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.

 

[Jan Olšan]

FLAC encoding is multithreaded.

flac: Encodes, decodes and tests FLAC files – The Geek Diary

flac is a command-line tool that provides functionality for encoding, decoding, and testing FLAC (Free Lossless Audio Codec) files. FLAC is a popular audio format known for its lossless compression, which means it retains the original audio quality while significantly reducing the file size. The...

www.thegeekdiary.com

Does the MT encoder achieve the same compression ratio though? There was even a GPGPU encoder for FLAC once, but that was precisely the problem.

(Just run more instances at once...)

 

[ondma]

Here is my take on multithreaded performance. I too thought it was a terrible decision to remove HT from ARL. Now, I am not so sure. Still doesnt make sense to me, except maybe from a security point of view. However, MT performance of ARL in fine in the apps that it works well for. The fact that it does so well in some nT apps and badly in others makes me think there is some other bottleneck or flaw in the architecture. There does seem to be sufficient raw computing power there when it can be effectively utilized.

As to how important nT is to the average user, this is my point of view. I use my PC primarily for gaming, but also for some light video editing. Gaming performance is by far priority one. It doesnt really matter that much to me whether I can encode a video in 15 min or 20. I can just start the process and go do something else until it finishes. OTOH, there is nothing I can do if the cpu doesnt provide good gaming performance, other than perhaps changing settings. So I can easily work around limitations to nT performance, while I cant do that for gaming without sacrificing video quality.

Now obviously, depending on ones use case, the priority could vary. For sure if one is doing video editing professionally, it is essential to complete the job as quickly as possible, so the priority would be more to nT performance.

 

[Hulk]

It is hard to argue against this:


and even harder when you consider this:


ARL is competitive but Zen 5 wins the game. Also, the apps where ARL wins by a good margin are probably enhanced by the NPU, which Zen 5 does not have.

 

[Kocicak]

Here is my take on multithreaded performance. I too thought it was a terrible decision to remove HT from ARL. Now, I am not so sure. ... It doesnt really matter that much to me whether I can encode a video in 15 min or 20. ...

With 24 core Intel CPU, the HT benefit is about 10%, if the app can really take advantage of it. So you are looking at 15 vs. 17 minutes.

The story is a little bit different with AMD CPUs, when all 16 cores can do multithreading, there the benefit is higher.

 

[ondma]

With 24 core Intel CPU, the HT benefit is about 10%, if the app can really take advantage of it. So you are looking at 15 vs. 17 minutes.

The story is a little bit different with AMD CPUs, when all 16 cores can do multithreading, there the benefit is higher.

Of course, the 15 vs 20 min was just an off the cuff number to make my point. Losing HT probably will hurt more on lower end of the stack CPUs though, since they have fewer E cores. Something I didnt mention though, is that whatever reason Intel used to decide to remove HT seems like an indication of the fact that the P core is a poor design, since AMD uses it on all cores and still is more efficient.

 

[ondma]

It is hard to argue against this:
View attachment 110465

and even harder when you consider this:
View attachment 110466

ARL is competitive but Zen 5 wins the game. Also, the apps where ARL wins by a good margin are probably enhanced by the NPU, which Zen 5 does not have.

Actually, that makes ARL look better than I thought it would. If you remove the -68% outlier, ARL is nearly equal in applications, and competitive in power usage. The big problem is the horrible and inconsistent gaming performance.

 

[desrever]

Actually, that makes ARL look better than I thought it would. If you remove the -68% outlier, ARL is nearly equal in applications, and competitive in power usage. The big problem is the horrible and inconsistent gaming performance.

Remove the AI bullshit benchmarks in there and its slower in 90% of applications. Why would you remove photoshop as an outlier when it is a legit application?

 

[igor_kavinski]

If you remove the -68% outlier, ARL is nearly equal in applications

With HT, ARL could've probably won in all those lost benchmarks, except for the -68% which is just too high to overcome without special help from AVX-512.

 

[ondma]

Remove the AI bullshit benchmarks in there and its slower in 90% of applications. Why would you remove photoshop as an outlier when it is a legit application?

Of course, you can remove any benchmarks you want. You could look at it the other way and say ARL is more "future proof" because it utilizes AI. My point simply was that the benchmark is clearly an outlier.

 

[igor_kavinski]

My point simply was that the benchmark is clearly an outlier.

May change if the developer decides to code that specific benchmark or application to use Arrow Lake's NPU.

 

[Hulk]

Let's talk use cases. I have two main rigs. My 14900K desktop and my old 8250u Skylake MS Surface 2, which is overmatched for much of what I do.

Here are the applications and operations that when I do them on my laptop I wish I was on the desktop. I am going to buy a new a new laptop soon but really want a Surface Laptop with Lunar Lake so that is why I haven't purchased already.

PureRaw for RAW photo conversion. I can batch this on either system but it painful low without a GPU. 15 seconds per photo with ARC750 desktop, 264 seconds per photo with Intel Graphics 620 in the Surface Laptop 2. That is more than an order of magnitude.

