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	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.)

SpudLobby · Apr 11, 2024

SiliconFly said:
You can't compare nodes based on their products. Thats foundry 101.

Oh I agree with that, I’m not talking about comparing nodes. I’m talking about comparing products with the reasoned premise they have similar manufacturing technology underpinning the transistors, similar market targets, and even similar area overhead.

With Lunar Lake vs the M3 that’s pretty perfect.

With Lunar Lake vs the M1 or M2, or X Elite, it’s funny, because the advantage is for Lunar Lake on a node basis.

Hitman928 · Apr 11, 2024

SiliconFly said:
You can't compare nodes based on their products. Thats foundry 101.

You can compare nodes based on their products only if both are exact same (even the libraries have to be very similar). Otherwise it makes no sense.

The way I read it, @SpudLobby is saying the opposite. He's saying that now that Intel products will have node parity (the most advanced node available), any shortcoming in the product will be on the design alone and they can't blame a node deficiency as a reason for any shortcomings.

SiliconFly · Apr 11, 2024

Hitman928 said:
The way I read it, @SpudLobby is saying the opposite. He's saying that now that Intel products will have node parity (the most advanced node available), any shortcoming in the product will be on the design alone and they can't blame a node deficiency as a reason for any shortcomings.

Yep. No more blame game. And hopefully, no more shortcomings too. LNC is supposed to fix that.

Like I mentioned a while ago, Q3 2024 is the first time Intel Core is ditching all archaic stuff (like Intel 7/4, RWC, Gracemont, etc), in favor of cutting edge stuff (like 20A, N3B, LNC, Skymont, etc). A lot is riding on LNC. And I think it's gonna deliver after all.

The coming months are gonna be very interesting.

Hans Gruber · Apr 11, 2024

SiliconFly said:
Yep. No more blame game. And hopefully, no more shortcomings too. LNC is supposed to fix that.

Like I mentioned a while ago, Q3 2024 is the first time Intel Core is ditching all archaic stuff (like Intel 7/4, RWC, Gracemont, etc), in favor of cutting edge stuff (like 20A, N3B, LNC, Skymont, etc). A lot is riding on LNC. And I think it's gonna deliver after all.

The coming months are gonna be very interesting.

There is a lot of Intel bashing on this forum. I agree with naysayers about the fake generations that Intel has released in the last few years. Arrow Lake is a legit new node on a new process 20A. Going from 10nm down to 5nm (20A). AMD has a problem with not being on TSMC 3nm for the efficiency gains out of the gate. N4 (4nm) is a good clip better than N5 but at least 10% less efficient than N3 (3nm).

Intel said awhile back they would gain market leadership (beyond node parity) with 18A and they were comparing 18A to TSMC 2nm which is not even in production yet. Intel also said that 18A would have 10% better efficiency than 20A. It would seem that Intel is basing their performance numbers more on a performance boost rather than efficiency gains. TSMC is very power efficient when you get to 3nm but the performance uplift isn't there. I would argue the performance uplift 7nm (TSMC) provided has not been seen on 5nm or 3nm. The efficiency gains are substantial but probably not worth the price premium TSMC has been charging their customers.

Intel said they are dropping hyperthreading from Arrow Lake. They said their efficiency cores would more than offset the performance gains that hyperthreading provides. Implying hyperthreading technology is old and outdated. Since Zen 5 Ryzen processors will first appear on 4nm. Intel and AMD are at node/silicon parity with Zen 5 and Arrow Lake.

DavidC1 · Apr 11, 2024

Ghostsonplanets said:
You need to take these tests with a grain of salt, though. The VideoPlayback one is basically local playback and can be chalked up to the Compute Tile being deactivated while only the Media Engine, Display Unit and LPE cores being active.
....
I do prefer Geekerwan test bench.

Anandtech reviews are in the toilet. That's why we go to OTHER sites for reviews but come to this forum to talk about it.

The lack of quality control on the latest review comparable to $1 Chinese ebay stuff definitely doesn't help.

DAPUNISHER · Apr 11, 2024

Hans Gruber said:
There is a lot of Intel bashing on this forum.

Let me clarify a few things:

If you see someone doing it in an Intel thread, report it.

Taking IHVs to task in threads devoted to the subject is allowed Example: the UE games crashing thread. As is some mild trash talk about the competition in that brand's threads. Example: Comparing here in this very thread the faster boot times on Intel vs AM5. Pointing out how fussy AM5 can be about memory compared to Intel. That kind of stuff. If AMD users don't like reading it, they should stay out of Intel threads, and vice versa. Coming in here to start beef about the statements/claims is not permitted.

