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

[Tigerick]

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	8 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 (20A)	Arrow Lake (N3B)	Arrow Lake Refresh (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Desktop Only	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2025 ?	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	8P + 32E	4P + 4E	4P + 8E
LLC	24 MB	24 MB ?	36 MB ?	?	8 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.)

 

[SteinFG]

Idk why people argue whether or not intel will make 8+16 meteor lake, this picture clearly shows 3 compute tile sizes
2+8 and 6+8 already confirmed, and 8+16 is left.
But what rubs me the wrong way is the "Expanded IO tile".
For socket compatibility all IO must be the same throughout the lineup, but this suggests that 6+8 will get less IO than 8+16? kinda strange.
Maybe this IO tile is just a proposal

 

[IntelUser2000]

Idk why people argue whether or not intel will make 8+16 meteor lake, this picture clearly shows 3 compute tile sizes
2+8 and 6+8 already confirmed, and 8+16 is left.
But what rubs me the wrong way is the "Expanded IO tile".
For socket compatibility all IO must be the same throughout the lineup, but this suggests that 6+8 will get less IO than 8+16? kinda strange.
Maybe this IO tile is just a proposal
View attachment 66821

Remember that is a Foveros presentation which includes Arrowlake, not just Meteorlake.

 

[SteinFG]

Remember that is a Foveros presentation which includes Arrowlake, not just Meteorlake.

This was during the section on meteor lake, not generally foveros https://www.servethehome.com/intel-disaggregates-client-chips-with-meteor-lake-hc34/

 

[IntelUser2000]

This was during the section on meteor lake, not generally foveros https://www.servethehome.com/intel-disaggregates-client-chips-with-meteor-lake-hc34/

Yup, you are right. The middle die could also be 4+8. I guess we still don't know for sure but skipping MTL for high end and getting Arrowlake out very soon after would be the only way to meet their "5 nodes in 4" years.

@Exist50 Going to 20A from N3 would be the 180 degree change under Gelsinger to not abandon fabbing chips internally but gain leadership and strengthen it.

 

[jpiniero]

@Exist50 Going to 20A from N3 would be the 180 degree change under Gelsinger to not abandon fabbing chips internally but gain leadership and strengthen it.

Yeah... but Intel says a lot of things. A lot of things that turn out not to be anywhere near true. Hell, check out these two articles just on 7 nm.

https://www.anandtech.com/show/13683/intel-euvenabled-7nm-process-tech-is-on-track (from 2018)
https://www.anandtech.com/show/1431...r-products-in-1h-2020-accelerated-7nm-in-2021 (from 2019)

Getting Free Money from the Government is great, but you still have to sell products. And be able to make them.

 

[DrMrLordX]

Idk why people argue whether or not intel will make 8+16 meteor lake, this picture clearly shows 3 compute tile sizes

That slide is pretty vague. Also if you look at it, the second and third mockups have a compute tile of the same area. It definitely hints at a slightly smaller SoC, larger I/O, and larger iGPU for the third pic. Whatever that means!

 

[Zucker2k]

Yeah... but Intel says a lot of things. A lot of things that turn out not to be anywhere near true. Hell, check out these two articles just on 7 nm.

https://www.anandtech.com/show/13683/intel-euvenabled-7nm-process-tech-is-on-track (from 2018)
https://www.anandtech.com/show/1431...r-products-in-1h-2020-accelerated-7nm-in-2021 (from 2019)

Getting Free Money from the Government is great, but you still have to sell products. And be able to make them.

2018 and 2019? That's practically a different era in Intel management history.

 

[dullard]

Yeah... but Intel says a lot of things. A lot of things that turn out not to be anywhere near true. Hell, check out these two articles just on 7 nm.

https://www.anandtech.com/show/13683/intel-euvenabled-7nm-process-tech-is-on-track (from 2018)
https://www.anandtech.com/show/1431...r-products-in-1h-2020-accelerated-7nm-in-2021 (from 2019)

Those articles are not yet proof of Intel being not "anywhere near true". They may become examples, but the clock has not yet ran out.

Lets consider your second article from May 2019: "While Intel's first 7 nm product will be launched in 2021, Intel stresses that high-volume manufacturing (HVM) using the technology will begin in 2022 when the technology will be used not only for a server GPU, but also a server CPU. So, expect more 7 nm products three or four years down the road." When is 3 to 4 years from that article? That would be May 2022 to May 2023. So, if Intel gets a 7 nm server GPU and CPU out in the next 8 months, then that article was spot on.

