Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/intel-nova-lake-s-uses-tsmc-n2-process-tape-out.23172/
        )

    [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] => 2021770
            [XFI] => 1050270
        )

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

Intel Nova Lake-S uses TSMC N2 process Tape-Out

Daniel Nenni

Admin
Staff member
It is reported that Intel's next-generation client CPU flagship product Nova Lake-S has been taped out at TSMC's wafer fab in Taiwan. We previously speculated based on rumors that Intel will adopt its own internal 18A process and rely on TSMC's 2nm mass production technology. But according to SemiAccurate's report, Intel has taped out a computing module on TSMC's N2 process, which means that the computing module of Nova Lake-S will most likely use both 18A and TSMC's N2 processes. One possible reason for Intel's decision is that if the 18A process cannot be delivered, or if demand is expected to be too high and internal production capacity cannot be met, Intel is building a reliable backup plan. In any case, customers can expect the product to be delivered on time in the second half of 2026, but there may be some interesting solutions behind it.

Copilot_20250711_132029.jpg


As for the specific date, it will take several months from Tape-Out to the delivery of the final product. Currently, the chips completed by Tape-Out are undergoing power-on testing in Intel laboratories, running various tests to test the performance of the chips under multiple uses and check the correctness of their operation. Normally it takes a few weeks to a month to power on, and final mass production will begin a few months later. It will take another two to three months to manufacture and ship, which means that Nova Lake-S will most likely be released in the third quarter of 2026. It should be reminded that this CPU will integrate up to 52 cores (16 P cores, 32 E cores and 4 LPE cores), with an 8,800MT/s memory controller, as well as Xe3 Celestial for graphics rendering and Xe4 Druid for media and display tasks. This undoubtedly makes it an extremely attractive product, but due to its heterogeneous complexity, its manufacturing difficulty is also quite high.

 
TSMC N2 is going to absolutely dominate with 100% support from the ecosystems. Incredible!


 
No real surprise here. At the BoA Technology Conference back on June 3rd Michelle Johnston-Holthaus said: "And so, I've been very public that come our next generation of product, Nova Lake, I'm using both TSMC and Intel Foundry. And I'm doing that because I think it allows me to deliver a more competitive product for our customers. And at the end of the day, the best product wins."
 
TSMC N2 is going to absolutely dominate with 100% support from the ecosystems. Incredible!


Nice Table Dan 😍😍 but i think it needs to be updated since 20A is no more we can compare 18A with Intel 3
 
No real surprise here. At the BoA Technology Conference back on June 3rd Michelle Johnston-Holthaus said: "And so, I've been very public that come our next generation of product, Nova Lake, I'm using both TSMC and Intel Foundry. And I'm doing that because I think it allows me to deliver a more competitive product for our customers. And at the end of the day, the best product wins."

Or the best yielding product wins. Hard to compete with TSMC on yield.
 
TSMC N2 is going to absolutely dominate with 100% support from the ecosystems. Incredible!
I remember you previously mentioned that Intel had no plans to use TSMC’s N2, but it now seems the strategy has shifted. Does this imply that Intel 18A may not be sufficiently mature, even for internal applications?
 
I remember you previously mentioned that Intel had no plans to use TSMC’s N2, but it now seems the strategy has shifted. Does this imply that Intel 18A may not be sufficiently mature, even for internal applications?

I said Intel had not signed a TSMC N2 agreement. That was more than a year ago. I was told that all but one TSMC N3 customer had signed an N2 agreement and that was Intel.

I had hoped Intel would use 18A for all internal designs but clearly that is not the case. I do think that TSMC N2 was plan B and now with the reorg and cutbacks it is a very good plan B.
 
I said Intel had not signed a TSMC N2 agreement. That was more than a year ago. I was told that all but one TSMC N3 customer had signed an N2 agreement and that was Intel.
I had hoped Intel would use 18A for all internal designs but clearly that is not the case. I do think that TSMC N2 was plan B and now with the reorg and cutbacks it is a very good plan B.
I agree it's a good plan. Right now, Intel's biggest challenges lie in resolving its financial issues and boosting product competitiveness.
It also reminds me of a joke: using TSMC is like trying a drug—once you start, it's hard to quit.
 
TSMC N2 is going to absolutely dominate with 100% support from the ecosystems. Incredible!


I really like these process pages, especially the 'development timeline' section is unique.
 
But that's compared to N2, not N2P or A16 -- which makes A14 look good but is kind of cheating... ;-)
Got it. I was thinking of what the improvement will be from the current A18 Pro chip, which is N3E. Still seems amazing to me. :)
 
Back
Top