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

Khato · Tuesday at 1:55 PM

Hitman928 said:
The point is that in the real world, Intel software doesn’t make up such a large portion of used software

This is an important point all too often ignored in server CPU discussions in particular. General benchmark suites like those run by Phoronix are only going to be applicable to some server use cases. Servethehome does a bit better in that respect, but still doesn't provide a complete picture.

My guess is that it's safe to say that AMD is going to retain a lead on whatever HPC/float compute workloads which remain in the CPU realm. AMD will potentially remain a better choice for general purpose high ST performance servers as well, but not the clear lead they once had. Meanwhile Intel will be the processor of choice in any scenario where their accelerators come into play or where there's cause to disable SMT.

hemedans · Tuesday at 1:56 PM

FlameTail said:
That is truly remarkable. This is as good as Snapdragon X Elite, if not better?

Can you compare A720, X4, Apple P and E with this new Skymont score?

FlameTail · Tuesday at 2:06 PM

hemedans said:
Can you compare A720, X4, Apple P and E with this new Skymont score?

Sure, but it will take some time to assemble the data, and it will have to come from multiple sources.
Meanwhile, here's Notebookcheck's efficiency data;

Lunar Lake is big improvement over Meteor Lake, but still a quite behind Qualcomm/Apple in single-core efficiency; bit closer in multi-core.

511 · Tuesday at 2:12 PM

Khato said:
This is an important point all too often ignored in server CPU discussions in particular. General benchmark suites like those run by Phoronix are only going to be applicable to some server use cases. Servethehome does a bit better in that respect, but still doesn't provide a complete picture.

This is valid for everything
can't believe i am saying this The current thing is AI which require good SW/HW Intel will dominate due to their software in that and AMD will loose to it even Zen5

Khato said:
My guess is that it's safe to say that AMD is going to retain a lead on whatever HPC/float compute workloads which remain in the CPU realm. AMD will potentially remain a better choice for general purpose high ST performance servers as well, but not the clear lead they once had. Meanwhile Intel will be the processor of choice in any scenario where their accelerators come into play or where there's cause to disable SMT.

I agree but the lead will be totally workload dependent now instead of outright domination
customer have to be very picky now when choosing between Intel and AMD cause they will need to understand geometry of their workloads even more cause GNR has Memory Bandwidth advantage AMD doesn't have Multiplexer DIMM AMD will have superior Per core performance
We will se on 10 October but i can comfortably say AMD will not get easy wins anymore goes to show how important Fabrication Process is lol

Abwx · Tuesday at 2:21 PM

FlameTail said:
Sure, but it will take some time to assemble the data, and it will have to come from multiple sources.
Meanwhile, here's Notebookcheck's efficiency data;
View attachment 108166
Lunar Lake is big improvement over Meteor Lake, but still a quite behind Qualcomm/Apple in single-core efficiency; bit closer in multi-core.

Those numbers are meaningless because that s the whole laptop power measured at the main, for CPU efficency they did more accurate measurements, at least for MT.

And no real iprovement vs MTL at 28W, only in the 15-20W range, and against
the 10C/20T 365 that s just a lost cause at whatever power.

511 · Tuesday at 2:25 PM

Embarassing Xeon Benchmark some firmware/kernel issue

cebri1 · Tuesday at 2:39 PM

511 said:
Embarassing Xeon Benchmark some firmware/kernel issue
View attachment 108169

If it’s a bug why is it embarrassing? Do you know there is a specific thread for server chips?

It’s a zero value post. If it’s slower than SPR there is clearly a testing issue.

511 · Tuesday at 2:40 PM

cebri1 said:
If it’s a bug why is it embarrassing? Do you know there is a specific thread for server chips?

The embarassing part is it made it to launch and NAMD is one of the popular HPC benchmark

cebri1 · Tuesday at 2:45 PM

511 said:
The embarassing part is it made it to launch and NAMD is one of the popular HPC benchmark

How do you know is an issue with the firmware and not with the software stack? Have you tested it yourself?

hemedans · Tuesday at 3:15 PM

FlameTail said:
Sure, but it will take some time to assemble the data, and it will have to come from multiple sources.
Meanwhile, here's Notebookcheck's efficiency data;
View attachment 108166
Lunar Lake is big improvement over Meteor Lake, but still a quite behind Qualcomm/Apple in single-core efficiency; bit closer in multi-core.

Well thanx, but I meant skymont, not LNC

Huang has Skymont data but I couldn't understand all of it. From this article

月光下的新探索：Lunar Lake CPU (Lion Cove / Skymont) 微架构测试 | David Huang's Blog

511 · Tuesday at 3:25 PM

cebri1 said:
How do you know is an issue with the firmware and not with the software stack? Have you tested it yourself?