The first article is harder to analyze since there aren't any real details. The only Intel quote of relevance is "So, we are very, very focused on getting 7 nm out according to our original internal plans." (1) We don't know what the internal plans are and (2) being focused on meeting those plans is not a promise to meet them.

 

[IntelUser2000]

2018 and 2019? That's practically a different era in Intel management history.

Exactly. Don't know why @jpiniero constantly ignores this. BK management did absolutely insane things like fire the server validation team(according to @Exist50 the entire team) and put no rehire! He fired Mooly Eden and Dadi Perlmutter for fear that they would take over, since BK was doing a horrible job. Remember those two guys are from the Haifa team after Pentium M and Core 2 success.

Many people said if there was one company that could make the ambitious plans of 10nm work, it would be the team in Intel.

Yes, tons of people also left during the era. Not only BK, but around the time he became CEO Otellini fired 10K employees for cost cutting. People that laser focus on tech without realizing they are created by people are very naive. So the problems span over a decade!

Under Gelsinger, many came back. He hired 12,000 employees, most in the engineering and technical side. Even the skeptics like Charlie Demerjian is changing his tune and even said the new management is "strong". Gelsinger is the same guy that was got 99% CEO approval(#1) from employees while at VMware.

 

[DrMrLordX]

Under Gelsinger, many came back. He hired 12,000 employees, most in the engineering and technical side. Even the skeptics like Charlie Demerjian is changing his tune and even said the new management is "strong". Gelsinger is the same guy that was got 99% CEO approval(#1) from employees while at VMware.

That's all well and good, but we've yet to see the actual fruit of these changes. Admittedly it will take time. But the real issue isn't just who Gelsinger hired but what it is that these new hires and rehires have to work with now that people like Murthy are gone.

 

[jpiniero]

Yes, tons of people also left during the era. Not only BK, but around the time he became CEO Otellini fired 10K employees for cost cutting. People that laser focus on tech without realizing they are created by people are very naive. So the problems span over a decade!

Funny you say that. I am expecting big time layoffs at Intel by the end of the year. It is only a prediction at this point but the server business looks about ready to implode.

 

[shady28]

Exactly. Don't know why @jpiniero constantly ignores this. BK management did absolutely insane things like fire the server validation team(according to @Exist50 the entire team) and put no rehire! He fired Mooly Eden and Dadi Perlmutter for fear that they would take over, since BK was doing a horrible job. Remember those two guys are from the Haifa team after Pentium M and Core 2 success.

Many people said if there was one company that could make the ambitious plans of 10nm work, it would be the team in Intel.

Yes, tons of people also left during the era. Not only BK, but around the time he became CEO Otellini fired 10K employees for cost cutting. People that laser focus on tech without realizing they are created by people are very naive. So the problems span over a decade!

Under Gelsinger, many came back. He hired 12,000 employees, most in the engineering and technical side. Even the skeptics like Charlie Demerjian is changing his tune and even said the new management is "strong". Gelsinger is the same guy that was got 99% CEO approval(#1) from employees while at VMware.

I actually have a lot of faith in Gelsinger's ability.

It's one of the reasons I believe 2020-2025 will play out a lot like 2007-2012. And if you look at that history closely - the parallel for the first two years is strikingly uncanny.

I could be wrong, but if Intel sticks on its roadmap more or less, I won't be. AMD has to stay one node ahead of intel to maintain a performance lead, and they won't be able to do that if Intel can stay near its current roadmap.

 

[moinmoin]

Funny you say that. I am expecting big time layoffs at Intel by the end of the year. It is only a prediction at this point but the server business looks about ready to implode.

Indeed. If we take AMD as a comparison point, it was essentially in 2012 when today's team was formed there. Intel currently is where AMD was in 2013. Of course I hope Intel is faster at turning around its business, and Intel has to try to salvage much more of its existing business. But AMD's case shows that this is a marathon, not a sprint.

 

[jpiniero]

[Intel-gfx] [PATCH 0/8] i915: Add "standalone media" support for MTL

Meteor Lake has apparently separated the media engine from the GPU at the hardware level. Maybe eventually it'll be it's own tile?

 

[Exist50]

Then what else can they offer in 2023?

