Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads

Tigerick · Aug 22, 2022

As Hot Chips 34 starting this week, Intel will unveil technical information of upcoming Meteor Lake (MTL) and Arrow Lake (ARL), new generation platform after Raptor Lake. Both MTL and ARL represent new direction which Intel will move to multiple chiplets and combine as one SoC platform.

MTL also represents new compute tile that based on Intel 4 process which is based on EUV lithography, a first from Intel. Intel expects to ship MTL mobile SoC in 2023.

ARL will come after MTL so Intel should be shipping it in 2024, that is what Intel roadmap is telling us. ARL compute tile will be manufactured by Intel 20A process, a first from Intel to use GAA transistors called RibbonFET.

Comparison of upcoming Intel's U-series CPU: Core Ultra 100U, Lunar Lake and Panther Lake

Model	Code-Name	Date	TDP	Node	Tiles	Main Tile	CPU	LP E-Core	LLC	GPU	Xe-cores
Core Ultra 100U	Meteor Lake	Q4 2023	15 - 57 W	Intel 4 + N5 + N6	4	tCPU	2P + 8E	2	12 MB	Intel Graphics	4
?	Lunar Lake	Q4 2024	17 - 30 W	N3B + N6	2	CPU + GPU & IMC	4P + 4E	0	12 MB	Arc	8
?	Panther Lake	Q1 2026 ?	?	Intel 18A + N3E	3	CPU + MC	4P + 8E	4	?	Arc	12

Comparison of die size of Each Tile of Meteor Lake, Arrow Lake, Lunar Lake and Panther Lake

	Meteor Lake	Arrow Lake (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop & Mobile H&HX	Mobile U Only	Mobile H
Process Node	Intel 4	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	8P + 16E	4P + 4E	4P + 8E
LLC	24 MB	36 MB ?	12 MB	?
tCPU	66.48
tGPU	44.45
SoC	96.77
IOE	44.45
Total	252.15

Intel Core Ultra 100 - Meteor Lake

As mentioned by Tomshardware, TSMC will manufacture the I/O, SoC, and GPU tiles. That means Intel will manufacture only the CPU and Foveros tiles. (Notably, Intel calls the I/O tile an 'I/O Expander,' hence the IOE moniker.)

cannedlake240 · Dec 7, 2024

N4X by TSMC numbers isn't really that far behind N3E, P in perf. N2 is 10-15% over that, but 18A-P is also another 10% perf. So overall 18A vs TSMC is slightly behind in density and perf but significantly (>15%) behind in perf/W?

oak8292 · Dec 7, 2024

511 said:
I mean smartphone SOC for the complete node package of HP/SOC Libraries

Is this licensed ARM cores like Mediatek? OS? Apps? Software moat? They developed x86 Android back when they were doing contra revenue at 14 nm but the software ecosystem-system is way past that. Do you expect Intel to be vertically integrated like Apple or reference designs and somebody else figures out software compatibility?

ajsdkflsdjfio · Dec 7, 2024

vanplayer said:
Internal data of 18A was roughly equals to N4X. Maybe there is still some chance to improve but yeah that's so far it was looked like.

???? Powervia plus GAA and it's barely better than N5 or even their own intel 4/3? Where do you get this stuff lol. If 18a was this mediocre you wouldn't see future Intel products being produced on it as they would've known these projections years ago. Pat would've also been fired a long time ago if earlier projections showed 18a at N4X levels after betting the whole company on it... What you are claiming is equivalent to claiming that Intel is going to implode within the decade.

poke01 · Dec 7, 2024

18A should be roughly N3P equivalent.

cannedlake240 · Dec 7, 2024

ajsdkflsdjfio said:
???? Powervia plus GAA and it's barely better than N5 or even their own intel 4/3? Where do you get this stuff lol. If 18a was this mediocre you wouldn't see future Intel products being produced on it as they would've known these projections years ago. Pat would've also been fired a long time ago if earlier projections showed 18a at N4X levels after betting the whole company on it... What you are claiming is equivalent to claiming that Intel is going to implode within the decade.

Nah, N3P isn't leaps above N4X. It's at best 20% better than N5. 18A still has 3Q to go til production to match or even exceed N3P. N2 brings 10-15% better perf over N3E, meaning the perf gap between 18A and N2 could be in single digits. Now perf/W could be a much larger disadvantage for Intel and density is probably slightly below too.
Decent progress going from 10nm to half node behind in 3 years

DrMrLordX · Dec 7, 2024

vanplayer said:
Internal data of 18A was roughly equals to N4X. Maybe there is still some chance to improve but yeah that's so far it was looked like.

