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

[Tigerick]

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.)

 

[majord]

For the 285K.



Previously It was 10212 pts in CB R23@30W, wich is about the same score at same power than a 18 months old 6C/12T 7540U, so they are quite late to the party, dunno why the excitement.

A 2 1/2 year old 8c 6800/6850U (zen 3+) does 10233 CB23 @ 23w

This is a complete dud @ throughput perf and perf/watt , despite half of its 8 cores being the fabled Skymont . Either Lion cove is shockingly bad, or Skymont isn't that great

its ST is strong courtesy of lion cove's brute performance , but at what power remains to be seen

 

[mikk]

A 2 1/2 year old 8c 6800/6850U (zen 3+) does 10233 CB23 @ 23w

This is a complete dud @ throughput perf and perf/watt , despite half of its 8 cores being the fabled Skymont . Either Lion cove is shockingly bad, or Skymont isn't that great

its ST is strong courtesy of lion cove's brute performance , but at what power remains to be seen

8 threads vs 16 threads? In Cinebench is has a big impact.

 

[Philste]

2% faster than RPL in MTL to start with

What are you talking about? There is no "RPL IN MTL" RPL is RPL.

Zen 5 is more than 10% faster than Zen 4,
Zen 5c is the same architecture as Zen 5.

It's not in SpecINT, which is the more important of the two Spec Scores. Also different implementations of the same Arch don't necessarily have the same IPC.

Meteor Lakes LPE-Cores have way lower IPC than the one on ring and ZEN5c in Strix also gets hurt by only 8MB L3 for 8 Cores. So yes, Skymont on Ring in ARL-H should have the same IPC as Strix ZEN5c Cores and Clocks higher. ZEN5c clocks at 3.3GHz on both SKUs, even the LPE-Cores on LNL clock higher than that.

 

[Hulk]

That CB R23 LL score has got to be wrong. Either clocks are horrendously low or that score is for P cores only. Actual score should be around 15,000.

 

[majord]

What are you talking about? There is no "RPL IN MTL" RPL is RPL.

It's not in SpecINT, which is the more important of the two Spec Scores. Also different implementations of the same Arch don't necessarily have the same IPC.

Meteor Lakes LPE-Cores have way lower IPC than the one on ring and ZEN5c in Strix also gets hurt by only 8MB L3 for 8 Cores. So yes, Skymont on Ring in ARL-H should have the same IPC as Strix ZEN5c Cores and Clocks higher. ZEN5c clocks at 3.3GHz on both SKUs, even the LPE-Cores on LNL clock higher than that.

Apologies, I meant RWC on MTL . I though it was already established that Intel compared this specific implementation of RWC (i.e in meteor lake) to Skymont for IPC claims.

Pretty sure the true max clock speed of 5c on strix is 'unexplored' and assumed to be approx 4Ghz , but in any case, both Skymont and 5c are well out of their efficiency at Fmax , so probably moot.

8 threads vs 16 threads? In Cinebench is has a big impact.

About 1.2 x at the most? , doesn't really change much does it. turn SMT off and it still isn't looking flash if those numbers are legit.

 

[Philste]

I though it was already established that Intel compared this specific implementation of RWC (i.e in meteor lake) to Skymont for IPC claims.

Skymont was compared to Raptor Cove, Lion Cove was compared to Redwood Cove.

 

[igor_kavinski]

I see mac devices being used by many SW developers. And we need fast and efficient parallel builds. My 4 yo MBP can last all day running parallel builds and running mostly ST compute heavy tasks. It remains to be seen if Intel will be able to match that.

Is that a niche? Surely 😀

Curious question about that. Are these developers fully married to the Mac ecosystem? Whatever binaries they are generating, can only be used on other Macs, if my understanding is correct? Maybe with some slight modification, their open source projects can work on Linux and other BSD derivative distros. And ok, maybe on Windows too with WSL2. But that's it, right? It's a niche of a niche. As far as I understand, they can't be doing anything with their Macs that sells well enough to non-Apple users. That's why I like the idea of Lunar Lake and Snapdragon Elite X. At least it will bring some happiness to regular Windows users who form the majority of the global personal computing userbase.

 

[ikjadoon]

New 268V GB6 scores today:

Geekbench Search - Geekbench

Highest 1T: 2915 1T, 11448 nT
Highest nT: 2903 1T, 11481 nT

 

[Abwx]

Skymont was compared to Raptor Cove, Lion Cove was compared to Redwood Cove.

SKT score about 1595 pts at 4.6GHz in CB R23, and i would estimate that it score in the SSE4.2 restrained CB R15 something like 245 pts at this same frequency, there s no way it could match RPL.

 

[yuri69]

As far as I understand, they can't be doing anything with their Macs that sells well enough to non-Apple users.

