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.)

Abwx · Nov 24, 2023

....................

tamz_msc · Nov 24, 2023

TESKATLIPOKA said:
NBC uses a 25 cycle loop for Cinebench R15. That's good for getting rid of the short burst, what is nonsensical is that they use the first result in their chart.

The longer you run crap like Cinebench on laptops with Tiger Lake, or later CPUs, the more the power consumption figures as reported in graphs and charts deviate from the actual average package power.

Abwx · Nov 24, 2023

tamz_msc said:
The longer you run crap like Cinebench on laptops with Tiger Lake, or later CPUs, the more the power consumption figures as reported in graphs and charts deviate from the actual average package power.

They also make measurements at the main with an external monitor, so far the numbers are accurate in respect of the package powers.

Beside, if power was to vary it would show in the scores, you are that much left out of bad faith arguments that you dont even realise that you re contradicting yourself, we re talking of CPU that output the same score over 10 runs, prove that power is totaly stable....

Anyway a 13700H@45W perform like a 7840U@24W, now do your calculations on this basis to evaluate the necessary improvements for MTL to match PHX1.

tamz_msc · Nov 24, 2023

Abwx said:
Lol, you negate the numbers once they dont suit your intel bias, live with it, numbers can be checked, it s just that you are disgruntled by your favourite brand lagging AMD.

NBC put all runs including the first one, we can see the perf at peak power and then at throttled power, here we have a 45W 13700H wich is not throttled; it score 1800 pts from the first to the last run :

https://www.notebookcheck.net/Schenker-Vision-16-2023-Review-16-inch-lightweight-with-many-ports.738862.0.html

And a 7840U set at 30/24W, at 30W it score 1930 pts and at 24W the score decrease to 1800 pts :

https://www.notebookcheck.net/Lenovo-ThinkPad-T14s-G4-review-Business-laptop-is-better-with-AMD-Zen4.763581.0.html

So much for clulessness, numbers are numbers, and they say that you dont even have the slightest idea about Intel s huge disadvantage currently, that s why you are overestimating MTL s eventual numbers...

Okay, so let me take your own numbers from the two links you provided:

13700H Avg package power = 40 W; Average clock-speed = 2503 MHz; Average temp = 62 degrees.

7840U Avg package power = 24 W; Average clock-speed = 3055 Mhz; Average temp = 69 degrees.

The first one is Intel 7, the second one is TSMC N4.

The first one is 6P+8E, the second one is 8 "P".

So much for "inefficiency", when it is clearly due to Intel cramming much more on a worse node.

tamz_msc · Nov 24, 2023

Abwx said:
They also make measurements at the main with an external monitor, so far the numbers are accurate in respect of the package powers.

Beside, if power was to vary it would show in the scores, you are that much left out of bad faith arguments that you dont even realise that you re contradicting yourself, we re talking of CPU that output the same score over 10 runs, prove that power is totaly stable....

Anyway a 13700H@45W perform like a 7840U@24W, now do your calculations on this basis to evaluate the necessary improvements for MTL to match PHX1.

Tell me you haven't used a modern laptop without telling me you haven't used a modern laptop.

Abwx · Nov 24, 2023

tamz_msc said:
Okay, so let me take your own numbers from the two links you provided:

13700H Avg package power = 40 W; Average clock-speed = 2503 MHz; Average temp = 62 degrees.

7840U Avg package power = 24 W; Average clock-speed = 3055 Mhz; Average temp = 69 degrees.

The first one is Intel 7, the second one is TSMC N4.

The first one is 6P+8E, the second one is 8 "P".

So much for "inefficiency", when it is clearly due to Intel cramming much more on a worse node.

That s not the point, if 13700H@45W = 7840U@24W it means that MTL will have to use a process that is about 45/24 = 1.875x more efficient to match the 7840U, or said otherwise a process that provide the same perf at 0.53x the power.

tamz_msc · Nov 24, 2023

Abwx said:
That s not the point, if 13700H@45W = 7840U@24W it means that MTL will have to use a process that is about 45/24 = 1.875x more efficient to match the 7840U, or said otherwise a process that provide the same perf at 0.53x the power.

Bozo, the 7840U does not consume 24W during the first run. It consumes closer to 30W. So 30/45 = 0.67x. And according to Semiwiki's analysis, Intel 4 is 40% lower power at iso-performance versus Intel 7.

BorisTheBlade82 · Nov 24, 2023

H433x0n said:
Did they really say that?

It was in an interview back in those Raven Ridge and Picasso days where they were still clearly lagging in idle consumption. Of course this was not 100% serious.

BTT Depending on how effectively this works it might be quite a feat, that the competition might have to follow.

FlameTail · Nov 24, 2023

Such a profound comment. Infinite money glitches indeed: first it was the crypto boom, now the AI boom...

Henry swagger · Nov 24, 2023

FlameTail said:
View attachment 89279

Such a profound comment. Infine money glitches indeed: first it was the crypto boom, now the AI boom...

