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

[Tigerick]

Wildcat Lake (WCL) Preliminary Specs

Intel Wildcat Lake (WCL) is upcoming mobile SoC replacing ADL-N. WCL consists of 2 tiles: compute tile and PCD tile. It is true single die consists of CPU, GPU and NPU that is fabbed by 18-A process. Last time I checked, PCD tile is fabbed by TSMC N6 process. They are connected through UCIe, not D2D; a first from Intel. Expecting launching in Q2/Computex 2026. In case people don't remember AlderLake-N, I have created a table below to compare the detail specs of ADL-N and WCL. Just for fun, I am throwing LNL and upcoming Mediatek D9500 SoC.

		Intel Alder Lake - N	Intel Wildcat Lake	Intel Lunar Lake		Mediatek D9500
Launch Date		Q1-2023	Q2-2026 ?	Q3-2024		Q3-2025
Model		Intel N300	?	Core Ultra 7 268V		Dimensity 9500 5G
Dies		2	2	2		1
Node		Intel 7 + ?	Intel 18-A + TSMC N6	TSMC N3B + N6		TSMC N3P
CPU		8 E-cores	2 P-core + 4 LP E-cores	4 P-core + 4 LP E-cores		C1 1+3+4
Threads		8	6	8		8
Max Clock		3.8 GHz	?	5 GHz
L3 Cache		6 MB	?	12 MB
TDP		7 W	Fanless ?	17 W		Fanless
Memory		64-bit LPDDR5-4800	64-bit LPDDR5-6800 ?	128-bit LPDDR5X-8533		64-bit LPDDR5X-10667
Size		16 GB	?	32 GB		24 GB ?
Bandwidth			~ 55 GB/s	136 GB/s		85.6 GB/s
GPU		UHD Graphics		Arc 140V		G1 Ultra
EU / Xe		32 EU	2 Xe	8 Xe		12
Max Clock		1.25 GHz		2 GHz
NPU		NA	18 TOPS	48 TOPS		100 TOPS ?

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.

 

[511]

I dunno about those differences but I do know it won'
Yes, considering most people use a browser all the time, then most likely a chat program like discord, all the other background apps and before you know it a lot of RAM is gone. Not even talking about apps getting less and less optimized.
Even if I just use Office suits (at least 2-3 open), likely also Adobe PDF, the amount of apps coming together is already at least 6 and that's before even starting on power using. I wouldn't say ordinary people these days don't need lots more than a decade ago.

Yeah but ordinary people are fine with 32 Max it's enthusiasts like us who need 64GB Ram

 

[igor_kavinski]

So with Arrow Lake, and Skymont e-cores having Raptor Cove IPC, maybe could buy the Core ultra 265K or 285K and disable Lion Cove cores and have a 12-16 Raptor Cove substitute cores with excellent and more importantly consistent core to core latency in mid 30s ms between all 12-16 Skymont cores regardless o which node they are in., And the Skymont cores run at 4.6GHz. Probably can get them overclocked to 5GHz and perfectly setup 12 P core Raptor Lake equivalent with slight lower 5GHz clock. I am excited.

I don't think it's possible to disable all P-cores. At least one P-core must be active for the PC to boot. Are you speaking from personal experience? Does your current mobo let you disable all P-cores and boot with E-cores only?

 

[Wolverine2349]

I don't think it's possible to disable all P-cores. At least one P-core must be active for the PC to boot. Are you speaking from personal experience? Does your current mobo let you disable all P-cores and boot with E-cores only?

II know it does not with Alder and Raptor Lake.

Perhaps and hopefully it will be different hopefully with Arrow Lake??

The Skymont cores are supposed to be so lightyears ahead of the weak FP and even somewhat weak INTEGER Gracemont cores anyways so do not see why there would not be an option to disable all P cores if Skymont truly does have Raptor Cove IPC and real world clock normalized performance the same as it.

The Skymont cores should be able to fly and handle themselves well. The Gracemont were gimped and more limited perhaps that's why you could not disable all P cores on Raptor and Alder? Or was it just Intel and mobo makers being dicks about it.