Going into a thread about a brand, Example: this very Intel thread, and bashing Intel is forbidden. As is trying to deflect, obfuscate, gaslight, or astroturf in order to derail the thread and/or upset Intel users.

I hope this clears up any confusion concerning the forum policies. Intel users should feel like Intel threads are a safe space for them to discuss the company and their products without getting harangued by haters. Report violation. We are a small moderation team and miss stuff. We can't enforce the rules if we didn't see the violation.

- CPU mod DAPUNISHER

SiliconFly · Apr 11, 2024

Hans Gruber said:
Intel and AMD are at node/silicon parity with Zen 5 and Arrow Lake.

Thats not true anymore. For the first time in many years, Intel actually has the lead with Arrow Lake on advanced nodes 20A & N3B, whereas Zen5 is coming out on N4 first and on N3 only later I think (probably after 6 months? or 1 year? dunno when). That gives Intel the edge.

Tiles, advanced packaging, advanced nodes... they have it all. Afaik, the Core is the sole issue now.

Like I always say, for Intel everything now squarely boils down to the performance of LNC. If it tanks, Intel may not survive, considering many key future designs are also based on LNC and most importantly, how thin they're spread already. I mean it'll still survive in some form, but it won't be the Intel we know.

Hans Gruber said:
Going from 10nm down to 5nm (20A).

20A isn't 5nm. It's not equivalent to TSMC 5nm node.

Ghostsonplanets · Apr 11, 2024

SiliconFly said:
Zen5 is coming out on N4 first and on N3 only later I think

Turin Dense will use N3. Consumer Zen 5 will be fabbed on N4P only.

N3B doesn't improve much over N4P on most metrics, sans density, that we can say the node will make one product shine brighter than the other. Design and execution is key and will be showed with each product.

If Intel fumbles even with the usage of the latest and leading edge node, then it's an IDC issue. There will be no mother foundry to cover-up potential issues. And vice-versa.

Hans Gruber · Apr 11, 2024

SiliconFly said:
Thats not true anymore. For the first time in many years, Intel actually has the lead with Arrow Lake on advanced nodes 20A & N3B, whereas Zen5 is coming out on N4 first and on N3 only later I think (probably after 6 months? or 1 year? dunno when). That gives Intel the edge.

Tiles, advanced packaging, advanced nodes... they have it all. Afaik, the Core is the sole issue now.

Like I always say, for Intel everything now squarely boils down to the performance of LNC. If it tanks, Intel may not survive, considering many key future designs are also based on LNC and most importantly, how thin they're spread already. I mean it'll still survive in some form, but it won't be the Intel we know.

20A isn't 5nm. It's not equivalent to TSMC 5nm node.

I agree with some of what you are saying but I want to wait and see what Arrow Lake does on 20A. In other posts I said that AMD really needed to be on 3nm for the efficiency gains. N4 is called 4nm because it's enhanced 5nm silicon with better efficiency and performance gains. It's still not the best 4nm TSMC silicon but it's better than N5. I have been waiting to see what TDP Intel will use for Arrow Lake. I have read 125w for K series chips.

20A is 5nm silicon but Intel says their silicon is not the same as TSMC silicon. Intel says their 5nm silicon is equivalent to TSMC 3nm. 18A is supposed to be better than 2nm TSMC silicon and I think they call the 18A 1.8nm. Intel has changed their stance on what 20A and 18A are equivalent to over the last several years.

All of the people sitting on the sidelines have been waiting for 20A. A true node shrink with a new CPU architecture and not a fake new process. Alder Lake was the last real new Intel generation with a shrink from 14nm down to 10nm. That's why I expect a big performance gain with Arrow Lake in IPC and efficiency.

Here is an article from late 2021 talking about 20A and it being 5nm.

What Is Intel 20A? The Process-Node Classification Explained | Digital Trends

Intel will change its process-node naming convention starting in 2021. Beyond Intel 7, Intel 4, and Intel 3 comes the angstrom era, starting with Intel 20A.

www.digitaltrends.com

mikk · Apr 11, 2024

Tigerick said:
Where have you heard about PCD??? I just heard from my source as explained above

This is what xino told last month about PTL and I trust him. That's why I asked you about the PCD tile you didn't mention.

