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.

 

[yuri69]

It will be a brutal next to years
L in Server
L in Desktop
Game over in GPU
Foundry going nowhere
Okay in Laptop

* The best server product since Rome arrived
* Desktop is a super tiny niche and Intel's gains are comparable to Zen 5 - both are rather small
* GPU has been a standard "Koduri project" - think of AMD Vega
* Foundry remains to be seen
* Laptop is very good and actually is not a niche like desktop or discrete GPUs

Intel is fine

 

[Hulk]

I have to think about this some more but I am not so sure the bulk of the efficiency gain is coming from the TMSC process. Architectural gains, especially Skymont 50+% IPC increase allows for relaxed clocks. As we all know backing down clocks even 10 or 15% can provide huge efficiency gains.

Ultimately, I think Intel may have gone TMSC more for transistor density (economics) rather than efficiency. Meaning ARL on Intel 7 would have been physically and economically huge.

 

[gdansk]

I want to thank the Intel engineer who changed the PL2 from 253W to 250W. Much more pleasant number.

 

[poke01]

The 285K uses more power than 9950X, interesting.

And the results are likewise close but opposite in Blender 4.1, Linux.

Spoiler

View attachment 110221
And as the scene becomes larger it seems 9950X pulls ahead

ahh, I give up. okay now I’m on the same boat as abwx, does Intel really optimise the hell out of Cinebench?

148 ST doesn’t really match in other ST benchmarks and MT is somewhat dubious. Is it scheduling issues, which doesn’t effect effect Cinebench2024? 🧐

 

[Abwx]

NBC review, they have a lot of benches that are not used by other sites even if often
quite relevant.

https://www.notebookcheck.net/Intel-Core-Ultra-9-285K-Arrow-Lake-S-Review-589-CPU-sacrifices-gaming-crown-to-match-AMD-Zen-5-Ryzen-9000-in-power-efficiency.901799.0.html

The 285K uses more power than 9950X, interesting.

ahh, I give up. okay now I’m on the same boat as abwx, does Intel really optimise the hell out of Cinebench?

148 ST doesn’t really match in other ST benchmarks and MT is somewhat dubious. Is it scheduling issues, which doesn’t effect effect Cinebench2024? 🧐

Mainly in CB 2024, there s power and scores comparisons in the review above, for CB R15/20/23/2024.

 

[GTracing]

I have to think about this some more but I am not so sure the bulk of the efficiency gain is coming from the TMSC process. Architectural gains, especially Skymont 50+% IPC increase allows for relaxed clocks. As we all know backing down clocks even 10 or 15% can provide huge efficiency gains.

Ultimately, I think Intel may have gone TMSC more for transistor density (economics) rather than efficiency. Meaning ARL on Intel 7 would have been physically and economically huge.

In regard to the choice of tiled architecture, I think part of the issue is that they've built out the manufacturing to do tiles, but they have no customers for it. If Intel doesn't make tiled CPUs, they still have the costs from the idle or shuttered production lines.

 

[AcrosTinus]

Doesn't look like it. It appears ARL launch is indeed a disaster. All the reviews so far are quite tragic.



Thats a bold claim considering all the negative reviews on X/Twitter. I hope you can back it up with some actual numbers.

And it appears it's lagging in MT too.

Read the Phoronix Linux bench, the chip is not bad. Windows and the power plans seem to be limiting performance for the sake of lower appearing power consumption.

This generation is for anyone on a 10900K and wants to stay on Intel. You will get better performance in any case, just not class leading performance.

 

[Hitman928]

* The best server product since Rome arrived
* Desktop is a super tiny niche and Intel's gains are comparable to Zen 5 - both are rather small
* GPU has been a standard "Koduri project" - think of AMD Vega
* Foundry remains to be seen
* Laptop is very good and actually is not a niche like desktop or discrete GPUs

Intel is fine

• Intel had the lead in server for a couple of weeks until Turin launched. They're far closer now than before, but it's still going to be a tough task to stop AMD from advancing.
• It's not super tiny niche, that's only super high end desktop systems.
  • Not sure if we're looking at the same charts, but Intel has regressed as much as it has gained in performance with ARL.
• dGPU is on life support. Integrated is looking better but they need some actual volume design wins for that to matter (e.g., steam deck).
• Agreed. Biggest thing is they need real customers which they hope to start to get in 2 years time. It's gonna be rough until then, maybe longer if they can't convince people on 14A.
• They are doing pretty well on laptop, they just need to get back on their own nodes to help with margins.

