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.)

AcrosTinus · Nov 18, 2024

I went through the hassle of testing the storage performance of my Raptor Lake system 14900K unleashed vs 7950x3D vs U9 285K.

Conclusion: 14900K@5,8Ghz>7950x3D>>>> U9 285K.

The test was done with 2x900P 280GB in Raid0 as well as a P5800x 400GB using CrystalDiskMark 8.0.4 on Windows 11 Pro 24H2 with virtualization on as well as the core isolation enabled.

The drives were connected if possible to the CPU Lanes using a U.2 adapter card or a m.2 to slim SAS with a redriver (switches were left default). Cable Length is 50cm.

Sequential throughput is mostly the same across all platforms.
Biggest Difference in the Q1T1 RND4K
- 14900K:
  - > 450MB/s Read // 396MB/s Write (P5800x)
  - > 310MB/s Read // 295MB/s Write (2x900P SW Raid0)
- 7950x:
  - > 360MB/s Read // 285MB/s Write (P5800x)
  - > 250MB/s Read // 276MB/s Write (2x900P SW Raid0)
- U9 285K:
  - > 180MB/s Read // 120MB/s Write (P5800x)
  - > 165MB/s Read // 76MB/s Write (2x900P SW Raid0)

Funny result, the IO die on Intel is a shame, even AMD can realize the Optane advantage. Only the U285 had worse load times in all respects.

AMD has somehow managed to craft a lower latency chiplet architecture that has latency so good that you won't be feeling the difference even on Optane.

SAD!

DavidC1 · Nov 18, 2024

When the original Athlon came out it outperformed the Pentium III even in IO operations. History rhymes again.

DZero · Nov 18, 2024

511 said:
ARL-U SKU(RWC+ and CMT+)

https://twitter.com/x/status/1858299099608752597
The frequency increase is more than the node (18%) gains lmao at 12/15W
View attachment 111768

Big question... ARL-U has HT?
If not... damn Intel is speedrunning to end the U tier.

Hulk · Nov 18, 2024

AcrosTinus said:
I went through the hassle of testing the storage performance of my Raptor Lake system 14900K unleashed vs 7950x3D vs U9 285K.

Conclusion: 14900K@5,8Ghz>7950x3D>>>> U9 285K.

The test was done with 2x900P 280GB in Raid0 as well as a P5800x 400GB using CrystalDiskMark 8.0.4 on Windows 11 Pro 24H2 with virtualization on as well as the core isolation enabled.

The drives were connected if possible to the CPU Lanes using a U.2 adapter card or a m.2 to slim SAS with a redriver (switches were left default). Cable Length is 50cm.

Sequential throughput is mostly the same across all platforms.

Biggest Difference in the Q1T1 RND4K

14900K:

> 450MB/s Read // 396MB/s Write (P5800x)

> 310MB/s Read // 295MB/s Write (2x900P SW Raid0)

7950x:

> 360MB/s Read // 285MB/s Write (P5800x)

> 250MB/s Read // 276MB/s Write (2x900P SW Raid0)

U9 285K:

> 180MB/s Read // 120MB/s Write (P5800x)

> 165MB/s Read // 76MB/s Write (2x900P SW Raid0)

Funny result, the IO die on Intel is a shame, even AMD can realize the Optane advantage. Only the U285 had worse load times in all respects.

AMD has somehow managed to craft a lower latency chiplet architecture that has latency so good that you won't be feeling the difference even on Optane.

SAD!

Do you have a 285K rig? If so can you run our Handbrake benchmark?

DavidC1 · Nov 18, 2024

DZero said:
Big question... ARL-U has HT?
If not... damn Intel is speedrunning to end the U tier.

Very likely. Lion Cove doesn't have HT at all. ARL-U uses Redwood Cove core.

The things exist50 says is amusing. He's basically completely discrediting Intel's claim that MoP on Lunarlake is to save cost, because ultimately a worst product will cancel out any lower costs.

There is some merit to this. My laptop with Kabylake-Y and LPDDR3 memory doesn't go below 0.6W. On Lunarlake I saw it low as 0.2W. Now, MoP isn't enough, you have to optimize for it. But in Lunarlake it does. Lunarlake's MoP is offering anywhere from 10-15% battery life increase by itself.

So Intel was focusing too much on margins that led them to this mess today, and because of that mess and resulting low revenue they are focusing on margins even more.

511 · Nov 19, 2024

AcrosTinus said:
I went through the hassle of testing the storage performance of my Raptor Lake system 14900K unleashed vs 7950x3D vs U9 285K.

Conclusion: 14900K@5,8Ghz>7950x3D>>>> U9 285K.

The test was done with 2x900P 280GB in Raid0 as well as a P5800x 400GB using CrystalDiskMark 8.0.4 on Windows 11 Pro 24H2 with virtualization on as well as the core isolation enabled.

