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

[Tigerick]

Wildcat Lake (WCL) Specs

Intel Wildcat Lake (WCL) is upcoming mobile SoC replacing Raptor Lake-U. 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 Q1 2026.

		Intel Raptor Lake U	Intel Wildcat Lake 15W?	Intel Lunar Lake		Intel Panther Lake 4+4+4
Launch Date		Q1-2024	Q2-2026	Q3-2024		Q1-2026
Model		Intel 150U	Intel Core 7	Core Ultra 7 268V		Core Ultra 7 365
Dies		2	2	2		3
Node		Intel 7 + ?	Intel 18-A + TSMC N6	TSMC N3B + N6		Intel 18-A + Intel 3 + TSMC N6
CPU		2 P-core + 8 E-cores	2 P-core + 4 LP E-cores	4 P-core + 4 LP E-cores		4 P-core + 4 LP E-cores
Threads		12	6	8		8
Max Clock		5.4 GHz	?	5 GHz		4.8 GHz
L3 Cache		12 MB		12 MB		12 MB
TDP		15 - 55 W	15 W ?	17 - 37 W		25 - 55 W
Memory		128-bit LPDDR5-5200	64-bit LPDDR5	128-bit LPDDR5x-8533		128-bit LPDDR5x-7467
Size		96 GB		32 GB		128 GB
Bandwidth				136 GB/s
GPU		Intel Graphics	Intel Graphics	Arc 140V		Intel Graphics
RT		No	No	YES		YES
EU / Xe		96 EU	2 Xe	8 Xe		4 Xe
Max Clock		1.3 GHz	?	2 GHz		2.5 GHz
NPU		GNA 3.0	18 TOPS	48 TOPS		49 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.

 

[DavidC1]

Comparison against RTX 4050

RTX 4050 is 10% away from GTX 1080.

Though it'll take another decade before that performance comes at $70.

It is. You should check the picture I posted. This is 25W WCL die; Intel should release low-power WCL later. At least Intel able to fit in NPU5 in the same SoC die (CPU+GPU+NPU)

So it's like Intel calling N100 series Alderlake-N.

@jdubs03 If Pantherlake outperforms Arrowlake-H in ST at lower clocks, again it's a very good result, backing up TPU preview saying Cougar Cove has 5-10% perf/clock gains. So despite all the ups and downs about whether 5.1GHz for PTL would be enough or that Geekbench scores are too low, it looks like a very good product overall, especially in the iGPU space which won't have a competitor for quite a while.

 

[regen1]

It is. You should check the picture I posted. This is 25W WCL die; Intel should release low-power WCL later. At least Intel able to fit in NPU5 in the same SoC die (CPU+GPU+NPU)

View attachment 136260

WCL config looks similar to the first pic you posted but those U5 332 and 322 on the 2nd pic with 46TOPS NPU are part of Panther Lake series, not WCL.
WCL has the same NPU architecture as PTL but only 1 tile instead of 3.

 

[Hulk]

Some interesting claims in there. If validated, it’s a good step. Though the X2E(E) will still be out of reach for now.

Up to 50% lower processor power with similar performance vs. AMD Strix Point with an Intel® Core™ Ultra X9 388H vs. AMD Ryzen™ AI 9 HX 370

As measured by Cinebench 2024 Single Core (PTL at 10W, AMD at 20W)​

 

[Tigerick]

Guys, I have updated the specs of WCL with table. Seems like Intel is positioned WCL to be low-end ICU 5 with 128-bit LPDDR5-6800 memory bus. As for low-power WCL-U series which is supposed to replace Raptor Lake 150U, the specs look subpar. As usual, Intel EUV sucks at Fmax compared to old DUV-powered Raptor Lake. Here is last picture of PTL-H to show 4+4+4 configuration in case some people still not get it:

 

[511]

Guys, I have updated the specs of WCL with table. Seems like Intel is positioned WCL to be low-end ICU 5 with 128-bit LPDDR5-6800 memory bus. As for low-power WCL-U series which is supposed to replace Raptor Lake 150U, the specs look subpar. As usual, Intel EUV sucks at Fmax compared to old DUV-powered Raptor Lake. Here is last picture of PTL-H to show 4+4+4 configuration in case some people still not get it:

by thing logic everything is meh cause Intel 7 Ultra is just frequency maxed node at the cost of power and area even AMD on N4P is having trouble reaching those clocks lmao on a less wide design

 

[jdubs03]

@Hulk Yeah single-core, not multi-core. You’re conflating the two.