The only thing else would be to try to refresh Raptor Lake. Yeah...

Yeah, but that's gonna make the shareholders really mad. Margins margins margins.

It's not like I know what Intel's paying for all these different wafers, but I'd expect for comparable CPU performance, Meteor Lake will be significantly more expensive to produce than Raptor Lake, given the use of Intel 4 and TSMC N6, plus Foveros. The only saving grace is they could probably leave the GPU off entirely for some SKUs.

@Exist50 Going to 20A from N3 would be the 180 degree change under Gelsinger to not abandon fabbing chips internally but gain leadership and strengthen it.

I'd expect that if they do use both, the N3 products would come first. I think they'll definitely try to avoid straight up dual sourcing (because that would either make 20A look only as good as N3 or great weird performance discrepancies), so that leaves either different product lines, a quick 20A refresh, or both.

 

[Exist50]

For socket compatibility all IO must be the same throughout the lineup, but this suggests that 6+8 will get less IO than 8+16? kinda strange.
Maybe this IO tile is just a proposal

That slide is pretty vague. Also if you look at it, the second and third mockups have a compute tile of the same area. It definitely hints at a slightly smaller SoC, larger I/O, and larger iGPU for the third pic. Whatever that means!

I don't think they meant for that slide to be taken so literally. Think the general idea is just that they have the ability to scale different components (CPU, GPU, IO) depending on the market.

 

[Exist50]

Yeah... but Intel says a lot of things. A lot of things that turn out not to be anywhere near true. Hell, check out these two articles just on 7 nm.

You know what? I completely agree. Intel's fab efforts have been an unmitigated failure for the better part of the last decade. No one can be blamed for doubting that this time™ things will be different, especially after 7nm/Intel 4 has already been delayed once, and perhaps de facto by a bit more. I do have reason to be optimistic about the direction they are heading, but nothing about that is certain until the products ship.

Funny you say that. I am expecting big time layoffs at Intel by the end of the year. It is only a prediction at this point but the server business looks about ready to implode.

I think that will be a good test of Gelsinger as CEO. Because let's be real; any of the last three CEOs would be perfectly happy to start laying people off in the current circumstances. Gelsinger has been willing to put his money where his month is in terms of manufacturing, even if it's spooking investors, but I want to see him demonstrate the same backbone when it comes to design.

Indeed. If we take AMD as a comparison point, it was essentially in 2012 when today's team was formed there. Intel currently is where AMD was in 2013. Of course I hope Intel is faster at turning around its business, and Intel has to try to salvage much more of its existing business. But AMD's case shows that this is a marathon, not a sprint.

I agree with the sentiment, but probably not the timeline. Things are obviously bad now, but they weren't any better a year or two ago. I'm hoping that Granite Rapids is at least a "Naples" moment, and we see more steady improvement from there.

 

[IntelUser2000]

That's all well and good, but we've yet to see the actual fruit of these changes. Admittedly it will take time. But the real issue isn't just who Gelsinger hired but what it is that these new hires and rehires have to work with now that people like Murthy are gone.

Keep in mind that the year Kraznich became CEO was the year Haswell was introduced. So it took almost 2 years with the 14nm delay bad style of management took effect. It's way easier to destroy than rebuild and it still took 2 years!

Meteorlake is the timeframe when the new management takes effect. Pretty much everything now is BK and also Swan. Though I read many times they brought Swan in to stabilize before bringing in Gelsinger for the real deal.

I think that will be a good test of Gelsinger as CEO. Because let's be real; any of the last three CEOs would be perfectly happy to start laying people off in the current circumstances. Gelsinger has been willing to put his money where his month is in terms of manufacturing, even if it's spooking investors, but I want to see him demonstrate the same backbone when it comes to design.

The last few years have shook the company beyond normalcy. So doubts exist, even for me. Only time will tell.

I actually have a lot of faith in Gelsinger's ability.

It's one of the reasons I believe 2020-2025 will play out a lot like 2007-2012. And if you look at that history closely - the parallel for the first two years is strikingly uncanny.

I could be wrong, but if Intel sticks on its roadmap more or less, I won't be. AMD has to stay one node ahead of intel to maintain a performance lead, and they won't be able to do that if Intel can stay near its current roadmap.