The drives were connected if possible to the CPU Lanes using a U.2 adapter card or a m.2 to slim SAS with a redriver (switches were left default). Cable Length is 50cm.

Sequential throughput is mostly the same across all platforms.

Biggest Difference in the Q1T1 RND4K

14900K:

> 450MB/s Read // 396MB/s Write (P5800x)

> 310MB/s Read // 295MB/s Write (2x900P SW Raid0)

7950x:

> 360MB/s Read // 285MB/s Write (P5800x)

> 250MB/s Read // 276MB/s Write (2x900P SW Raid0)

U9 285K:

> 180MB/s Read // 120MB/s Write (P5800x)

> 165MB/s Read // 76MB/s Write (2x900P SW Raid0)

Funny result, the IO die on Intel is a shame, even AMD can realize the Optane advantage. Only the U285 had worse load times in all respects.

AMD has somehow managed to craft a lower latency chiplet architecture that has latency so good that you won't be feeling the difference even on Optane.

SAD!

Also Intel has accidentally crafted 13/14th gen that is giving everyone trouble in area like these it's 10th -> 11th gen fiasco all over again 🤣

511 · Nov 19, 2024

DavidC1 said:
Very likely. Lion Cove doesn't have HT at all. ARL-U uses Redwood Cove core.

The things exist50 says is amusing. He's basically completely discrediting Intel's claim that MoP on Lunarlake is to save cost, because ultimately a worst product will cancel out any lower costs.

If it was Intel's Fab they would have saved the cost

DavidC1 said:
There is some merit to this. My laptop with Kabylake-Y and LPDDR3 memory doesn't go below 0.6W. On Lunarlake I saw it low as 0.2W. Now, MoP isn't enough, you have to optimize for it. But in Lunarlake it does. Lunarlake's MoP is offering anywhere from 10-15% battery life increase by itself.

So Intel was focusing too much on margins that led them to this mess today, and because of that mess and resulting low revenue they are focusing on margins even more.

As for this they specifically said they optimised the PHY for low power and it is showing PTL is going to regress how much depends on whether PMICs are used or not cause MOP ks not coming

majord · Nov 19, 2024

DavidC1 said:
ARL-U uses Redwood Cove core.

Why is this?

SteinFG · Nov 19, 2024

majord said:
Why is this?

Because it's a meteor lake update, and because they want mass-market chips to be manufactured on their fabs (Redwood cove is intel 4 / intel 3 nm core)

majord · Nov 19, 2024

thanks. Sorry i'm a bit behind on the full notebook stack , Seems strange to use the Arrowlike nomenclature without LNC/ . So these are sort of budget, mass-market chips? are there no lion cove /Skymont based higher core count U parts? I'm just trying to visualise what the halo parts are in various TDP classes. Obviously Luna lake there for very low core count, low power, then it seems like a performance gap between that and the LNC/SKY based H / HX parts?

Kocicak · Nov 19, 2024

AcrosTinus said:
Conclusion: 14900K@5,8Ghz>7950x3D>>>> U9 285K.

Sequential throughput is mostly the same across all platforms.

Biggest Difference in the Q1T1 RND4K

14900K:

> 450MB/s Read // 396MB/s Write (P5800x)

> 310MB/s Read // 295MB/s Write (2x900P SW Raid0)

7950x:

> 360MB/s Read // 285MB/s Write (P5800x)

> 250MB/s Read // 276MB/s Write (2x900P SW Raid0)

U9 285K:

> 180MB/s Read // 120MB/s Write (P5800x)

> 165MB/s Read // 76MB/s Write (2x900P SW Raid0)

Interesting results, do you have the complete result screenshots? Would anything change, if you used a "leashed" 14900K?

coercitiv · Nov 19, 2024

majord said:
thanks. Sorry i'm a bit behind on the full notebook stack , Seems strange to use the Arrowlike nomenclature without LNC/ . So these are sort of budget, mass-market chips? are there no lion cove /Skymont based higher core count U parts? I'm just trying to visualise what the halo parts are in various TDP classes. Obviously Luna lake there for very low core count, low power, then it seems like a performance gap between that and the LNC/SKY based H / HX parts?

The halo part for the U class is Lunar Lake. It might struggle against ARL-U @ 28W in "Cinebench", but should be superior in all aspects that really matter, except cost ofc.

SteinFG · Nov 19, 2024

majord said:
Seems strange to use the Arrowlike nomenclature without LNC

Intel is being shady. It is an small upgrade over Meteor lake, cause ARL-U chips have higher boost clock, but that's pretty much it

majord said:
So these are sort of budget, mass-market chips?

Yeah, 2P+8E for mainstream.

majord said:
are there no lion cove /Skymont based higher core count U parts?

