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+0+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.

 

[coercitiv]

But of course, if the main benefit is that the NEXT CPU is easier to make means the first CPU gets few of the benefits and all of the drawbacks.

The most import benefit of tiles is significant reduction in validation times. Allows them to iterate very rapidly.

We'll have to wait and see how their execution improves with the next few product launches. I would still have liked to see tiles delivered in desktop and HEDT first, as the mobile chips need to squeeze efficiency from every source possible. Also, let's hope they don't have a change of heart with regards to the optimal number of tiles in their mobile lineup, otherwise the first two generations will get the few benefit / all drawbacks treatment.

That being said, ANY tile approach is a welcome addition from Intel in my book, I said before they need experience with disaggregation before they learn how to make it work competitively. Like I said already, this experience will pay dividends in the future.

 

[SiliconFly]

.

 

[Hulk]

It's 100% better than RPL without a doubt (I was just talking abt ST).

Almost everything in MTL is awesome. The tiles, GPU, NPU, new process, SoC, NoC Interconnect, Foveros, etc. Everything is perfect; except RWC.

And tbh, I'm not a big fan of RWC. It let me down. Thank god they're killing it sooner rather than later.

I believe that the minor changes to Rapter Cove to turn in into Redwood Cove were simply to offset the latency penalty in moving to the tile based format. I write this because the only evidence we have to any changes to Raptor Cove>Redwood Cove are the "larger L2 cache" as written about by Anandtech. This would be the obvious simple change to offset the tiled latency.

Raptor Cove and Redwood Cove in their respective parts have essentially the same IPC. The E's (Crestmont) in MTL do seem to show some minor but measurable IPC improvement. It was reported in Anandtech to be 4-6%.

As someone else noted the MTL will be the baseline for ARL, which is good for Intel. While MTL is not a bad part, and is exciting on paper, in execution there isn't much to be cheering about. It is however serving it's purpose of making the move to tiles from monolithic for Intel. For ARL they only need to work with the CPU tile since the rest of it is basically the same as MTL.

 

[igor_kavinski]

And tbh, I'm not a big fan of RWC. It let me down. Thank god they're killing it sooner rather than later.

Maybe it's not the compute tile's fault. Maybe it's coz the other engineers botched the job assigned to them: https://forums.anandtech.com/thread...-prediction-for-core-to-core-latency.2618699/

I mean, look at those horrendous latencies. It's a miracle that Intel still came out with a sellable product!

 

[Ghostsonplanets]

Chips 'N Cheese has been dropping some amazing content on MTL GFX IP. Intel really has a solid and very competitive part, if let down by still immature software.

Intel’s Ambitious Meteor Lake iGPU

Intel and AMD both tried to ship iGPUs fast enough to compete with low end discrete cards over the past 10 years with mixed results.

chipsandcheese.com

Raytracing on Meteor Lake’s iGPU

Meteor Lake iGPU is a testament to Intel’s mobile graphics aspirations.

chipsandcheese.com

 

[DrMrLordX]

The most import benefit of tiles is significant reduction in validation times. Allows them to iterate very rapidly.

Tiles allow Intel to partially evade yield problems on new/fishy nodes by fabbing smaller dice.

 

[FlameTail]

Lunar Laks sounds like it will be perfect for gaming handhelds, but there is a potential issue:

The 8-core CPU: Well, what's the problem, you ask. Most games are optimized for 8-core CPUs, right?

But the caveat is that when people say that- they mean a CPU having 8 P-cores with SMT.

Lunar Lake has a 4P+4E CPU, meaning half of the cores are E-cores. And non of them have SMT!

 

[dullard]

The 8-core CPU: Well, what's the problem, you ask. Most games are optimized for 8-core CPUs, right?

Most games are optimized for 4-core or 6-core CPUs. This is because that is what was available for quite a few gaming systems for many years. But maybe you want to exclude the entire history of games and only focus on new games. Some new games do better with 8 cores and very, very few benefit from more than 8 cores.

I'll talk about AMD CPUs here since they make the core count easier to explain without complications of E cores. Compare the Ryzen 7700X (8 cores) vs 7600X (6 cores). Compare the Ryzen 7700 (8 cores) vs 7600 (6 cores). Notice that there is only a tiny difference in gaming performance.



https://www.techspot.com/review/2602-amd-ryzen-7600-7700-7900/

 

[Ghostsonplanets]

Lunar Laks sounds like it will be perfect for gaming handhelds, but there is a potential issue:

The 8-core CPU: Well, what's the problem, you ask. Most games are optimized for 8-core CPUs, right?

But the caveat is that when people say that- they mean a CPU having 8 P-cores with SMT.

Lunar Lake has a 4P+4E CPU, meaning half of the cores are E-cores. And non of them have SMT!