Phoronix said it in their review

AMDK11 · Tuesday at 3:27 PM

The measurements reveal data not disclosed by Intel:

Re-Order Buffer / Retire Queue:

LionCove(ArrowLake) - 576
LionCove(LunarLake) - 557
RaptorCove - 512
Zen5 - 446
Skymont - 411
Gracemont - 253

Scalar Register File:

LionCove(LunarLake) - 268
Skymont(LunarLake) - 250
RaptorCove - 242
Zen5 - 198
Gracemont - 193

Scalar Load Queue:

Zen5 - 202
RaptorCove - 192
LionCove(LunarLake) - 184
Skymont - 114
Gracemont - 80

Scalar Store Queue:

LionCove(LunarLake) - 120
RaptorCove - 112
Zen5 - 104
Skymont - 56
Gracemont - 48

256-bit Vector Register File:

Zen5 - 392
LionCove(LunarLake) - 379
RaptorCove - 307
Skymont - 131
Gracemont - 94

Scalar Add Scheduler:

LionCove(LunarLake) - 107
RaptorCove - 97
Zen5 - 86

128-bit Vector Add Scheduler:

Zen5 - 163
LionCove(LunarLake) - 158
RaptorCove - 101

I have a feeling that LunarLake received a stripped down variant of LionCove.

controlflow · Tuesday at 3:28 PM

If Microsoft hadn't pushed for this useless copilot nonsense, LNL could have reallocated the huge amount of die area eaten up by the NPU to support another 4-8 Skymont cores. LNL as a 4+12 or 6+8 configuration with the same die area minus the NPU could have been pretty great.

poke01 · Tuesday at 3:46 PM

ikjadoon said:
CB23 1T tests, after SemiAccurate's claims (confirmed by ASUS to KitGuru)

Seems to be smaller than ASUS predicted. Perhaps some mitigation? The update itself also failed to update a few times, but finally went through:

CB23 1T score on the 258V
1859 (before the "bad" Windows update)
1819 (after the "bad" Windows Update)

What does this update do and does it effect other tests like CB 2024?

cebri1 · Tuesday at 3:59 PM

511 said:
Phoronix said it in their review

No, he didn’t.

Presumably some issue at the kernel or firmware/microcode level in these select cases or perhaps where the compiler is going awry with Granite Rapids.

KompuKare · Tuesday at 4:09 PM

controlflow said:
If Microsoft hadn't pushed for this useless copilot nonsense, LNL could have reallocated the huge amount of die area eaten up by the NPU to support another 4-8 Skymont cores. LNL as a 4+12 or 6+8 configuration with the same die area minus the NPU could have been pretty great.

But surely the exact same thing can be said about SnapDragon X and Strix too since both also invest huge areas in the NPU which would have been better invested in CPU or GPU cores or cache etc.

However, Microsoft seems to be calling the shots while the other dance!

controlflow · Tuesday at 4:17 PM

KompuKare said:
But surely the exact same thing can be said about SnapDragon X and Strix too since both also invest huge areas in the NPU which would have been better invested in CPU or GPU cores or cache etc.

However, Microsoft seems to be calling the shots while the other dance!

For sure, I think the entire PC ecosystem would be better off if the die space was re-allocated into more CPU cores or larger iGPUs.

I could certainly be wrong but so far I'm very skeptical in Microsoft's ability to come up with truly compelling AI use cases that run on the NPU. A lot of their use cases so far still require an internet connection and they will still police what the user can and cannot do on their own NPU. The whole thing just seems like a way for Microsoft to save on costs by pushing processing onto the users.

The Hardcard · Tuesday at 4:17 PM

controlflow said:
If Microsoft hadn't pushed for this useless copilot nonsense, LNL could have reallocated the huge amount of die area eaten up by the NPU to support another 4-8 Skymont cores. LNL as a 4+12 or 6+8 configuration with the same die area minus the NPU could have been pretty great.

It didn’t have to be either/or. Why wouldn’t they just make 2 tiers? Were the LNL team on a short leash? I’m not seeing any technical or market reasons why there couldn’t be a larger die version.

poke01 · Tuesday at 4:26 PM

Intel seems to limit lunar lake ST performance on battery by a large amount

controlflow · Tuesday at 4:29 PM

poke01 said:
View attachment 108171
Intel seems to limit lunar lake ST performance on battery by a large amount

Seems the hit on battery happens if you use power saver or balanced? Performance doesn't seem to drop much at all on performance profile.

FlameTail · Tuesday at 4:30 PM

poke01 said:
View attachment 108171
Intel seems to limit lunar lake ST performance on battery by a large amount

Intel or the OEMs?
Why is the Snapdragon laptop at the top also dropping ST performance on battery?

poke01 · Tuesday at 4:35 PM

FlameTail said:
Intel or the OEMs?
Why is the Snapdragon laptop at the top also dropping ST performance on battery?

OEMs and vendor?

Josh128 · Tuesday at 4:37 PM

poke01 said:
I was right. 15 watts @ 5.1 GHz
View attachment 108147

Where do you see 5.1 GHz??

poke01 · Tuesday at 4:43 PM

Josh128 said:
Where do you see 5.1 GHz??

gdansk · Tuesday at 5:05 PM

Ah, I'm an idiot. I just now went to the Anandtech homepage looking for Lunar Lake coverage. Forgot about that, oops.

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

Senior member

Attachments

Golden Member

Senior member

Diamond Member

Senior member

Lifer

Attachments

Senior member

Senior member

Senior member

Senior member

Senior member

Senior member

Senior member

Member

Platinum Member

Senior member

Golden Member

Member

Senior member

Platinum Member

Member

Diamond Member

Platinum Member

Senior member

Platinum Member

Platinum Member