Wut? Every heard of VMs or web?

Nearly every React zoomer runs a Mac. The same applies to your average enterprise Java/Python backend dev.

 

[ikjadoon]

New 228V GB6 scores today:

Geekbench Search - Geekbench

Highest 1T & nT: 2621 1T, 10431 nT

 

[igor_kavinski]

New 268V GB6 scores today:

Geekbench Search - Geekbench

Highest 1T: 2915 1T, 11448 nT
Highest nT: 2903 1T, 11481 nT

Decent if the laptop is cool and quiet.

 

[DokiDoki]

A 2 1/2 year old 8c 6800/6850U (zen 3+) does 10233 CB23 @ 23w

No it doesn't.

From Geekerwan:

 

[igor_kavinski]

Wut? Every heard of VMs or web?

Nearly every React zoomer runs a Mac. The same applies to your average enterprise Java/Python backend dev.

How does that help the masses? Websites I can understand but how many websites are running on Macs? There are no real Mac servers.

 

[DokiDoki]

As it can be seen from the above graph, Jaykhin's results mean that LL is (presumably at 17 W PL1) competitive with Strix Point at 15 W package power

 

[DokiDoki]

How does that help the masses? Websites I can understand but how many websites are running on Macs? There are no real Mac servers.

What does that even mean?

 

[511]

No it doesn't.

From Geekerwan:

View attachment 106157

it looks like at 17W LNL including RAM in package power is better than 8945H/185H at 17W loses to HX370 but it has more cores at 30W it loses to all three of them LNL shines at 17W any higher and it loses efficiency

 

[ikjadoon]

Decent if the laptop is cool and quiet.

Agreed. These are actually mini-PCs from Khadas.

With better cooling & slightly more stable power vs the laptop results so far, LNL's Lion Cove 1T Pts / GHz appears stable at 584 Pts / GHz (comparison chart).

Best 228V in Khadas mini-PC: 2621 Pts / 4.487 GHz = 584.1 Pts / GHz
Best 268V in Khadas mini-PC: 2915 Pts / 4.989 GHz = 584.3 Pts / GHz

 

[Philste]

SKT score about 1595 pts at 4.6GHz in CB R23,

Source?

 

[yuri69]

How does that help the masses? Websites I can understand but how many websites are running on Macs? There are no real Mac servers.

Mac as development tool probably provides faster feedback loop (running tests, building locally, etc.) leading to shorter development time. That's the end user benefit.

 

[Abwx]

As it can be seen from the above graph, Jaykhin's results mean that LL is (presumably at 17 W PL1) competitive with Strix Point at 15 W package power

At 30W it score 10200 while Strix Point in this curve score 10000 at 15W.

Source?

Assuming that ARL s P cores have only 1% better IPC than Zen 5 in CB R23, wich is dubbious since CB traditionaly favour Intel but let s take this basis to not under estimate SKT, then at their max all core frequency, wich is 5.4GHz, ARL s 8 P cores score 17600 pts.

Since the total score is 43118 pts this imply that the 16 e cores at 4.6GHz score 25518 pts, wich amount to 1595 pts/core/4.6GHz and about 4% better than a 5600X wich score 1536 pts at about the same frequency.

And given that Intel got a 10% perf/clock uplift vs AMD in CB R23 comparatively to CB R15 then Zen 3 should have 1.1/1.04 = 5.7% better IPC in CB R15, so overall SKT and Zen 3 should have comparable IPC as far as we re talking CB R23 + CB R15.

 

[DokiDoki]

At 30W it score 10200 while Strix Point in this curve score 10000 at 15W.

It can score ZERO in Cinebench for all I care. It has no relevance to how people use their laptops.

 

[Abwx]

It can score ZERO in Cinebench for all I care. It has no relevance to how people use their laptops.

It had a relevance a few posts ago :

As it can be seen from the above graph, Jaykhin's results mean that LL is (presumably at 17 W PL1) competitive with Strix Point at 15 W package power

Lol of a lol, talk of being of bad faith, guess that the ignore button is the way.

 

[DokiDoki]

It had a relevance a few posts ago :



Lol of a lol, talk of being of bad faith, guess that the ignore button is the way.

That comment was me correcting @majord's false claim that Zen 3+ scores 10000 points at 23W. Beyond that it has no relevance to me.

The only one who engages in masturbation using Cinebench scores in this forum, and, more often than not, by using ONLY one source - Computerbase - is YOU.

Don't worry - I won't be flooding your notifications by replying to you any longer.

 

[mikk]

No it doesn't.

From Geekerwan:

View attachment 106157

It falls apart at 10W. Pretty sure in this power range LNL won't fall apart. I'm curious about the scores at 8W in fanless mode.