Where are we getting this internal data?

ajsdkflsdjfio · Dec 7, 2024

cannedlake240 said:
Nah, N3P isn't leaps above N4X. It's at best 20% better than N5. 18A still has 3Q to go til production to match or even exceed N3P. N2 brings 10-15% better perf over N3E, meaning the perf gap between 18A and N2 could be in single digits. Now perf/W could be a much larger disadvantage for Intel and density is probably slightly below too.
Decent progress going from 10nm to half node behind in 3 years

Performance is the only metric in which N4X is close to N3(non-P) and N3P widens that gap more. Unless vanplayer was referring solely to performance, he's talking BS. Besides, if you are talking about performance why compare 18A to N4X rather than N3P. I think it's much more likely 18A performance-wise is at east at N3P levels if not higher given intel's history of focusing on performance for their nodes, If only one thing about 18A were superior to N3 it'd be performance.

Also wym by 3Q till production, time to improve. Their node specs are already 99% defined only 3Q away from production. What they are focusing right now isn't changing the silicon but improving yields and getting ready for mass production there is no magical extra 10% performance to be squeezed out.

FlameTail · Dec 7, 2024

18A

Performance/Power = N2
Density = N3P

My guess.

cannedlake240 · Dec 7, 2024

ajsdkflsdjfio said:
Performance is the only metric in which N4X is close to N3(non-P) and N3P widens that gap more. Unless vanplayer was referring solely to performance, he's talking BS. Besides, if you are talking about performance why compare 18A to N4X rather than N3P. I think it's much more likely 18A performance-wise is at east at N3P levels if not higher given intel's history of focusing on performance for their nodes, If only one thing about 18A were superior to N3 it'd be performance.

Also wym by 3Q till production, time to improve. Their node specs are already 99% defined only 3Q away from production. What they are focusing right now isn't changing the silicon but improving yields and getting ready for mass production there is no magical extra 10% performance to be squeezed out.

Yeah I overestimated how good N4X is. 18A vs N2P I guess is going to be 10% worse perf and density and up to 25% lower perf/W. Still good progress over 10nm Imo. P1280/14A even has a chance to take the lead in 2027

511 · Dec 7, 2024

oak8292 said:
Is this licensed ARM cores like Mediatek? OS? Apps? Software moat? They developed x86 Android back when they were doing contra revenue at 14 nm but the software ecosystem-system is way past that. Do you expect Intel to be vertically integrated like Apple or reference designs and somebody else figures out software compatibility?

I meant for qualcomm SOC also one of Intels developer told me that Ingel internally has a AOSP build that they still maintain

511 · Dec 7, 2024

cannedlake240 said:
Yeah I overestimated how good N4X is. 18A vs N2P I guess is going to be 10% worse perf and density and up to 25% lower perf/W. Still good progress over 10nm Imo. P1280/14A even has a chance to take the lead in 2027

I have heard N3X level of performance from Semiwiki for 18A the Sram is same as N3.
I estimate it at like 5% behind N2 in perf and 15% behind in Area 18AP will have +5-10% perf increase (like slightly better than N2) and +5% density increase but that is a massive improvement vs Intel 10nm and TSMC N5 ( 1 whole node difference)
It is more than enough for their internal use granted design doesn't F*** Up for upcoming years to not outsource to TSMC

511 · Dec 7, 2024

This image makes more sense Intel is small buyer for N2 majority is 18A going forward

SiliconFly · Dec 7, 2024

In a very unusual move, Pat Gelsinger himself took to Twitter/X today to address the ongoing rumor about 18A's abysmal 10% yield. In a reply to Patrick Moorhead, he confirmed that its fake news.

The most interesting part about this drama is how deep-rooted it was. Appears to be well coordinated with 100s of tech and not-so-tech sites all reporting the same in lock step. Looks like some of the Taiwanese friends are hard at work.

dttprofessor · Dec 7, 2024

cannedlake240 said:
N4X by TSMC numbers isn't really that far behind N3E, P in perf. N2 is 10-15% over that, but 18A-P is also another 10% perf. So overall 18A vs TSMC is slightly behind in density and perf but significantly (>15%) behind in perf/W?

n4p/n3b 5.7g, then n2 6.3g-6.6g?

Meteor Late · Dec 7, 2024

dttprofessor said:
n4p/n3b 5.7g, then n2 6.3g-6.6g?

Doesn't work like that, usually those performance numbers are not at the top of the voltage curve, which is where these very high frequencies sit at.

511 · Dec 7, 2024

dttprofessor said:
n4p/n3b 5.7g, then n2 6.3g-6.6g?

It's not a Intel node lmao it is usually bottom or mid curve

poke01 · Dec 7, 2024

SiliconFly said:
In a very unusual move, Pat Gelsinger himself took to Twitter/X today to address the ongoing rumor about 18A's abysmal 10% yield. In a reply to Patrick Moorhead, he confirmed that its fake news.

The most interesting part about this drama is how deep-rooted it was. Appears to be well coordinated with 100s of tech and not-so-tech sites all reporting the same in lock step. Looks like some of the Taiwanese friends are hard at work.