Magix Vegas Pro 21 for video editing/encoding. There are two parts to this timeline preview and rendering/encoding a project. I need the strong P cores and my ARC750 to make editing creatively possible and not painful. It is painful on my laptop. Encoding is also much faster on the desktop but not impossible on the laptop as I can run it overnight if needed. Encoding using Voukoder in Vegas will load up the P cores to 90%+ and the E's to around 60%. Encoding two projects simultaneously (2 Vegas instances open) will load all cores to 100%. It's rare that I need to do that.

Photoshop. I can do some photoshop work on my laptop but it is cumbersome and I'm waiting too much.

Presonus Studio One. Multitrack recording is no problem on the laptop as most of the heavy lifting is done by the interface. The P cores are needed for mixing when the plug-ins get heavy.

Topaz Photo AI - Relies heavily on both GPU and some CPU. I can process a photo or two in a pinch on the laptop but it's quite the painful process.

Topaz Video AI - Forget about it on the laptop. No way. It's quite taxing on the desktop with the ARC750 as well but doable.

Time Xpress. This is payroll software I use for work. Running a report takes a long time, like 5 or 10 minutes on the laptop. Again, doable, but there are things I could be doing with this time.

Opening CorelDraw, MS Phone Link, and a few others I can't remember right now seem to take forever on the laptop. Of course this is more a slow memory/SSD/storage system issue rather than a compute one I would think.

 

[igor_kavinski]

PureRaw for RAW photo conversion. I can batch this on either system but it painful low without a GPU. 15 seconds per photo with ARC750 desktop, 264 seconds per photo with Intel Graphics 620 in the Surface Laptop 2. That is more than an order of magnitude.

Magix Vegas Pro 21 for video editing/encoding. There are two parts to this timeline preview and rendering/encoding a project. I need the strong P cores and my ARC750 to make editing creatively possible and not painful. It is painful on my laptop. Encoding is also much faster on the desktop but not impossible on the laptop as I can run it overnight if needed. Encoding using Voukoder in Vegas will load up the P cores to 90%+ and the E's to around 60%. Encoding two projects simultaneously (2 Vegas instances open) will load all cores to 100%. It's rare that I need to do that.

Photoshop. I can do some photoshop work on my laptop but it is cumbersome and I'm waiting too much.

Topaz Photo AI - Relies heavily on both GPU and some CPU. I can process a photo or two in a pinch on the laptop but it's quite the painful process.

Topaz Video AI - Forget about it on the laptop. No way. It's quite taxing on the desktop with the ARC750 as well but doable.

These may get a higher uplift with a Geforce GPU than a CPU upgrade.

 

[igor_kavinski]

Intel Core Ultra 9 285K, Core Ultra 7 265K and Core Ultra 5 245K Gaming and Workstation Review - When more suddenly becomes less | Page 6 | igor´sLAB

The time has finally come and the results for Intel’s new Core Ultra 200 series are on the table. I can already spoil the fact that we will see very interesting, but partly mixed results today…

www.igorslab.de

285K setting new records in some of Igor's professional workstation benchmarks.

Just imagine how invincible it could've been in these same benchmarks with an extra 15 to 30% "free" HT performance.

I think the current CEO has some sort of curse following him around which is why he is so afraid of the Devil and keeps tweeting Bible verses. He hasn't had a single slam dunk since he joined Intel.

 

[GTracing]

Intel Core Ultra 9 285K, Core Ultra 7 265K and Core Ultra 5 245K Gaming and Workstation Review - When more suddenly becomes less | Page 6 | igor´sLAB

The time has finally come and the results for Intel’s new Core Ultra 200 series are on the table. I can already spoil the fact that we will see very interesting, but partly mixed results today…

www.igorslab.de

285K setting new records in some of Igor's professional workstation benchmarks.

Just imagine how invincible it could've been in these same benchmarks with an extra 15 to 30% "free" HT performance.

I think the current CEO has some sort of curse following him around which is why he is so afraid of the Devil and keeps tweeting Bible verses. He hasn't had a single slam dunk since he joined Intel.

What's your issue with Pat? Did he poison your dog or something?

 

[desrever]

What's your issue with Pat? Did he poison your dog or something?

Pat has been a terrible CEO so far, why are you defending him? Do you personally know him or something?

 

[Tup3x]

FLAC encoding is multithreaded.

flac: Encodes, decodes and tests FLAC files – The Geek Diary

flac is a command-line tool that provides functionality for encoding, decoding, and testing FLAC (Free Lossless Audio Codec) files. FLAC is a popular audio format known for its lossless compression, which means it retains the original audio quality while significantly reducing the file size. The...

www.thegeekdiary.com

LibFLAC is not multithreaded. Pretty much all multithreaded audio encoders are more or less experimental. In any case, it makes more sense to just convert multiple tracks parallel. You can basically treat is at as pure single threaded operation in 99% of software out there that people would use.

 

[SiliconFly]

.

 

[GTracing]

Pat has been a terrible CEO so far, why are you defending him? Do you personally know him or something?

I'm not defending him? I'm just tired of reading comment after comment of armchair CEOs bashing him for every reason they can think of.

 

[MS_AT]

ARL is competitive but Zen 5 wins the game. Also, the apps where ARL wins by a good margin are probably enhanced by the NPU, which Zen 5 does not have.