 

[naukkis]

Intel did give Lunarlake cross latencies. E-core interlatencies in module are lower than big cores in ring. And big core to e -core latency is 55ns and thats for cores that are in different cache domains.

 

[igor_kavinski]

Or was it just Intel and mobo makers being dicks about it.

The N305 chip for example is all E-cores so it should be possible to disable the P-cores but then we don't know if N305 is a special chip with only E-cores or is it the lowest quality P-core+E-core die with all P-cores disabled?

 

[MS_AT]

Raptor Lake it seems there is a little more hit going to e-core clusters in latency where as Meteor Lake, there is none at all and in fact core to core latency even better between Crestmont cores even in different 4 core clusters *as long as on same ring as LPE cores of course are bad which desktop counterparts will not have and are excluded)

https://images.anandtech.com/doci/17601/Core%20i9-13900K%20Core%20to%20core%20latency%203.jpg

Though its a mild increase and pretty consistent even among the 4 different 4 core Gracemont clusters though not quite as good as Raptor Cove to Raptor Cove communication. And despite consistent 40s ns score to core latency GFracemont cores are Skylake IPC or weaker in certain aspects.

Consistency is key. AMD has even better core to core latency within a CCX than Intel as all numbers are in green. But once it exits a CCX/CCD, latency jumps enormously and it is bad as not consistent as a massive latency jump through IF.,

Supposedly AMD is supposed to improve it with Zen 6, but then again with Zen 5 was supposed to be much better, but cross CCD/CCX latency actually much worse than Zen 4 as with Zen 4 jumps into high 70s to 80s ns where as Zen 5 100s of ns almost like Laptop LPE e-cores of Meteor Lake which are irreverent for desktop on Intel side.

I am exited looking at the Core Ultra 155H graph and seeing consistent 30s core to core latency of the Crestmont cores even in different clusters on same ring. Figure Arrow Lake Skymont cores should be no different if not better and just as consistent. My dream chip may be there if I can disable the Lion Cove cores and use only Skymont with a mild 5GHz overclock.

Do you know which method Intel used to produce their latency numbers, as I recall Cheese from Chips&Cheese suggested they are comparing apples to oranges since they used C&C number for Meteor Lake but did not provide how they obtained Lunar Lake numbers

 

[Wolverine2349]

Do you know which method Intel used to produce their latency numbers, as I recall Cheese from Chips&Cheese suggested they are comparing apples to oranges since they used C&C number for Meteor Lake but did not provide how they obtained Lunar Lake numbers

I have no idea. I was just referencing a core to core latency chart Anandtech reviewer used for both the 13900K and Meteor Lake based Core Ultra 155H.

Many have stated they think core to core latencies should be the same between Arrow Lake and Meteor Lake so I was using Meteor Lake Core Ultra 155H as a reference point.

And the core to core latencies between all 8 crestmont e-cores look consistent and good in the mid 30s nanoseconds even across clusters on Meteor Lake per Anandtech review So that is hopeful for Arrow Lake.

https://images.anandtech.com/doci/21282/Intel%20Core%20Ultra%207%20155H%20Core%20to%20Core%20Latency.png

 

[dr1337]

I must be really old, I remember upgrading to 4gb on my pentium 4 and feeling utterly destroyed by windows XP 32bit not letting me use the full amount, even if it was only by a few hundred MBs.

I simply cannot relate with anyone that wants less performance on new hardware. Have any of these people talking about not needing 64gb ever even used a system with that much ram?


All I have open on this PC are two web browsers, discord, and steam. And I am already very near 32GB of RAM.


I've already heard all the arguments, stop using bloated browsers, don't use discord, its only due to windows memory allocation, I don't care. If the software I use wants RAM, why wouldn't I give it RAM? And also if we consider windows itself, don't forget caching is a thing. And with that counted I'm already at 40gb aka way past 32gb.

Now, please consider a high end intel SOC with a decent iGPU. You'd probably want to give at least 8gb of RAM dedicated to graphics right? On a 32gb laptop that cuts you down to 24gb, and that's already less ram than my PC is using to run chrome and firefox. 64gb on laptops, especially ones with brand new extremely powerful chips, shouldn't be scoffed at.