LNL (lunar lake) is 4P+4E, tdp is configurable from 8 to 30W. It's a halo part in its TDP range.
ARL-H is 6P+8E. most likely 30 to 65W tdp (following MTL-H tdp numbers). It's a halo part in its TDP range.

Both lunar lake and arrow lake-H have new cores (lion cove and skymont). And with such core counts and TDP ranges, there is no need for something in-between.

SteinFG · Nov 19, 2024

coercitiv said:
The halo part for the U class is Lunar Lake.

The halo part for low-power class is Lunar Lake V series.
Sorry, had to nitpick a bit 😛 though I totally get what you're saying

511 · Nov 19, 2024

coercitiv said:
The halo part for the U class is Lunar Lake. It might struggle against ARL-U @ 28W in "Cinebench", but should be superior in all aspects that really matter, except cost ofc.

ARL-U is way inexpensive compared to LNL that it is hilarious and regarding battery life MTL-U/ARL-U have better battery life than H comparison not good as lnl but very good and for cheap
96mm2 N6 SOC
10mm2 IO
40mm2 CPU on Intel 3
24mm2 GPU Die

coercitiv · Nov 19, 2024

SteinFG said:
The halo part for low-power class is Lunar Lake V series.
Sorry, had to nitpick a bit 😛 though I totally get what you're saying

I stand by my correct use of "U class" according to certified British classification:

Mobile low power - U Class
Mobile performance - H Class
Desktop low power - T class
Desktop performance - S class

On a more serious note, while I agree LNL is supposed to be the halo for a "low-power class" mobile chips, I disagree with the idea that "low-power" goes up to 30W and therefore I don't consider U series to be "low-power" CPUs. That's why I used the rather awkward "U class" term.

511 · Nov 19, 2024

I guess this is why we see low supply for ARL U9

https://twitter.com/x/status/1858816934583955719

dttprofessor · Nov 19, 2024

511 said:
ARL-U is way inexpensive compared to LNL that it is hilarious and regarding battery life MTL-U/ARL-U have better battery life than H comparison not good as lnl but very good and for cheap
96mm2 N6 SOC
10mm2 IO
40mm2 CPU on Intel 3
24mm2 GPU Die

It‘s not good compare to 8845,but not bad for U15.

511 · Nov 19, 2024

dttprofessor said:
It‘s not good compare to 8845,but not bad for U15.

It is not good as performance comparison but the cost is less

cannedlake240 · Nov 19, 2024

511 said:
It is not good as performance comparison but the cost is less

Doubt it's less expensive than hawk point, TSMC keeps raising prices and theres Foveros overhead as well

511 · Nov 19, 2024

cannedlake240 said:
Doubt it's less expensive than hawk point, TSMC keeps raising prices and theres Foveros overhead as well

You know that tsmc raised Price on 5/3 nm class node N6 was low utilisation vs 5/3 nm and the only N5 die is 23mm2 which is getting raised which will not affect it much and the price raise is customer dependent we all know for sure who is getting the price raised the most everyone will not get the same increase the highest leak was 10% 🙂

AcrosTinus · Nov 19, 2024

Kocicak said:
Interesting results, do you have the complete result screenshots? Would anything change, if you used a "leashed" 14900K?

Yeah, I can attach some later.
The Random 4K reads go down by about 15MB/s, so not much change.

AcrosTinus · Nov 19, 2024

Hulk said:
Do you have a 285K rig? If so can you run our Handbrake benchmark?

No, it is not my own rig, a friend wanted to build a new machine (he is still on a 6700K,1070Ti) and I convinced him to buy Arrow Lake, what a good friend I am I know. While assembling the machine I ran some benches.

511 · Nov 19, 2024

AcrosTinus said:
No, it is not my own rig, a friend wanted to build a new machine (he is still on a 6700K,1070Ti) and I convinced him to buy Arrow Lake, what a good friend I am I know. While assembling the machine I ran some benches.

Lol should have at least bought him 265K and put the money in 4090 or saved the money 🤣

igor_kavinski · Nov 19, 2024

AcrosTinus said:
No, it is not my own rig, a friend wanted to build a new machine (he is still on a 6700K,1070Ti) and I convinced him to buy Arrow Lake, what a good friend I am I know. While assembling the machine I ran some benches.

No, I fully support you being a GOOD friend because coming from a 6700K, that is INDEED a huge upgrade for him.

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

Senior member

Attachments

Member

Golden Member

Senior member

Diamond Member

Golden Member

Golden Member

Golden Member

Senior member

Senior member

Senior member

Golden Member

Diamond Member

Senior member

Senior member

Golden Member

Diamond Member

Golden Member

Member

Golden Member

Senior member

Golden Member

Member

Member

Golden Member

Lifer