Why are you posting this on here again ? .. are you bored sir lol

FlameTail · Nov 24, 2023

Henry swagger said:
Why are you posting this on here again ? .. are you bored sir lol

Oh sorry.

My bad, wrong thread.

Abwx · Nov 24, 2023

tamz_msc said:
Bozo, the 7840U does not consume 24W during the first run. It consumes closer to 30W. So 30/45 = 0.67x. And according to Semiwiki's analysis, Intel 4 is 40% lower power at iso-performance versus Intel 7.

You should quit tech discussions,...

In the first run it run at 30W and score 1930 pts, in the following runs it run at 24W and score 1800 pts, the same as the 13700H@45W.

Is it possible to be that dense..?...

Geddagod · Nov 24, 2023

I'm sorry, anyone who thinks a 6+8 CPU should be compared to a 10 P-core CPU rather than a 8 P-core CPU doesn't get to call other members "dense" lmao
Also anyone else experiencing slow load times on Anandtech or something? I didn't participate in the recent discussions for a couple days cuz Anandtech was loading pretty slowly for me for some reason...

coercitiv · Nov 24, 2023

Geddagod said:
I'm sorry, anyone who thinks a 6+8 CPU should be compared to a 10 P-core CPU rather than a 8 P-core CPU doesn't get to call other members "dense" lmao

how about calling them "bozo"?

Geddagod said:
Also anyone else experiencing slow load times on Anandtech or something? I didn't participate in the recent discussions for a couple days cuz Anandtech was loading pretty slowly for me for some reason...

you're not alone, barely working in the last few days

FlameTail · Nov 24, 2023

We are calling each other bozos, are we?

FlameTail · Nov 24, 2023

Geddagod said:
Also anyone else experiencing slow load times on Anandtech or something? I didn't participate in the recent discussions for a couple days cuz Anandtech was loading pretty slowly for me for some reason...

I have been too. I was wondering if my ISP was blocking the website or something.

Abwx · Nov 24, 2023

Geddagod said:
I'm sorry, anyone who thinks a 6+8 CPU should be compared to a 10 P-core CPU rather than a 8 P-core CPU doesn't get to call other members "dense" lmao
Also anyone else experiencing slow load times on Anandtech or something? I didn't participate in the recent discussions for a couple days cuz Anandtech was loading pretty slowly for me for some reason...

That s the metric, 2 E cores are supposed to have the same throughput as a a P core + SMT, this is obvious for the 13900K vs 7950X FI.

So a 6 P + 4 E is comparable to a 8C/16T and a 6 P + 8 E to a 10C/20T, it s just that Intel lagging by a node cant sustain those comparisons at same power, we ll see once MTL is released but i doubt that it will hold the comparison even with a 8C/16T given the current numbers and the expected perf/watt improvements.

FTR Intel stated that MTL "will be our most efficent mobile CPU ever", notice that they didnt state "the most effcient mobile CPU ever"..

tamz_msc · Nov 24, 2023

It is totally unreliable to have 100% faith in HWinfo's reporting alone to estimate power/frequency scaling at different performance levels for Intel mobile CPUs beginning with Tiger Lake.

Why?

Because of Intel Dynamic Tuning technology. I have the data to prove it. And as for NBC's testing methodologies, former-AT reviewer Billy Tallis finds them unreliable according to a few comments I exchanged with him on Reddit.

Geddagod · Nov 24, 2023

Abwx said:
That s the metric, 2 E cores are supposed to have the same throughput as a a P core + SMT, this is obvious for the 13900K vs 7950X FI.

So a 6 P + 4 E is comparable to a 8C/16T and a 6 P + 8 E to a 10C/20T, it s just that Intel lagging by a node cant sustain those comparisons at same power, we ll see once MTL is released but i doubt that it will hold the comparison even with a 8C/16T given the current numbers and the expected perf/watt improvements.

FTR Intel stated that MTL "will be our most efficent mobile CPU ever", notice that they didnt state "the most effcient mobile CPU ever"..

Idk how much simpler I can make this for you.
If Intel had no E-cores, a 6+8 CPU would be a 8 P-core CPU. That's how they literally slot into the floorplan of an Intel CPU.

The immense performance scalability when you throw efficiency of the E-cores out the window is a deliberate strength of Intel's heterogenous design, used by Intel to cheapen out on the die area it has to invest for their desktop products.

Abwx · Nov 24, 2023

Geddagod said:
Idk how much simpler I can make this for you.
If Intel had no E-cores, a 6+8 CPU would be a 8 P-core CPU. That's how they literally slot into the floorplan of an Intel CPU.

The immense performance scalability when you throw efficiency of the E-cores out the window is a deliberate strength of Intel's heterogenous design, used by Intel to cheapen out on the die area it has to invest for their desktop products.

That was the point, to have the same throughput with 2 E cores as a 1C/2T P core while using 0.5x the area of said P core.

So it s designed such that perf wise 2 E cores = 1 P core + SMT.