Lunar Lake setup is more like Apple M series rather than current Intel hybrid. The E cores are more like LPE cores. They don't have L3 cache neither have access to the 4P L3 (They're "off-ring"). So the P core needs to do C2C with the LPE cores.

So I don't know how that would fare with gaming workloads. Lack of HT will hurt here as the P cores could provide some additional performance in gaming if they had it.

Lunar Laks sounds like it will be perfect for gaming handhelds, but there is a potential issue:

But I would actually like to see LNL on HH PCs because it would mean it's available at cheaper price points and I can have hope (inhales hopium) LNL will also appear on <$800 laptops.

A fanless, highly-efficient x86 laptop would be "chief kiss"

 

[S'renne]

Lunar Lake setup is more like Apple M series rather than current Intel hybrid. The E cores are more like LPE cores. They don't have L3 cache neither have access to the 4P L3 (They're "off-ring"). So the P core needs to do C2C with the LPE cores.

So I don't know how that would fare with gaming workloads. Lack of HT will hurt here as the P cores could provide some additional performance in gaming if they had it.


But I would actually like to see LNL on HH PCs because it would mean it's available at cheaper price points and I can have hope (inhales hopium) LNL will also appear on <$800 laptops.

A fanless, highly-efficient x86 laptop would be "chief kiss"

Didn't some older games instead runs worse on cores with HT?

 

[coercitiv]

Didn't some older games instead runs worse on cores with HT?

Not only that, in power limited scenarios the number of cores is only part of the equation, perf/w within the limited power envelope is crucial.

 

[Ghostsonplanets]

Didn't some older games instead runs worse on cores with HT?

Yes in some games, no on others. Very game dependent.

Not only that, in power limited scenarios the number of cores is only part of the equation, perf/w within the limited power envelope is crucial.

True. LNL will have a limited power budget, specially in gaming scenarios where GPU and Memory subsystem are quintessential, so feeding less cores means more power can go towards the others components.

I was making my judgement based on the fact current gen consoles are/will all be 8C/8(16T) based. But Lion Cove should be >2x over Skylake PPC, so I guess it's fine.

 

[SiliconFly]

.

 

[Ghostsonplanets]

And most of the games are GPU bound. The tGPU in LNL is just entry level when it comes to real gaming. While gaming, the LNL tCPU is more than sufficient to keep the tGPU fully saturated at all times.

If the LNL is matching or surpassing MTL 8 Xe iGPU performance, I wouldn't call it entry level. From what I gather in reviews, MTL tGPU can comfortably run modern AAA games at low-medium 1080p30 or above. And in the more demanding cases, there's XeSS or FSR.

In a world where Apple fanless laptops can run AAA games (That are ported to Mac) and there's a push towards PC Handhelds, GPU performance at low power is a very important aspect.

Lunar Lake can have an edge here at novel designs due to the scalability and SKUs optimized for fanless designs.

Gaming problem arises, when we try to couple LNL with a good dGPU. Then things'll surely go haywire. But that isn't LNL's use case.

Completely agree here.

 

[SiliconFly]

.

 

[uzzi38]

Saw a leak a few days back (not sure where). Top LNL part is expected to roughly double the performance of top MTL part. This puts it roughly somewhere near RTX 3050 or AMD equivalent. It's very entry level (even with XeSS enabled).

I'm 100% certain whoever said that is wrong, sorry.

I'm sure Lunar Lake will be a huge step up over what Intel has now for 15W iGPU performance, but at this power level they're currently quite far behind the competition. Also 3050 performance tier is far, far, far too ambitious.

 

[Gzxy]

Is LNL's use case gaming? I would say ARL non U will be aimed for gaming. Intel wants the market share that Apple has gained just because of the battery life + AI capabilities. Cores play a big role especially when you are limited to 4C - 4T. I dont know if the 4E will have L3 or be outside the bus (stupid if that's the case) but Panther will retain the 4P and add +4E Skymont.

 

[HoveringStyle]

Saw a leak a few days back (not sure where). Top LNL part is expected to roughly double the performance of top MTL part. This puts it roughly somewhere near RTX 3050 or AMD equivalent. It's very entry level (even with XeSS enabled).

The only 2x articles I'm finding are about Sisoftware Sandra GPU processing scores being 2x MTL at the same number of "EUs" (64) and 26% faster than the top MTL part. Who knows how that'll translate into gaming.

 

[SiliconFly]

.

 

[SiliconFly]

.

 

[Ghostsonplanets]

MTL has around 1000 & LNL has around 1900+ (almost 2000).

Which benchmark? TimeSpy Graphics MTL easily does >3000+.

 

[SiliconFly]

.

 

[HoveringStyle]

Here. It's definitely the same leak but the MTL score you mention is for a 64EU/4 Xe core part, the full one gets just over 1500.

 

[SiliconFly]

.

 

[SiliconFly]

.