Tigerick said:
Don't worry, my source is the one mentioned about 12 XE3 before. Time will tell if above info are correct...

I've heard the same, at least Sunny told PTL offers a 12 Xe core GPU. And I trust him as well.

Hans Gruber · Apr 11, 2024

Ghostsonplanets said:
Turin Dense will use N3. Consumer Zen 5 will be fabbed on N4P only.

N3B doesn't improve much over N4P on most metrics, sans density, that we can say the node will make one product shine brighter than the other. Design and execution is key and will be showed with each product.

If Intel fumbles even with the usage of the latest and leading edge node, then it's an IDC issue. There will be no mother foundry to cover-up potential issues. And vice-versa.

Turin will use N3 from Zen 5 release. That's going to do really well efficiency wise. N4P doesn't exist yet. N4 is significantly behind the performance metrics of N4P. Even Nvidia is staying on the 4nm TSMC silicon for Blackwell. Nvidia will use N4P for their enterprise stuff and use N4P or 4N silicon for the consumer cards. 4N is better performance and efficiency wise than N4.

As the third major enhancement of TSMC’s 5nm family, N4P will deliver an 11% performance boost over the original N5 technology and a 6% boost over N4. Compared to N5, N4P will also deliver a 22% improvement in power efficiency as well as a 6% improvement in transistor density. In addition, N4P lowers process complexity and improves wafer cycle time by reducing the number of masks. N4P demonstrates TSMC’s pursuit and investment in continuous improvement of our process technologies.

It would be nice if TSMC would publish their own performance numbers for all the processes on their nodes in an easy to read chart. Below is the best I have for performance metrics.

TSMC Expands Advanced Technology Leadership with N4P Process|Taiwan Semiconductor Manufacturing Company Limited

Hsinchu, Taiwan, R.O.C., Oct. 26, 2021 - TSMC (TWSE: 2330, NYSE: TSM) today introduced its N4P process, a performance-focused enhancement of the 5-nanometer technology platform.

pr.tsmc.com

Tigerick · Apr 11, 2024

mikk said:
This is what xino told last month about PTL and I trust him. That's why I asked you about the PCD tile you didn't mention.

I've heard the same, at least Sunny told PTL offers a 12 Xe core GPU. And I trust him as well.

Yeah, didn't know Xino leaked before. Never heard about PCD as well.

@DavidC1 Now I believe you can trust my source cause the rest of technical details are not easy to fool people if you think through. And Xino also mentioned three tiles, do you believe now?

Ghostsonplanets · Apr 11, 2024

The fact the Compute Tile is coming back to Intel after the N3B is a good show of confidence towards i18A. So going external for the crown jewel (CPU) will be a one-off thing.

mikk said:
I've heard the same, at least Sunny told PTL offers a 12 Xe core GPU. And I trust him as well.

How many EUs would be 12 Xe? 192 EUs? So 1536 ALUs? Although Battlemage does change the internal structure? So maybe it's actually 2x that.

Hans Gruber said:
N4P doesn't exist yet

It does?

Hans Gruber said:
It would be nice if TSMC would publish their own performance numbers for all the processes on their nodes in an easy to read chart.

They always offer PPA advancements numbers over the past nodes. You can easily find it with a Google search. Even Anandtech themselves has a pretty up to date comparison charts.

SiliconFly · Apr 11, 2024

Hans Gruber said:
20A is 5nm silicon

Ghostsonplanets · Apr 11, 2024

20A is 20A, nothing more. 20A offers RibbonFET and BSPDN, which are far different from current FinFET nodes. Without technical details behind it, there's no point comparing it.

Glo. · Apr 11, 2024

Ghostsonplanets said:
The fact the Compute Tile is coming back to Intel after the N3B is a good show of confidence towards i18A. So going external for the crown jewel (CPU) will be a one-off thing.

How many EUs would be 12 Xe? 192 EUs? So 1536 ALUs? Although Battlemage does change the internal structure? So maybe it's actually 2x that.

It does?

They always offer PPA advancements numbers over the past nodes. You can easily find it with a Google search. Even Anandtech themselves has a pretty up to date comparison charts.

128 ALUs per Xe core.

8 Xe cores - 1024 ALUs.
12 Xe Cores - 1536 ALUs.

Hans Gruber · Apr 11, 2024

Ghostsonplanets said:
20A is 20A, nothing more. 20A offers RibbonFET and BSPDN, which are far different from current FinFET nodes. Without technical details behind it, there's no point comparing it.