This all started due to a South Korean news publication? The 10% yields was fake news

igor_kavinski · Dec 7, 2024

511 said:
This image makes more sense Intel is small buyer for N2 majority is 18A going forward

View attachment 112791

Excellent! Something like this should routinely be posted in the Leading Edge thread.

Doug S · Dec 7, 2024

DrMrLordX said:
Where are we getting this internal data?

Back when Intel was doing well rumors about low yields or judging a new process as worse than TSMC's current one would be almost completely discounted, but now they are much easier to believe. Likewise due to Intel's disastrous last few years, there will be well deserved skepticism about rumors claiming things are all rosy with 18A and they've finally righted the ship.

Because of that the stock price will be much easier to move based on rumor than it would have been in the past. As a result we're inevitably going to simultaneously see leaks of real data and people making stuff up, both designed to move the stock price. We probably won't know the truth until we see 18A stuff hit the market.

OneEng2 · Dec 7, 2024

poke01 said:
18A should be roughly N3P equivalent.

Fundamentally, 18A should have much better density than N2 in practical applications (vs transistor metrics) due to BSPDN. BSPDN reportedly allows up to 30% higher density in applications where lots of power via are needed (I am guessing in high speed logic circuits).

I spend some time on Semiwiki forums with the pro's over there. They seem to believe that there are issues with BSPDN in cooling hotspots (which I admittedly don't completely understand) compared to FSPD. Additionally, they say that TSMC's A16 BSPDN implementation is more complex than Intel's 18A as well as being less dense in order to accommodate both FSPD and BSPD libraries which are backwards compatible with N2.... so A16 may have some yield issues as well.

cannedlake240 said:
Nah, N3P isn't leaps above N4X. It's at best 20% better than N5. 18A still has 3Q to go til production to match or even exceed N3P. N2 brings 10-15% better perf over N3E, meaning the perf gap between 18A and N2 could be in single digits. Now perf/W could be a much larger disadvantage for Intel and density is probably slightly below too.
Decent progress going from 10nm to half node behind in 3 years

If true, then I am guessing that the Intel statement that Panther Lake would be delivered in "Mid 2025" was incorrect?

I am guessing (based on conversations at SemiWiki) that the metrics on 18A are going to be surprisingly good. It is the yields that remain in question.

Doug S said:
Back when Intel was doing well rumors about low yields or judging a new process as worse than TSMC's current one would be almost completely discounted, but now they are much easier to believe. Likewise due to Intel's disastrous last few years, there will be well deserved skepticism about rumors claiming things are all rosy with 18A and they've finally righted the ship.

Because of that the stock price will be much easier to move based on rumor than it would have been in the past. As a result we're inevitably going to simultaneously see leaks of real data and people making stuff up, both designed to move the stock price. We probably won't know the truth until we see 18A stuff hit the market.

... and my thing is that if Intel's 18A was doing really well, they would be publishing information every week about it to keep their stock prices high. There is no reason I can think of to be hiding it under a bushel if it is really going well ...... so as such, I fear the worst. .

ondma · Dec 8, 2024

SiliconFly said:
In a very unusual move, Pat Gelsinger himself took to Twitter/X today to address the ongoing rumor about 18A's abysmal 10% yield. In a reply to Patrick Moorhead, he confirmed that its fake news.

The most interesting part about this drama is how deep-rooted it was. Appears to be well coordinated with 100s of tech and not-so-tech sites all reporting the same in lock step. Looks like some of the Taiwanese friends are hard at work.

I liked Gelsinger, but considering some of his statements recently, one has to take anything he says with a grain of salt. As someone else said, If 18A was in great shape, I dont think Gelsinger would have been forced out.

dttprofessor · Dec 8, 2024

ondma said:
I liked Gelsinger, but considering some of his statements recently, one has to take anything he says with a grain of salt. As someone else said, If 18A was in great shape, I dont think Gelsinger would have been forced out.

You belittle the BOD.

SiliconFly · Dec 8, 2024

OneEng2 said:
I am guessing (based on conversations at SemiWiki) that the metrics on 18A are going to be surprisingly good. It is the yields that remain in question.

Not yields. It's hitting volume on time that remain in question. Same happened to Intel 4 (and also Intel 3 I think). For comparison, TSMC hits volume on time every time. Intel, no so much.

LightningZ71 · Dec 8, 2024

I'm still interested in seeing what Intel3 does for Meteor Lake refresh. The specs for frequencies were promising at least.

Win2012R2 · Dec 8, 2024

SiliconFly said:
It's hitting volume on time that remain in question

And cost surely? Even if the yields are good, that can still cost too much

Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads

Senior member

Attachments

Senior member

Member

Member

Diamond Member

Senior member

Lifer

Member

Diamond Member

Senior member

Golden Member

Golden Member

Golden Member

Golden Member

Member

Senior member

Golden Member

Diamond Member

Lifer

Diamond Member

Senior member

Diamond Member

Member

Golden Member

Platinum Member

Senior member