I doubt any of the benchmarks used by TPU are NPU enabled and NPU in ArrowLake is nothing to write home about.

With 24 core Intel CPU, the HT benefit is about 10%, if the app can really take advantage of it. So you are looking at 15 vs. 17 minutes.

You are estimating HT gain of LionCove based on RaptorCove and Lion is noticeably wider core than Raptor. It stands to reason it might have higher HT uplift than Raptor, since more of the core is likely to sit idle, but we will never know.

Just imagine how invincible it could've been in these same benchmarks with an extra 15 to 30% "free" HT performance.

If it was free they would have implemented it

 

[desrever]

He's done a better job than his predecessors. He's no superman. But not a wimp either.

By what metric? Intel is just as bad now if not worse off than under any previous CEO. Their future is also worse than ever. Pat is the worst CEO in Intel history at this point since he is driving the company to bankruptcy. Intel stock in down >60% since he took over.

I'm not defending him? I'm just tired of reading comment after comment of armchair CEOs bashing him for every reason they can think of.

Because the CEO is responsible for the company and its in their job description? Why does anyone need to be an "armchair CEO" to give criticism?

 

[adroc_thurston]

Intel is just as bad now if not worse off than under any previous CEO

Well, no, the current crop of products wasn't made under his tenure. Pat will be judged by foundry efforts and stuff like Unified Core in the end.

Pat is the worst CEO in Intel history at this point since he is driving the company to bankruptcy. Intel stock in down >60% since he took over.

They're just 2012-era AMD. They may recover.

 

[GTracing]

Because the CEO is responsible for the company and its in their job description? Why does anyone need to be an "armchair CEO" to give criticism?

Making fun of him for being religious is not criticism.

 

[MoistOintment]

By what metric? Intel is just as bad now if not worse off than under any previous CEO. Their future is also worse than ever. Pat is the worst CEO in Intel history at this point since he is driving the company to bankruptcy. Intel stock in down >60% since he took over.


Because the CEO is responsible for the company and its in their job description? Why does anyone need to be an "armchair CEO" to give criticism?

But how much of that specifically due to decisions he made? The biggest damage he did to the stock was cutting the dividends, even though that's what was needed for the company. Lip-Bu Tan resigned because Pat was laying off only 15% of the workforce - he wanted much bigger layoffs. MTL/ARL, SPR, RPL - all inherited messes that had the wheels in motion already.

Realistically, what choices would you have done differently?

I like that one of the first things he did was cancel stock buybacks. That he's forcing fab front and center with separate reporting with industry standard PDKs and tooling. With Intel design required to use industry standard designs that are foundry portable. I like the renewed focused on mobile, even if it hurts me personally due to the lack of focus on desktop.

Intel as a company has a lot of missteps due to past errors: Entering AI and dGPU too late. Adopting EUV too late. 10nm failures with no backup plan. The mess that is MTL/ARL - all inherited messes that he's trying to navigate through the consequences of.

I'm bullish on Intel's long term future and bearish on their short term. I think he's laying a lot of important ground work success that the next CEO will get credit for.

 

[Hulk]

Intel Core Ultra 9 285K, Core Ultra 7 265K and Core Ultra 5 245K Gaming and Workstation Review - When more suddenly becomes less | Page 6 | igor´sLAB

The time has finally come and the results for Intel’s new Core Ultra 200 series are on the table. I can already spoil the fact that we will see very interesting, but partly mixed results today…

www.igorslab.de

285K setting new records in some of Igor's professional workstation benchmarks.

Just imagine how invincible it could've been in these same benchmarks with an extra 15 to 30% "free" HT performance.

I think the current CEO has some sort of curse following him around which is why he is so afraid of the Devil and keeps tweeting Bible verses. He hasn't had a single slam dunk since he joined Intel.

Free? Aren't extra transistors associated with HT? Extra transistors mean additional die area, hot spots, and potentially reduced frequency, which is the point of the P cores.

Furthermore you are talking about extra performance from the 8 P cores so let's make sure we are throwing around numbers that make sense. Before doing calculations I have a feeling 15 to 30% is wildly optimistic consider the fact that Skymont carries the weight of the MT performance for ARL.

Assuming 5.4GHz could be maintained when HT is actually improving P cores performance by 35%, which I doubt without serious cooling, then 35% additional for the 8P cores would result in +14% overall performance in CB R24, which is higher than your low estimate and probably the best case for such a situation.

The reality of the situation as explained to us by the stupid Intel engineers, which all of us are smarter than, and the really stupid CEO, who any of us could replace tonight and we could turn the company around by next week, is that anything that hinders ST performance is illogical and a detriment to the ARL design philosophy. You see thermal considerations and transistor budget are a real thing for those dumb engineers. Of course we know better but they are stuck in their old ways I guess.

 

[Hulk]

These may get a higher uplift with a Geforce GPU than a CPU upgrade.

Of course that is true. But the reality is that on average it takes me longer than 15 seconds to open, edit, and save a photo in PS so the PureRaw conversions are working faster in the background than I am in the foreground. There is a point where fast enough is fast enough for me.