Dunno what cant be understood here, a 6P + 4E is meant to compete with a 8C/16T from AMD, it s just that they lag process wise and that even a 6P + 8E RPL is unable to compete perf/Watt wise with a 8C/16T 7840U.

TESKATLIPOKA · Nov 24, 2023

Geddagod said:
I'm sorry, anyone who thinks a 6+8 CPU should be compared to a 10 P-core CPU rather than a 8 P-core CPU doesn't get to call other members "dense" lmao

Geddagod said:
Idk how much simpler I can make this for you.
If Intel had no E-cores, a 6+8 CPU would be a 8 P-core CPU. That's how they literally slot into the floorplan of an Intel CPU.

The immense performance scalability when you throw efficiency of the E-cores out the window is a deliberate strength of Intel's heterogenous design, used by Intel to cheapen out on the die area it has to invest for their desktop products.

6P+8E RPL performs comparably to imaginary 10P RPL.
I don't understand why you compare It based on die space, what's the point?
It's not like 8C Zen4 is fighting against an 8P CPU, but against a 6P+8E CPU.
So while AMD has a process advantage, Intel has the core advantage.

MTL will be on a better process, so AMD will lose their advantage, but Intel will still keep It at least until Strix Point is out. Then AMD will have the core advantage.

igor_kavinski · Nov 24, 2023

tamz_msc said:
Because of Intel Dynamic Tuning technology. I have the data to prove it.

You got your data with/without Intel DTT?

Geddagod · Nov 24, 2023

Abwx said:
That was the point, to have the same throughput with 2 E cores as a 1C/2T P core while using 0.5x the area of said P core.

So it s designed such that 2 E cores = 1 P core + SMT.

Dunno what cant be understood here, a 6P + 4E is meant to compete with a 8C/16T from AMD, it s just that they lag process wise and that even a 6P + 8E RPL is unable to compete perf/Watt wise with a 8C/16T 7840U.

Again, only when you throw away efficiency to the toilet does 2E cores = 1 P core+SMT.

A 6+8 CPU is more efficient than a 8 P core iso performance. It takes as much space as a 8 P-core. But the neat thing is that it can scale up to a 10P core CPU, when you don't care about perf/watt.
It's incredibly stupid to claim a 6+8 design should be compared to a 10 P core CPU design in theory, since that would be comparing a 8 P core design vs a 10 P core design esentially. You could do the same thing with a 8 regular P core CPU vs a regular 10 P-core CPU btw, just make the 8 P cores clock higher and consume more power, and it can compete with the 10 P-core CPU. The only difference is that the 8 P-core CPU, would obviously consume way more power.

TESKATLIPOKA said:
6P+8E performs comparably to 10P at least at high TDPs.

Exactly. At high TDPs.

TESKATLIPOKA said:
I don't understand why you compare It based on die space, what's the point?

Because that's what an Intel without E-cores would look like. A 6+8 Core CPU would become a 8 core CPU. Not a 10 P-core one.

TESKATLIPOKA said:
It's not like 8C Phoenix is fighting against an 8P CPU, but against a 6P+8E CPU.

Oh. Ye sure.

Abwx · Nov 24, 2023

Geddagod said:
A 6+8 CPU is more efficient than a 8 P core iso performance. It takes as much space as a 8 P-core. But the neat thing is that it can scale up to a 10P core CPU, when you don't care about perf/watt.

It is more efficent the same way as at same throughput a 10C/20T is more efficient than a 8C/20T.

A 6 + 4 has 16 threads, 8 P cores have also 16 threads...

A 6 + 8 has 20 threads, so does a 10C/20T, you didnt notice this point..?..

Intel thought that they could use this trick to have a manufacturing cost advantage, but that only works if they are on par process wise, and at the time these hybrid uarches were designed they didnt thought that they would be that late on the process many years later...

Now even if we take account of the process difference their hybrid uarch is not that efficient overall, to match the 7840U@25W Intel has to improve perf/Watt of a 6 + 8 MTL by 1.8x at same throughput and power than a 6 + 8 RPL.

TESKATLIPOKA · Nov 24, 2023

Geddagod said:
Exactly. At high TDPs.

Then do you have data which shows that the situation is different at low TDPs?

Geddagod said:
Because that's what an Intel without E-cores would look like. A 6+8 Core CPU would become a 8 core CPU. Not a 10 P-core one.

Why should this matter exactly?
There is no written rule that Intel has to keep the same die size.
If Intel wants to keep die size then It would be 8P core CPU, but loose performance.
If they want to keep the performance then It would be a 10P core CPU, but obviously It would be larger.

Geddagod said:
Oh. Ye sure.

What?

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

Senior member

Attachments

Lifer

Diamond Member

Lifer

Diamond Member

Diamond Member

Lifer

Diamond Member

Senior member

Platinum Member

Senior member

Platinum Member

Lifer

Golden Member

Diamond Member

Platinum Member

Platinum Member

Lifer

Diamond Member

Golden Member

Lifer

Platinum Member

Lifer

Golden Member

Lifer

Platinum Member