 

[Hulk]

@Hulk Yeah single-core, not multi-core. You’re conflating the two.

Yes you are correct. Still, curious to see how PL does.

 

[regen1]

Guys, I have updated the specs of WCL with table. Seems like Intel is positioned WCL to be low-end ICU 5 with 128-bit LPDDR5-6800 memory bus. As for low-power WCL-U series which is supposed to replace Raptor Lake 150U, the specs look subpar. As usual, Intel EUV sucks at Fmax compared to old DUV-powered Raptor Lake. Here is last picture of PTL-H to show 4+4+4 configuration in case some people still not get it:

View attachment 136269

Hmm, you added Ultra 5 332 with 46 TOPS NPU to WildCat Lake when the Intel page says it's Panther Lake.



Anyway moving to some WCL pics

https://twitter.com/x/status/2008344805382787146

https://twitter.com/x/status/2008345220900082173

Spoiler: Wildcat Lake

 

[Tigerick]

Hmm, you added Ultra 5 332 with 46 TOPS NPU to WildCat Lake when the Intel page says it's Panther Lake.
View attachment 136270


Anyway moving to some WCL pics

https://twitter.com/x/status/2008344805382787146

https://twitter.com/x/status/2008345220900082173

Spoiler: Wildcat Lake

View attachment 136271
View attachment 136272
View attachment 136273

Hmm, interesting. The specs of PDF and ARK are conflicting. Let's see which one is correct. BTW, the picture of WCL has the same topology as PDF....

 

[fastandfurious6]

WCL PTL summary:

- no real progress in CPU, only GPU

- only real market usecase is handhelds

- Halo #1, PTL #2, STX #3

- STX bad product and Halo too expensive = PTL wins many handheld contracts

PTL would be 100% DOA if AMD had made mobile RDNA4 + 8cores Halo...

Medusa comes way too late

 

[511]

Hmm, you added Ultra 5 332 with 46 TOPS NPU to WildCat Lake when the Intel page says it's Panther Lake.
View attachment 136270


Anyway moving to some WCL pics

https://twitter.com/x/status/2008344805382787146

https://twitter.com/x/status/2008345220900082173

Spoiler: Wildcat Lake

View attachment 136271
View attachment 136272
View attachment 136273

this is a downbin

 

[511]

WCL PTL summary:

- no real progress in CPU, only GPU

the only improvement we get in CPU is better perf/watt that's it

- only real market usecase is handhelds

- Halo #1, PTL #2, STX #3

Halo is a different tier but yeah it's like that

- STX bad product and Halo too expensive = PTL wins many handheld contracts

PTL would be 100% DOA if AMD had made mobile RDNA4 + 8cores Halo...

Yeah nope

Medusa comes way too late

it will be against Nova which has AVX-512 as well so nothing missed on ISA front as well

 

[fastandfurious6]

RDNA3.5 Halo >>>>>> PTL

RDNA4 Halo < PTL ?

lol

 

[511]

RDNA3.5 Halo >>>>>> PTL

RDNA4 Halo < PTL ?

You meant halo RDNA 4 yeah PTL would loose but the problem is 256bit BUS is not cheap in cost PTL would still end up cheaper and thus not DOA

lol

 

[ToTTenTranz]

PTL would be 100% DOA if AMD had made mobile RDNA4 + 8cores Halo...

I'm not so sure about this. Had AMD done e.g. a 16CU RDNA4 + 4x Zen5 + 4x Zen5c on 128bit without a significant change in the cache system, it'd probably still be significantly slower than Panther Lake.

The performance delta from 12CU RDNA3 on Hawk/Phoenix Point to 16CU RDNA3.5 on Strix Point is abysmal because they're both choking on lackluster memory bandwidth. It's probably part of the reason AMD didn't even bother upgrading the iGPU arch and isn't bothering to do so on MDS-P either.


Panther Lake's performance on low power comes from not only support for LPDDR5X 9600MT/s but also from a cache structure where the iGPU has 16MB L2 (Strix Point's has 2MB), plus access to 18MB of the SoC's L3 and then there's 8MB cache in the memory controller. PTL-H's effective bandwidth on the iGPU is a world apart from Strix Point's.