 

[poke01]

I must be really old, I remember upgrading to 4gb on my pentium 4 and feeling utterly destroyed by windows XP 32bit not letting me use the full amount, even if it was only by a few hundred MBs.

I simply cannot relate with anyone that wants less performance on new hardware. Have any of these people talking about not needing 64gb ever even used a system with that much ram?

View attachment 107353
All I have open on this PC are two web browsers, discord, and steam. And I am already very near 32GB of RAM.


I've already heard all the arguments, stop using bloated browsers, don't use discord, its only due to windows memory allocation, I don't care. If the software I use wants RAM, why wouldn't I give it RAM? And also if we consider windows itself, don't forget caching is a thing. And with that counted I'm already at 40gb aka way past 32gb.

Now, please consider a high end intel SOC with a decent iGPU. You'd probably want to give at least 8gb of RAM dedicated to graphics right? On a 32gb laptop that cuts you down to 24gb, and that's already less ram than my PC is using to run chrome and firefox. 64gb on laptops, especially ones with brand new extremely powerful chips, shouldn't be scoffed at.

Ok then what about people who want 128gb they get stuffed?

Why does X Elite top at 64GB or Lunar at 32GB or M3 at 24GB. I guess those people get the M3 Max or Strix Halo.

 

[dr1337]

Ok then what about people who want 128gb they get stuffed?

If its possible to put it on a laptop then IDK why OEMs wouldn't offer it? Apple can do it so why not everyone else eh? I'm pretty sure both MTL and LNL can do at least 64gb but maybe not at the highest speeds? would be extremely sad if they couldn't.

I think no customer should be stuffed, other than stuffed full of good specs.

 

[Anhiel]

Windows caching is recommend to be 1.5x RAM. Then again that's more about your harddrive.

If its possible to put it on a laptop then IDK why OEMs wouldn't offer it? Apple can do it so why not everyone else eh? I'm pretty sure both MTL and LNL can do at least 64gb but maybe not at the highest speeds? would be extremely sad if they couldn't.

I think no customer should be stuffed, other than stuffed full of good specs.

I agree in principle.
It's due to Microsoft. It may not be well known but different Windows licenses have different limits for RAM amount you can use. It's a complicated jungle

Memory Limits for Windows and Windows Server Releases - Win32 apps

Describes the memory limits for supported Windows and Windows Server releases and provides lists of memory limits.

learn.microsoft.com

 

[SiliconFly]

.

 

[SiliconFly]

.

 

[Doug S]

If its possible to put it on a laptop then IDK why OEMs wouldn't offer it? Apple can do it so why not everyone else eh? I'm pretty sure both MTL and LNL can do at least 64gb but maybe not at the highest speeds? would be extremely sad if they couldn't.

I think no customer should be stuffed, other than stuffed full of good specs.

If you're soldering the RAM onto the CPU carrier the higher memory configurations are effectively a separate SKU. You have to have enough volume to make it worth building/stocking that. Intel might figure they have enough volume to do that in the higher end configs that would go into workstation type laptops, but they can't offer it across the whole line. Who is going to buy an i3 with 64 GB. Maybe a few would, but not enough to make it a viable product.

 

[Wolverine2349]

First they don't clock well and second are the IPC comparisons truly apples-to-apples?

I believe that for a given clock, LNC may outperform SKT in any given workload. Just a thought. P cores are still superior when it comes to performance.

That may be true. Though even if they do not clock well, if I can push 5GHz, I would be very happy with Golden Cove real world performance 10-12 P core chip HT disabled and Golden Cove P cores clock 5GHz all core reliably and GLC has similar IPC to Raptor Cove only 5% behind. If such a chip existed and and no stability issues, I would have bought it already.

My hope is Skymont only can be my homogenous 5GHz Raptor/Golden Cove real world performance 5GHz equivalent of 12-16 cores HT disabled.

Yes Lion Cove may perform better clock for clock, but unless Skymont is really gimped compared to GLC/RPC, there is only 8 Lion Cove cores. If only there were 10-12 I would be a buyer and use only Lion Cove cores. But that will not be an option at initial release and likely not ever.

