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.

 

[poke01]

Nova lake looks more interesting as the days go by

 

[poke01]

Next is SOCAMM, that’s the future of memory on PC

 

[gdansk]

Next is SOCAMM, that’s the future of memory on PC

Really? Why? Won't it be expensive if ML farms are buying all of it?

Nova lake looks more interesting as the days go by

It has its Perseus too. But that's if it survives.

 

[poke01]

Really? Why? Won't it be expensive if ML farms are buying all of it?

If Framework wants modular RAM with fast bandwidth, this is the only option.

 

[adroc_thurston]

Won't it be expensive if ML farms are buying all of it?

buying all of what? DRAM is DRAM, sold in wafers.

 

[gdansk]

buying all of what? DRAM is DRAM, sold in wafers.

I thought it was sold in reels.

 

[poke01]

I thought it was sold in reels.

It’s sold in trays

 

[511]

If Framework wants modular RAM with fast bandwidth, this is the only option.

CAMM/LPCAMM exists

 

[poke01]

CAMM/LPCAMM exists

Why didn’t FW use it for Strix Halo then?

 

[511]

Why didn’t FW use it for Strix Halo then?

Ask FW not me lol also SOCAMM has not been standardize by JEDEC yet

 

[moinmoin]

Why didn’t FW use it for Strix Halo then?

FW CEO said on LTT that AMD looked into it but couldn't manage Strix Halo to ensure the necessary signal integrity.

 

[511]

FW CEO said on LTT that AMD looked into it but couldn't manage Strix Halo to ensure the necessary signal integrity.

Now is this a problem on AMD Side or LPCAMM side we don't know

 

[jpiniero]

I recall someone on this thread proclaiming that Intel's introduction of parts incompatible with AVX-512 would be a trump card for Intel, because Intel's bigger market share would instantly make the incompatible part some sort of industry standard.

That was the height of "cope" logic.

Thankfully, Intel dropped that insanity today...

I wouldn't end up being surprised that ends up not being the case and client doesn't have AVX10 either.

 

[511]

I wouldn't end up being surprised that ends up not being the case and client doesn't have AVX10 either.

Client has AVX10 Confirmed by Intel Principal Engineer also this info is public next client will support APX+AVX 10

 

[jpiniero]

Client has AVX10 Confirmed by Intel Principal Engineer also this info is public next client will support APX+AVX 10

I had kind of assumed that was the AVX10 256-bit though. Which is now gone.

 

[igor_kavinski]

Which is now gone.

Good riddance.

 

[511]

I had kind of assumed that was the AVX10 256-bit though. Which is now gone.

Which means there is only 512 BIT Vector in AVX 10 and I think they are doing Zen 4 like AVX-512 in Arctic wolf and full AVX-512 in P cores

 

[DavidC1]

Now is this a problem on AMD Side or LPCAMM side we don't know

Neither.

When you are dealing with 1mm2 being a big difference, you are inevitably sacrificing signal integrity when making a module that is user replaceable. User replaceable memory CANNOT have the same signal integrity as a soldered down one, it's physically not possible.

 

[poke01]

Neither.

When you are dealing with 1mm2 being a big difference, you are inevitably sacrificing signal integrity when making a module that is user replaceable. User replaceable memory CANNOT have the same signal integrity as a soldered down one, it's physically not possible.

How does SOCAMM address this problem? It looks like it it provides more bandwidth than LPCAMM can.

 

[jpiniero]

Which means there is only 512 BIT Vector in AVX 10 and I think they are doing Zen 4 like AVX-512 in Arctic wolf and full AVX-512 in P cores

What I was implying was the idea that the E cores do not support AVX10's 512-bit mode, only 256-bit. Which is now gone.

 

[igor_kavinski]

What I was implying was the idea that the E cores do not support AVX10's 512-bit mode, only 256-bit. Which is now gone.

Shouldn't be too much problem for Intel to implement what AMD is doing with Zen 5c. Their E-core team is clearly superior in intellect.

 

[TESKATLIPOKA]

No but it is not rocket science given the die sizes and additional requirements on the Motherboard

But that's only the production cost. It doesn't say anything about how much Intel or AMD is asking for those chips and how will the OEMs price those laptops in the end. So we still don't know how much more expensive Strix Halo laptop is compared to a Panther Lake laptop.

According to notebookcheck(Link)
RX 8060S is 177% faster than ARC 140T in Cyberpunk at Full HD.
Even If Panther Lake provided 50% boost, Strix Halo would still be 85% faster.
So Panther Lake shouldn't cost more than 1350euro, or It would have worse gaming perf/€ than Asus Rog Flow(2509€ in my country), which already has bad perf/€ vs CPU+dGPUs.

 

[GTracing]

What I was implying was the idea that the E cores do not support AVX10's 512-bit mode, only 256-bit. Which is now gone.

I'm pretty sure AVX10 256bit is still valid. Are you thinking of how they ditched the 128bit version?

Edit: Ah, nevermind. Apparently this with the past week. Intel's AVX10.2 white paper no longer mentions a 256bit version.

 

[MS_AT]

I'm pretty sure AVX10 256bit is still valid. Are you thinking of how they ditched the 128bit version?

128b, 256b, 512b are all valid. The difference is before the latest change you could have a CPU that stopped at 256b and was not supporting 512b. With the new change every CPU claiming to support AVX10.2 has to be able to execute 512b ops so you don't need to have 2 binaries. But you can still write code that targets lower register widths.

 

[511]

But that's only the production cost. It doesn't say anything about how much Intel or AMD is asking for those chips and how will the OEMs price those laptops in the end. So we still don't know how much more expensive Strix Halo laptop is compared to a Panther Lake laptop.

It will still be proportional to the amount it cost to make them will it not ?

According to notebookcheck(Link)
RX 8060S is 177% faster than ARC 140T in Cyberpunk at Full HD.
Even If Panther Lake provided 50% boost, Strix Halo would still be 85% faster.
So Panther Lake shouldn't cost more than 1350euro, or It would have worse gaming perf/€ than Asus Rog Flow(2509€ in my country), which already has bad perf/€ vs CPU+dGPUs.

Yeah sure Strix Halo is indeed gonna be faster but how much cause Cyberpunk is CPU heavy in the same link a 285H with 140T is better than a Lunar Lake Arc 140V while 140V is clearly better in terms of architecture and it's HW capabilities.

I would say Halo would be 50-70% faster than the Xe3 iGPU