'll faced is going to be very different or easier to dealt with. I think you can see them stepping up the time window for Panther Lake.

And don't forget, Arrow Lake is basically Meteor Lake with two tiles updated. They are going to be using the same IO, SOC dies, same Foveros packaging. Why would you think they will slip to the end of 2024?
 
Not a chance. Arrow lake come out at very end of 2024 and some SKUs will be using 20A. Panther Lake come out at very end of 2025 and it will use 18A. So, really it's 2026 for 18A
and I bet that they definitely want to do that as soon as possible (releasing 18A products to market) before Apple new chips come out based on their roadmap. Clearly, they showed their intent to deliver at least four internal products (NO MATTER HOW MANY TO BE EXACT) before the first foundry designed coming out. Intel wants to give the best to itself first, which I don't think it's wrong. Not everyone is Apple. If based on what you said, 2026 for 18A, we will likely see first external 18A wafer start meaningfully in 2027, which isn't practical
1697822007464.png
 
well, no. you forgot clearwater forest, apparently there are going to be four 18A products that're going to come out, and Clearwater Forest is one of them. as well as Falcon Shore. and Ericsson Soc. so one client-side, one datacenter (cloud), one AI/GPU, and custom SoC for network. And you can bet 18A is much different from Intel 4 (the one we see in Meteor Lake) why? because Intel 4 is a completely new process, 1st gen EUV for Intel, first mass-produced 3D capabilities., which is why I think they deliver Meteor Lake at end of 2023. Panther Lake is different, 18A and 20A are same class process node. The level of complexity they'll faced is going to be very different or easier to dealt with. I think you can see them stepping up the time window for Panther Lake.

And don't forget, Arrow Lake is basically Meteor Lake with two tiles updated. They are going to be using the same IO, SOC dies, same Foveros packaging. Why would you think they will slip to the end of 2024?
Latest rumour from Moore's law says Q4 2024 for Arrow lake:
. Also, there is no rumours at all saying anything about clearwater forest being 18A
 
Latest rumour from Moore's law says Q4 2024 for Arrow lake:
. Also, there is no rumours at all saying anything about clearwater forest being 18A
:LOL: you said it's rumor, so let it be. but you didn't say clearwater forest being 18A is a rumor, so it is a fact, right?

Intel also revealed the first details of Clearwater Forest, it's second-gen efficiency Xeon that will debut in 2025. Intel skipped over its 20A process node for the more performant 18A for this new chip, which speaks volumes about its faith in the health of its future node. From Tom's Hardware

here is a fact for you
 

I did not know Intel skipped 20A in favor of 18A. I guess that is good news. So no products at 20A or just this one skipped?
 
@lefty @delong.height

Perhaps we should wait for actual launch on Intel 4 before speculating on Intel 20A, 18A

And remember: Intel said Intel 4 was in production in december 2022. which would mean Meteor lake was in production in Dec 2022. And It will be partially launched Dec 2023 (Most SKUs will launch in 2024.)

Following this Intel model. Arrow lake could be in "production" in Dec 2024.... with no products available in stores until Dec 2025.

2024 will tell us a lot about the actual roadmap vs planned roadmap
 
Following this Intel model. Arrow lake could be in "production" in Dec 2024.... with no products available in stores until Dec 2025.
Well, the Moore's law sources say that Arrow lake was originally planned to be 100% TSMC. The Intel 20A SKUs only got tacked on afterwards (I guess when they pulled it in)
 
@lefty @delong.height

Perhaps we should wait for actual launch on Intel 4 before speculating on Intel 20A, 18A

And remember: Intel said Intel 4 was in production in december 2022. which would mean Meteor lake was in production in Dec 2022. And It will be partially launched Dec 2023 (Most SKUs will launch in 2024.)

Following this Intel model. Arrow lake could be in "production" in Dec 2024.... with no products available in stores until Dec 2025.

2024 will tell us a lot about the actual roadmap vs planned roadmap
Absolutely - we will see how fast the follow on is for Intel 3 releases after Meteor Lake - seeing this jump from Intel 4 to 3 would arguably tell us about how long it would take Intel to jump from Intel 20A to 18A.

I do think that the tiles method allows for potentially faster launches, but it remains to be seen if Intel can execute on these faster launches.
 
I don't know if we can come to an observation that if entire Intel product rollout is tied to Intel Foundry Service's roadmap or progress, Intel will be in big trouble.

Intel's customers won't wait and Intel's competitors won't wait on the sideline either. At the end of each day, Intel needs competitive products to sell, no matter they're manufactured by IFS or external foundries. Pat Gelsinger is probably constantly thinking how to de-risk such challenges.
 
Taiwan Semiconductor Manufacturing's (NYSE:TSM) net revenue in October grew 34.8% month-over-month to about NT$243.2B, and year-over-year net revenue rose 15.7%. The chip giant had generated about NT$210.3B in net revenue in October 2022. Revenue for January through October (year-to-date) 2023 dipped 3.7% year-over-year to NT$1.779T.

Gonna be a great Q4!
 
Back
Top