No dealing with Big.Little hybird scheduling quirks if I can turly use Skymont and clock to 5GHz as GLC/RPC equivelent real world 99% gaming/other workloads

Not sure about improvement in core-to core latency, but MTL uses the NOC Fabric interconnect whereas Lunar Lake and Arrow Lake use the latest Coherent Interconnect Fabric for reduced latencies, better power efficiency and faster speeds. Makes a solid difference.

Perhaps even better core to core latency. Just hope its as consistent across all Skymont cores only/. And also hope Lion Cove cores can be fully disabled for a more than 8 strong homogenous non hybrid core config (12-16 Skymont cores as a 12 P core Raptor Lake equivalent but lower power is my hope).

 

[Anhiel]

Frame rates aren't directly dependent on amount of memory used. Just cos an Intel display driver uses only 128 MB or 256 MB of system RAM as VRAM, it does not affect refresh rate/frame rate.


Intel has always used dynamic allocation. And I'm pretty sure that unlike the older drivers, the later RDNA series also uses dynamic allocation. Fixed allocation just doesn't make sense anymore. It's just stupid imho.


Most of the apps, when they tend use lots of RAM, doesn't hurt overall system performance, unless the app is CPU intensive. Meaning, an app can use a terabyte of RAM without hurting system performance if it can adapt well to paging (like browsers with a zillion tabs open). Windows handles paging real good. Esp. with SSDs.

Problem arises only when an app tries to process gigabytes of data simultaneously like multiple large image manipulations parallelly. Lunar Lake isn't designed for such scenarios. For that, go with ARL-H or HX 370.


Office apps and adobe tend to use very little real ram cos they handle paging extremely well. In fact, in a theoretical system with only just Win64 and Office64 and adobe, a 4 GB system will just do the job extremely well (even with VBS added).

Your understanding of windows memory subsystem is just poor.

I never said anything about drivers and since when aren't these basic drivers not part of windows allocation?
I also never said FPS but windows since Vista needs to render, hence, need a buffer. The frame number I gave is about the buffer. You missed the point.

You were criticizing about the amount and now say have not limits would work great and no issues. Aren't grasping for reasons now?

My image test problem was for one single B&W image of 9.3312e8 pixel taking up 1GB of VRAM. And it was just about selecting one color and removing it after adding a transparency layer. It was a simple task done in 5 seconds on the 32b version. Loading was also 2x faster. Ofc it all fit within the system resources.

I was talking about how fast you could end up using lots of apps even for ordinary users and that could raise RAM usage without you realizing it. You missed the point again.

Nobody uses hard drives anymore. It's SSDs now. And with SSD, virtual memory works like a dream automatically in windows. Nothing to configure (windows takes care of everything by itself). Trust me bro!


All the limits you mention are usually well above typical general client user requirements. Usually in terabytes. Not worth wasting our time here unless you're one of those heavy duty server kinda person. Even the paltry lowest windows configuration supports 128 GB minimum. Don't sweat it.

Whether it's classic harddrive or SSD the point I said that was that it's not RAM. You missed the point again.

Aren't you mistaking those for the virtual amount? In any case, it was more significant on older ones. And are facing other layers of limitations, too.

 

[dullard]

All I have open on this PC are two web browsers, discord, and steam. And I am already very near 32GB of RAM.

Seems like your real need is to close browser tabs and learn to use the history to instantly load what you closed weeks/months/years ago. Chrome's grouped history is really helpful for that when you take the time to learn to use it.

Why purchase an expensive new system and then just clog it up, probably soon to a crawl, with unused web browser tabs?

 

[The Hardcard]

Ok then what about people who want 128gb they get stuffed?

Why does X Elite top at 64GB or Lunar at 32GB or M3 at 24GB. I guess those people get the M3 Max or Strix Halo.

Who would want 64GB let alone 128GB with Lunar’s relatively dinky CPU or X Elite’s relatively dinky GPU. 32GB is more than enough for these SOCs. Maybe if someone runs a lot of VMs on the Qualcomm, I suspect people with those needs will rarely consider 1st generation platforms.

The M3 Max has adequate CPU compute for 128 GB, far more than the others. The 40 core/5120 ALU GPU is also far more, yet it is constraining for RAM-filling workloads.