There is a point of reference. 10nm does not have the real estate that 5nm has for components and cores. It's an entirely new architecture on an entirely new core. The processor core advantage AMD had will be gone with Arrow Lake. The efficiency advantage is a big question. All the stereotypes about Intel processors being heaters or needing massive voltage to perform could be gone as well.

Nobody is comparing anything here. There are a lot of forum members here who do not realize the significance of 20A. Intel was on 14nm for more than 6 years. Alder Lake was the first 10nm Intel Processor. 20A will not be around for long. 18A is the process that will be around for a couple of years. 20A starts with some cutting edge technology that TSMC does not have.

mikk · Apr 11, 2024

Glo. said:
128 ALUs per Xe core.

8 Xe cores - 1024 ALUs.
12 Xe Cores - 1536 ALUs.

Assuming Xe3 won't change the number of EUs per Xe core or ALUs per Vector engine.

Kepler_L2 · Apr 11, 2024

Ghostsonplanets said:
The fact the Compute Tile is coming back to Intel after the N3B is a good show of confidence towards i18A. So going external for the crown jewel (CPU) will be a one-off thing.

IIRC Nova Lake uses N2 for some CPU tiles.

Ghostsonplanets · Apr 11, 2024

Kepler_L2 said:
IIRC Nova Lake uses N2 for some CPU tiles.

Will Intel keep flip-flopping between TSMC and Intel Foundry? But I guess that make sense as i18A pitch is supposed to be around the same as i3/i4 and Intel themselves have said i18A would still be behind the industry best in some metrics.

SiliconFly · Apr 11, 2024

Hans Gruber said:
10nm does not have the real estate that 5nm has...

20A in not 5nm. Even both preceding nodes have better PPA than TSMC 5nm.

Hans Gruber said:
There are a lot of forum members here who do not realize the significance of 20A

Really? You think lot of people here don't understand? Apparently, you're the one who lacks understanding.

Hans Gruber said:
20A starts with some cutting edge technology that TSMC does not have.

Interesting. They're called GAAFETs & BSPD. And you still say 20A equivalent to TSMC 5nm.

I'll let you in on a very interesting piece of information. Intel does not have a 5nm node.

Philste · Apr 12, 2024

Ghostsonplanets said:
How many EUs would be 12 Xe? 192 EUs? So 1536 ALUs? Although Battlemage does change the internal structure? So maybe it's actually 2x that.

PTL should be Celestial already, and we don't even know the changes of Battlemage officially. So there could be quite some changes about the meaning of 12 Xe Cores until then.

SiliconFly · Apr 12, 2024

Ghostsonplanets said:
Will Intel keep flip-flopping between TSMC and Intel Foundry? But I guess that make sense as

It's all about volume. 20A is not a high volume node. And 18A will need time to ramp up. They have no choice but to rely on TSMC for the foreseeable future.

Ghostsonplanets said:
...Intel themselves have said i18A would still be behind the industry best in some metrics.

Those are rumors. Intel never said that. According to them, when comparing 18A and N2, either of them have no "significant" advantage over one another. According to a slide leaked by Dr. Ian Cutress, 18A is expected to be slightly ahead of N2 in performance (Pat said the same). N2 I think is expected to be on par or slightly ahead of 18A in efficiency. Both are expected to have similar densities (and wafer cost too). In short, the PPA variations are negligible. But yield & volume are entirely different beasts. Only time can tell us, how well they ramp and what kinda volumes they can hit (the road ahead isn't rosy for intel foundry).

SiliconFly · Apr 12, 2024

Philste said:
PTL should be Celestial already...

That would be awesome. But can Intel really pull that off? (considering how many years they've spent with Alchemist.)

hemedans · Apr 12, 2024

Ghostsonplanets said:
Can't recommend enough Geekerwan review:

Lpe perfomance is better than expected, sadly they didn't compare power usage vs E, A520 and M3E cores.

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

Senior member

Attachments

Senior member

Diamond Member

Golden Member

Platinum Member

Senior member

Super Moderator CPU Forum Mod and Elite Member

Golden Member

Senior member

Platinum Member

Diamond Member

Attachments

Platinum Member

Senior member

Senior member

Golden Member

Senior member

Diamond Member

Platinum Member

Diamond Member

Senior member

Senior member

Golden Member

Member

Golden Member

Golden Member

Member