I do too. And my expectations are higher. Back then we thought PCs were all they had to worry about but that worry was nothing. This constant cyclical nature of two x86 vendors faltering every few years has to stop. They both should have had mobile chips going back then. I am hoping they can get it good enough to get into mobile organically - vendors will use it simply because it fits and it's good. Even if that means x86 devices going into smaller and more power efficient devices like the trend with mini gaming devices such as from GPD, and not into Android immediately.

 

[DrMrLordX]

I don't think they meant for that slide to be taken so literally. Think the general idea is just that they have the ability to scale different components (CPU, GPU, IO) depending on the market.

Perhaps so! It just seems interesting that the center package mirrors the 6+8 die we've all seen so many times already. So at least that item has some basis in reality.

Though I read many times they brought Swan in to stabilize before bringing in Gelsinger for the real deal.

Well again, BK was just the CEO, the people under him (and the ones he fired by no longer contributing to the company) had the real impact on Intel's output. It's good that you mentioned Bob Swan since I think a lot of people forget he was there for as long as he was, or at all. His ascension contrasted with Murthy's fall when people were expecting Murthy to assume power. A lot of what we're seeing now are probably decisions made by people post-Murthy and concurrent with Swan's tenure as CEO. Plus whatever quick fixes have been made to temporarily right the ship.

If people are expecting miracles from Gelsinger's people, they'll probably come with 20a or not at all.

 

[Exist50]

Meteorlake is the timeframe when the new management takes effect. Pretty much everything now is BK and also Swan.

Way I've heard things, not much has fundamentally changed about Intel's roadmap since Keller left. Delays, process changes, derivatives, etc., but nothing really grounds-up. I think it'll take till 2025 at minimum before we start seeing those.

Though I read many times they brought Swan in to stabilize before bringing in Gelsinger for the real deal.

IIRC, the rumor at the time was that the board asked several people, but each one shot them down, and Swan was left to occupy the position by default. I think that makes more sense, tbh.

The last few years have shook the company beyond normalcy. So doubts exist, even for me. Only time will tell.

I'm not yet convinced. When reality starts to set in during circumstances like this, the first instinct of many investors is to double down on cost cutting. Gelsinger will undoubtably be facing pressure to cut more than he already has, and it remains to be seen to what degree he's willing to hold the line. Though on the plus side, I recently heard that the Optane engineers (or at least many of them) have been redistributed throughout the company, so at least that wasn't an excuse for layoffs.

 

[RTX]

Suerte De Capote: Intel's Cannonlake Leaked In A Patent (NASDAQ:INTC)

Intel filed a patent application which reveals their Cannonlake processor design. The recent addition of Coffee Lake is telling of the advantages.

seekingalpha.com

Filed in 2014 with ideas for large/small cores and multiple pieces of silicon and it took Intel nearly 10 years to make this work in the form of 14th gen

 

[jpiniero]

Filed in 2014 with ideas for large/small cores and multiple pieces of silicon and it took Intel nearly 10 years to make this work in the form of 14th gen

Not entirely since the base die is still passive.

 

[IntelUser2000]

I'm not yet convinced. When reality starts to set in during circumstances like this, the first instinct of many investors is to double down on cost cutting. Gelsinger will undoubtably be facing pressure to cut more than he already has, and it remains to be seen to what degree he's willing to hold the line. Though on the plus side, I recently heard that the Optane engineers (or at least many of them) have been redistributed throughout the company, so at least that wasn't an excuse for layoffs.

And that should have been the normalcy for all companies. The founders of the company HP was known for very human focused management. When times got tough rather than resorting to layoffs as with virtually every company, they reduced pay temporarily for everyone for example. There are more than one way of doing things and an obvious better and worse ways.

 

[mikk]

https://twitter.com/x/status/1569219669411733504

15 months from powered on to RTS. Meteor Lake officially powered on in late April 2022, 15 months later it's late July 2023. Considering MTL is Intels first bigger tile generation for client it might require more validation than usual.

 

[witeken]

https://twitter.com/x/status/1569219669411733504

15 months from powered on to RTS. Meteor Lake officially powered on in late April 2022, 15 months later it's late July 2023. Considering MTL is Intels first bigger tile generation for client it might require more validation than usual.

Pat Gelsinger, Q3'21:

In this quarter, it came out of the fab and powered up, and within 30 minutes, with outstanding performance, right where we expected it to be. All told, this is one of the best lead product startups we have seen in recent memory, which speaks to the health of the process.