Don't the previous roadmaps such as that provided in the article show both "External" and "Intel 18A" as options for Lunar Lake & Beyond? Who's to say that Lunar Lake isn't entirely in the "External" category?Reply
Exactly, I'm surprised the article doesn't even entertain the idea of that being pure TSMC silicon shown on the stage yesterday, which some might say would be a masterclass in product roadmap deception.Reply
Riffusion is, at its core, Stable Diffusion finetuned to generate spectrograms. That is really cool, but Intel is "double dipping" here, essentially showing off the same model architecture twice.
Not that its anything to sneeze at, as Riffusion is quite a complex pipe!Reply
"However, other disclosures from Intel today indicate that they’re only going to be starting risk production of 18A silicon in Q1’2024. Which means that for Lunar Lake to be working today, it can’t be on 18A."
Intel will be fabbing test dies long before they even enter risk production. If they ended up with a 'golden sample' that worked far better than they were expecting, I could see them flexing by showing it off in an integrated system. Reply
I was sorta thinking the same thing: risk production obviously isn't mass production, but it is more than a few chips, so yeah, Intel is probably already doing test wafers and maybe got a few dies back that just happened to be super good and stable hahaReply
While I didn't attach it to this article (so that it wasn't overflowing with images), there is another slide where Intel indicates the first 18A design to go to the fab won't be until Q1'24.
@Ryan Smith: that slide definitely is important to show, because yeah, it doesn't say early risk production but the actual first silicon at all won't be going to the 18A fabs until 2024. That would mean "Lunar Lake" is definitely Intel 20A now. Thank you for sharing!Reply
In my understanding same as Intel 4/3 Intel 20A/18A is High performance/High Density node. The design library is not that much different to one another, or 18A is just a completed set of library vs a custom internal node.
What is more worrying me is that Intel going back to sail pace node advancement, no CFET discussed, just blank, as rumors now TSMC N2 is hitting delay, then that means that TSMC N2 will be on ASML High NA, since Intel will received High NA in December 2023, then if by Jan 2024 TSMC received a High NA machine, this will meant that 1 year for TSMC to produce N2.
If TSMC really delay (common after Chang left) then the only positive news is that TSMC N2 is going to produce in High NA (which might not required double patterning.
But the bad news is that Apple will not be slow down, they will not let intel have 2 years advantage and will move to Intel Foundry, there are rumor as well that TSMC slow down the fab build for 2nm, if that is true that Apple left therefore low demand for TSMC N2 then, we might see Intel Sail Pace again, which is not that great.Reply
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
9 Comments
Back to Article
Khato - Tuesday, September 19, 2023 - link
Don't the previous roadmaps such as that provided in the article show both "External" and "Intel 18A" as options for Lunar Lake & Beyond? Who's to say that Lunar Lake isn't entirely in the "External" category? Replyydeer - Wednesday, September 20, 2023 - link
Exactly, I'm surprised the article doesn't even entertain the idea of that being pure TSMC silicon shown on the stage yesterday, which some might say would be a masterclass in product roadmap deception. Replymy_wing - Thursday, September 21, 2023 - link
Because it is not.Because Pat already show am 18A wafer. Where is TSMC wafer????? Where is TSMC N3E wafer ????
So honestly, that is intel baked chip not from TSMC. Reply
brucethemoose - Wednesday, September 20, 2023 - link
Riffusion is, at its core, Stable Diffusion finetuned to generate spectrograms. That is really cool, but Intel is "double dipping" here, essentially showing off the same model architecture twice.Not that its anything to sneeze at, as Riffusion is quite a complex pipe! Reply
edzieba - Wednesday, September 20, 2023 - link
"However, other disclosures from Intel today indicate that they’re only going to be starting risk production of 18A silicon in Q1’2024. Which means that for Lunar Lake to be working today, it can’t be on 18A."Intel will be fabbing test dies long before they even enter risk production. If they ended up with a 'golden sample' that worked far better than they were expecting, I could see them flexing by showing it off in an integrated system. Reply
NextGen_Gamer - Wednesday, September 20, 2023 - link
I was sorta thinking the same thing: risk production obviously isn't mass production, but it is more than a few chips, so yeah, Intel is probably already doing test wafers and maybe got a few dies back that just happened to be super good and stable haha ReplyRyan Smith - Wednesday, September 20, 2023 - link
While I didn't attach it to this article (so that it wasn't overflowing with images), there is another slide where Intel indicates the first 18A design to go to the fab won't be until Q1'24.https://images.anandtech.com/doci/20060/6023140.jp... Reply
NextGen_Gamer - Thursday, September 21, 2023 - link
@Ryan Smith: that slide definitely is important to show, because yeah, it doesn't say early risk production but the actual first silicon at all won't be going to the 18A fabs until 2024. That would mean "Lunar Lake" is definitely Intel 20A now. Thank you for sharing! Replymy_wing - Thursday, September 21, 2023 - link
In my understanding same as Intel 4/3 Intel 20A/18A is High performance/High Density node. The design library is not that much different to one another, or 18A is just a completed set of library vs a custom internal node.What is more worrying me is that Intel going back to sail pace node advancement, no CFET discussed, just blank, as rumors now TSMC N2 is hitting delay, then that means that TSMC N2 will be on ASML High NA, since Intel will received High NA in December 2023, then if by Jan 2024 TSMC received a High NA machine, this will meant that 1 year for TSMC to produce N2.
If TSMC really delay (common after Chang left) then the only positive news is that TSMC N2 is going to produce in High NA (which might not required double patterning.
But the bad news is that Apple will not be slow down, they will not let intel have 2 years advantage and will move to Intel Foundry, there are rumor as well that TSMC slow down the fab build for 2nm, if that is true that Apple left therefore low demand for TSMC N2 then, we might see Intel Sail Pace again, which is not that great. Reply