 

[Anhiel]

No dealing with Big.Little hybird scheduling quirks if I can turly use Skymont and clock to 5GHz as GLC/RPC equivelent real world 99% gaming/other workloads

An interesting idea. But you realize are you're practically using only about half the silicon?
R9 9700X costs about 350 USD now. Ultra 9 285k is rumored to be about 700 USD.
So cost wise the Zen5 cores cost about the same as the E-cores. 350/8 vs 700/16
Assuming the e-cores @5GHz are about as fast as the Zen5 IPC wise but Zen5 has SMT.
According to GamerNexus 9700X uses 87.6W in Blender and the graph someone made for the e-cores puts it at about 7.2W
so zen5 8.8W/core presumely all-core @4.5GHz (GamersNexus)
8 zen5 cores: 4.5*1.25*8=45
16 e-cores: 5*1*16=80
so 1:1 it's 90:80 performance wise
90/175.2W vs 80/80W consumption wise
It's close or as good as the same. The actual IPC difference will decide it.

 

[SiliconFly]

.

 

[Wolverine2349]

An interesting idea. But you realize are you're practically using only about half the silicon?
R9 9700X costs about 350 USD now. Ultra 9 285k is rumored to be about 700 USD.
So cost wise the Zen5 cores cost about the same as the E-cores. 350/8 vs 700/16
Assuming the e-cores @5GHz are about as fast as the Zen5 IPC wise but Zen5 has SMT.
According to GamerNexus 9700X uses 87.6W in Blender and the graph someone made for the e-cores puts it at about 7.2W
so zen5 8.8W/core presumely all-core @4.5GHz (GamersNexus)
8 zen5 cores: 4.5*1.25*8=45
16 e-cores: 5*1*16=80
so 1:1 it's 90:80 performance wise
90/175.2W vs 80/80W consumption wise
It's close or as good as the same. The actual IPC difference will decide it.

Yeah will only be using half silicon. But what other choice do I have if I want more than 8 strong homogeneous cores of RPC/GLCZEN4/ZEN5 IPC equivalent on a single node. Skymont cores may be the only hope.

No such option exists. This may be my only hope for such a CPU.

IF Intel or AMD sold a CPU with 10 to 16 good homogeneous cores on a single node than I would buy that instead. But neither do.

AMD maxes at 8 cores per CCD. Intel maxes at 8 p cores. However the Intel Austin team has made such progress that Skymont cores being their e cores and more thsn i of them may be the new p cores performance I want and more than 8 of them to boot and oh on the same die.

We shall see if my idea can come true or not though. Benchmarks will tell compared to Raptor Cove clock normalized.

 

[majord]

Yeah will only be using half silicon. But what other choice do I have if I want more than 8 strong homogeneous cores of RPC/GLCZEN4/ZEN5 IPC equivalent on a single node. Skymont cores may be the only hope.

No such option exists. This may be my only hope for such a CPU.

IF Intel or AMD sold a CPU with 10 to 16 good homogeneous cores on a single node than I would buy that instead. But neither do.

AMD maxes at 8 cores per CCD. Intel maxes at 8 p cores. However the Intel Austin team has made such progress that Skymont cores being their e cores and more thsn i of them may be the new p cores performance I want and more than 8 of them to boot and oh on the same die.

We shall see if my idea can come true or not though. Benchmarks will tell compared to Raptor Cove clock normalized.

What are you trying to achieve though that you specifically need a monolithic solution?

 

[Wolverine2349]

What are you trying to achieve though that you specifically need a monolithic solution?

Want more than 8 good cores on single die. For gaming that is starting to slowly scale to more cores and best set and forget it solution for games without big.little or cross latency penalty.

 

[SiliconFly]

.

 

[majord]

Yeah indeed I'm just not sure it will have the outcome desired. highly unlikely to OC to 5Ghz without straying way outside efficiency curve at the very least.

I get the desire for high performance monolithic solutions with higher core counts, but they aint happening unfortunately. Can't push something that isn't into that realm though. the end outcome is performance at the end of the day, not latency figures measured in ns