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.)

DZero · Sunday at 2:54 PM

"Deletes Copilot App"
I do not see any loses on there

TESKATLIPOKA · Sunday at 2:59 PM

511 said:
The area efficiency sucks for battlemage sucks as a dGPU though they improve that as well with celestial and for second part strix halo exits but it is damm expensive vs Panther Lake ......

And what's the price of said Panther Lake?
BTW Strix Halo is still faster than this 12Xe² IGP.

Thibsie · Sunday at 3:25 PM

DZero said:
"Deletes Copilot App"
I do not see any loses on there

Copilot became so smart it figured it was better to delete itself. What a smart litle AI lol

511 · Sunday at 4:29 PM

TESKATLIPOKA said:
And what's the price of said Panther Lake?
BTW Strix Halo is still faster than this 12Xe² IGP.

Yes it is due to fact that it has very high memory bandwidth and a big MALL and a bigger GPU.
For Strix halo pricing it's using advanced packing two 69mm2 Zen5 dies and a big 300mm2 N4P dies vs a 114mm2 18A die and 53mm2 N3E die + 10mm2 N6 IO die and advanced packing (just si you know a N4P wafer is roughly 15K) and a N3 is 19K-20K Wafer. It is not rocket science which is more expensive. N6 is very cheap node

DavidC1 · Monday at 3:14 AM

511 said:
For 12Xe3 we have many Improvements also there is clear PPA Improvement as well vs Xe2 here are all the changes that have been leaked.

Most of the 1.6-2x gains are delivered by the 1.2x clock speed and 50% extra compute units. Yes they will not all scale well, and it's a challenge to do it, plus the memory bandwidth won't increase drastically.

It would be nice to see those gains come with minimal power increase.

Exist was talking about how Celestial's changes are bigger than the ones made with Battlemage. We'll see whether those changes are in terms of improved PPA. For pure performance and per shader Xe2 offered great improvements of 50% per watt and 70% per shader.

511 · Monday at 3:43 AM

DavidC1 said:
Most of the 1.6-2x gains are delivered by the 1.2x clock speed and 50% extra compute units. Yes they will not all scale well, and it's a challenge to do it, plus the memory bandwidth won't increase drastically.

It would be nice to see those gains come with minimal power increase.

Exist was talking about how Celestial's changes are bigger than the ones made with Battlemage. We'll see whether those changes are in terms of improved PPA. For pure performance and per shader Xe2 offered great improvements of 50% per watt and 70% per shader.

There was a node jump as well N3B -> N3E is more like going from N5 to n4 in terms of performance per watt and reverse in terms of density.

poke01 · Monday at 6:56 AM

What peoples thoughts on 18A clocks on PTL? Will it surpass LNL?

511 · Monday at 6:58 AM

poke01 said:
What peoples thoughts on 18A clocks on PTL? Will it surpass LNL?

Definitely lol LNL is a low bar even MTL Matched LNL Clocks the real question will it surpass ARL-H Clocks

poke01 · Monday at 8:02 AM

511 said:
Definitely lol LNL is a low bar even MTL Matched LNL Clocks the real question will it surpass ARL-H Clocks

Why did they choose N3E for the GPU and not 18A?

Cause I think the answer may lie in that decision unless N3E was locked in much earlier

Ghostsonplanets · Monday at 8:13 AM

poke01 said:
Why did they choose N3E for the GPU and not 18A?

Cause I think the answer may lie in that decision unless N3E was locked in much earlier

Ease of development? Mature PDK? Previous knowledge with the toolchain and foundry work? Timeline mismatch with i18A? Etc, etc, etc.

For what's worth, Intel 18A will be used to manufacture 2 Xe³ "tile" and Xe3P dGPUs, whatever these may be.

poke01 · Monday at 8:18 AM

Ghostsonplanets said:
Ease of development? Mature PDK? Previous knowledge with the toolchain and foundry work? Timeline mismatch with i18A? Etc, etc, etc.

For what's worth, Intel 18A will be used to manufacture 2 Xe³ "tile" and Xe3P dGPUs, whatever these may be.

Makes sense. Yeah I have faith in 18A now. Looking forward to Q3/Q4

511 · Monday at 9:21 AM

poke01 said:
Why did they choose N3E for the GPU and not 18A?

Cause I think the answer may lie in that decision unless N3E was locked in much earlier

Not optimized for GPUs yet? Xe IP is mainly on TSMC also as of now so i think those two were the main reason

Glo. · Monday at 9:44 AM

PTL-H with 12 Xe cores is not a competitor for Strix Halo, but for regular Strix/Medusa Point.

511 · Monday at 10:06 AM

Glo. said:
PTL-H with 12 Xe cores is not a competitor for Strix Halo, but for regular Strix/Medusa Point.

it is way Powerful to be compared with Regular Strix also Medusa Point is using RDNA 3.5 don't know how much Compute Units @Kepler_L2 would know more

Kepler_L2 · Monday at 10:21 AM

511 said:
it is way Powerful to be compared with Regular Strix also Medusa Point is using RDNA 3.5 don't know how much Compute Units @Kepler_L2 would know more

Should be 8 WGP still

Ghostsonplanets · Monday at 10:38 AM

511 said:
it is way Powerful to be compared with Regular Strix also Medusa Point is using RDNA 3.5 don't know how much Compute Units @Kepler_L2 would know more

I don't think they'll be way too powerful. But I do expect a performance leap over Strix Point. iGPUs biggest problem is cache and bandwidth and that's something Intel is very generous with.