 

[vanplayer]

the only improvement we get in CPU is better perf/watt that's it

At least from SPEC test I knew, the Panther Cove perf/power curve doesn't have any significant improvement compare to previous generations like ARL/LNL/AMD's STX

 

[511]

At least from SPEC test I knew, the Panther Cove perf/power curve doesn't have any significant improvement compare to previous generations like ARL/LNL/AMD's STX

It's Cougar Cove and it's like 10% perf/watt improvement that's more than what a half node gives you but less than a full node

 

[511]

https://twitter.com/x/status/2008566363309437327

WCL 18A die is 74mm2 decent imo

 

[MoistOintment]

RDNA3.5 Halo >>>>>> PTL

RDNA4 Halo < PTL ?

lol

Strix Halo is expensive, borderline vaporware that competes in an entirely different TDP bracket.

If I wanted a thicker, heavy laptop with power consumption over 100W, I'd be looking at something with a dGPU (Which I assume is what most of the market has done considering Strix Halo's limited success in laptop).

PTL Xe12 is in the power range I want from a laptop, and laptops should be judged by how their overall package suits their intended market - not just what the most powerful is, all else be damned.

 

[desrever]

PTL would be 100% DOA if AMD had made mobile RDNA4 + 8cores Halo...

PTL is probably DOA anyways since laptop prices will be through the roof with ram and storage prices so high.

I don't think anyone will be buying any consumer hardware unless they absolutely have to right now.

 

[DavidC1]

Panther Lake's performance on low power comes from not only support for LPDDR5X 9600MT/s but also from a cache structure where the iGPU has 16MB L2 (Strix Point's has 2MB), plus access to 18MB of the SoC's L3 and then there's 8MB cache in the memory controller. PTL-H's effective bandwidth on the iGPU is a world apart from Strix Point's.

Don't forget the improved uarch. Even 32MB is not big enough to run out in gaming scenarios, and that's for the top end part. Xe3 improves matters related to occlusion culling, which will improve in all scenarios, and will be disproportionately beneficial for lower end chips with smaller caches and structure. The 4 Xe core version only has 4MB L2. The slightly increased bandwidth and extra caches are made more effective by better occlusion culling.

Also, since Meteorlake it doesn't share the LL3 with the GPU, so it's just the 16MB, since we know the 8MB memory side cache is useless in terms of extra performance, and it's more for power. The "chiplet era" isn't a free lunch thing. It becomes more like a discreet addition system because now you need to create redundancy in areas not necessary with fully integrated systems like monolithic.

Intel has straight up better iGPU this generation.

PTL would be 100% DOA if AMD had made mobile RDNA4 + 8cores Halo...

Even if AMD had RDNA4, it would at best match it.

 

[ToTTenTranz]

Even if AMD had RDNA4, it would at best match it.

Strix Point with 16CUs RDNA4 and 16-24MB Infinity Cache would probably give hell to Panther Lake, but unfortunately the lack of competition led AMD to sit back and relax. So now the only thing they can realistically use against Panther Lake is Strix Halo sold for cheap.

 

[misuspita]

I hate corporations that get lazy. Intel did it and AMD came with a vengeance, then AMD got coxky and now it turns that Intel may have the upper hand. Which is good. We need competition so both are on their toes, giving their best effort, not some half assed combo

 

[DavidC1]

Strix Point with 16CUs RDNA4 and 16-24MB Infinity Cache would probably give hell to Panther Lake, but unfortunately the lack of competition led AMD to sit back and relax. So now the only thing they can realistically use against Panther Lake is Strix Halo sold for cheap.

It's just going to match it at the best.

I hate corporations that get lazy. Intel did it and AMD CAME with a vengeance, then AMD cot coxky and now it turns that Intel may have the upper hand. Which is good. We need competition so both are on their toes, giving their best effort, not some half assed combo

AMD is just focusing on AI that's why. They could have gave a large cache if it wasn't for the Die Waste Unit.

 

[fastandfurious6]

Strix Halo is expensive, borderline vaporware

I lost you there

Halo will remain the most powerful handheld until zen6

AMD should have scrapped STXpoint and just replace it with half halo: 9700X + half-CU

Half-halo > PTL at same TDP easy