Strix P 8WGP performs like a GTX 1650 - 1650 Super. So I think Intel can reach GTX 1660 - 1660 Super performance. Assuming Celestial PPA uplift is there.

511 · Monday at 10:45 AM

Kepler_L2 said:
Should be 8 WGP still

Looks like they will loose to PTL-iGPU (That's an idiotic decision done to minimize cost probably)

Ghostsonplanets said:
I don't think they'll be way too powerful. But I do expect a performance leap over Strix Point. iGPUs biggest problem is cache and bandwidth and that's something Intel is very generous with.

Strix P 8WGP performs like a GTX 1650 - 1650 Super. So I think Intel can reach GTX 1660 - 1660 Super performance. Assuming Celestial PPA uplift is there.

Well I was thinking of a 3050-3060 laptop at 40W in that range

TESKATLIPOKA · Monday at 2:54 PM

511 said:
Yes it is due to fact that it has very high memory bandwidth and a big MALL and a bigger GPU.
For Strix halo pricing it's using advanced packing two 69mm2 Zen5 dies and a big 300mm2 N4P dies vs a 114mm2 18A die and 53mm2 N3E die + 10mm2 N6 IO die and advanced packing (just si you know a N4P wafer is roughly 15K) and a N3 is 19K-20K Wafer. It is not rocket science which is more expensive. N6 is very cheap node

So you don't know what will be the price of Panther Lake.

Glo. · Monday at 4:09 PM

511 said:
Looks like they will loose to PTL-iGPU (That's an idiotic decision done to minimize cost probably)

Well I was thinking of a 3050-3060 laptop at 40W in that range
View attachment 120130

With 3050-3060 performance it is not touching the Strix Halo, let alone Medusa Halo.

511 said:
it is way Powerful to be compared with Regular Strix also Medusa Point is using RDNA 3.5 don't know how much Compute Units @Kepler_L2 would know more

Doesn't matter how much more powerful it will be. PTL-H is going to be 128 bit bus, and only 1536 ALU GPU. Yes, it will beat, most likely, Strix Point and Medusa point, but it doesn't change that fact that it is competing in the same, mainstream market.

It is not and never going to be the Halo competition.

dullard · Monday at 5:16 PM

poke01 said:
Why did they choose N3E for the GPU and not 18A?

Cause I think the answer may lie in that decision unless N3E was locked in much earlier

18A will be very limited capacity at the start (they just last week announced the first 18A test wafers out of the Arizona facility--it won't be running at full capacity for quite a while). They have to either have extremely limited quantity of products to sell or produce some parts on other nodes.

adroc_thurston · Monday at 5:40 PM

dullard said:
18A will be very limited capacity at the start (they just last week announced the first 18A test wafers out of the Arizona facility--it won't be running at full capacity for quite a while). They have to either have extremely limited quantity of products to sell or produce some parts on other nodes.

Big iGP SKUs will command a miniature slice of the total PTL volume.
The answer is N3e rocks and 18A ughhh. Yeah.

poke01 · Monday at 6:09 PM

Yeah that is what is strange to me. Compute the biggest title is on 18A would indicate good yields.

I wonder how early the EEP is this year, it being in September would show confidence.

Ghostsonplanets · Monday at 6:38 PM

adroc_thurston said:
The answer is N3e rocks and 18A ughhh. Yeah.

The most likely answer is that there was a timeline mismatch between Xe³ IP and Intel 18A.

The highest volume GPU tile of Panther Lake will be fabbed on Intel 3 after all.

And Wildcat Lake, a later project than PTL, will see the Xe³ IP fabbed on i18A.

adroc_thurston · Monday at 6:41 PM

Ghostsonplanets said:
The most likely answer is that there was a timeline mismatch between Xe³ IP and Intel 18A.

Nope.
18A non-P just blows for things not CPU.
Intel is very public about it too, 18A-P is their first proper SOC offering (whatever time the charm. 4th?).

Ghostsonplanets said:
The highest volume GPU tile of Panther Lake will be fabbed on Intel 3 after all.

Because it sucks but it doesn't matter that it sucks.
4core config is whatever, you're not buying it for the GPU perf.

Ghostsonplanets · Monday at 8:41 PM

adroc_thurston said:
Nope.
18A non-P just blows for things not CPU.
Intel is very public about it too, 18A-P is their first proper SOC offering (whatever time the charm. 4th?).

I can believe N3E provides better perf/W than i18A for GPU IP. But had the timeline afforded, I think they would very likely have used it for 12 Xe³.

What about Wildcat Lake though? That will offer a Xe³ IP implementation on i18A. Or does it fall in the case of "so small that it doesn't matter"?

(And your mention about i18A-P makes me wonder if Xe3-P is related to this node...)

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

Senior member

Attachments

Senior member

Platinum Member

Golden Member

Golden Member

Golden Member

Golden Member

Diamond Member

Golden Member

Diamond Member

Senior member

Diamond Member

Golden Member

Diamond Member

Golden Member

Senior member

Senior member

Golden Member

Platinum Member

Diamond Member

Elite Member

Diamond Member

Diamond Member

Senior member

Diamond Member

Senior member