 

[511]

For Intel Foundry to work it would require Intel to Switch from TSMC to Intel Fabs

 

[AcrosTinus]

* The best server product since Rome arrived
* Desktop is a super tiny niche and Intel's gains are comparable to Zen 5 - both are rather small
* GPU has been a standard "Koduri project" - think of AMD Vega
* Foundry remains to be seen
* Laptop is very good and actually is not a niche like desktop or discrete GPUs

Intel is fine

People forget that we are in a desktop gaming bubble. At work we have approved the new Notebooks for the team, AMD is nowhere to be seen. That is nearly 15K devices on Intel just from a small company under Microsoft. Imagine what other companies are doing....

Intel is fine but they cannot be fine forever

 

[adroc_thurston]

Laptop is very good and actually is not a niche like desktop or discrete GPUs

They're bleeding share in laptop and pretty aggressively so.
If you mean LNL, then it's more niche than discrete GPUs. But all tablet chips are.

 

[Saylick]

People forget that we are in a desktop gaming bubble. At work we have approved the new Notebooks for the team, AMD is nowhere to be seen. That is nearly 15K devices on Intel just from a small company under Microsoft. Imagine what other companies are doing....

Intel is fine but they cannot be fine forever

I've had nothing but Intel-based laptops at work, but then again it's Dell.

 

[Bouowmx]

Has there been tests on tile-to-tile and fabric OC?

 

[AcrosTinus]

I've had nothing but Intel-based laptops at work, but then again it's Dell.

We had Lenovo but Dell Latitudes as well now, they are fine...
One thing they do well is calibrated screen and keyboard, everything else....
The docks break down after 3M to 6M of use, the IT team hates me

 

[AcrosTinus]

View attachment 110225
Has there been tests on tile-to-tile and fabric OC?

Check of derBauer on Youtube, in the last few minutes he talks about that topic with more to come.

 

[yuri69]

• Intel had the lead in server for a couple of weeks until Turin launched. They're far closer now than before, but it's still going to be a tough task to stop AMD from advancing.
• It's not super tiny niche, that's only super high end desktop systems.
  • Not sure if we're looking at the same charts, but Intel has regressed as much as it has gained in performance with ARL.
• dGPU is on life support. Integrated is looking better but they need some actual volume design wins for that to matter (e.g., steam deck).
• Agreed. Biggest thing is they need real customers which they hope to start to get in 2 years time. It's gonna be rough until then, maybe longer if they can't convince people on 14A.
• They are doing pretty well on laptop, they just need to get back on their own nodes to help with margins.

* Intel *still* rides all the OEMs after all those years when AMD offered far more competitive products than now.
* Desktops aka the big boxes sitting at a desk are a niche - even detachable x-in-ones are larger
* dGPUs other than nVidia are on life support in general - AMD's been delaying their next-gen which even got the highend segment canned completely...
* I got no clue about their margins

 

[tsamolotoff]

https://m.alza.cz/intel-core-ultra-9-285k-recenze-a-testy

These guys have some huge issue with their ryzens, I can say that just by looking at their FPS in Starfield and Cyberpunk, my own is like 50% higher

24-10-2024, 20:12:49 Starfield.exe benchmark completed, 3083 frames rendered in 22.000 s
                     Average framerate  :  140.1 FPS
                     Minimum framerate  :  132.1 FPS
                     Maximum framerate  :  154.9 FPS
                     1% low framerate   :  109.3 FPS
                     0.1% low framerate :   99.8 FPS

 

[Abwx]

• Intel had the lead in server for a couple of weeks until Turin launched.

They had the lead because for the Intel review Phoronix removed all benches where Intel was performing unexpectedly bad, so he got an average of 1107 for the Intel chip and 990 for the AMD, but once he added those benches in the following reviews there was no more any lead as the real scores were actualy 1007 vs 1005 for the compared AMD chip.

Intel Xeon 6980P "Granite Rapids" Linux Benchmarks Review - Phoronix

www.phoronix.com

AMD EPYC 9755 / 9575F / 9965 Benchmarks Show Dominating Performance Review - Phoronix

www.phoronix.com

 

[Meteor Late]

I've seen Geekbench 6 is roughly on par with 14900K in single core, correct? this looks at a wide range of workloads so it's more indicative of overall performance instead of Cinebench which is more of less rendering and that's it.

All I miss is Spec, so I can definitely say it's barely an improvement in single core. Hopefully Geekerwan releases a video soon.

 

[cannedlake240]

Foundry will either save Intel in a couple years or completely take the rest of it down with itself. They are sacrificing the whole Intel products side to fund it. Oh they also have to fix the cpu uarch, soc design issues and build a competitive AI offering

 

[yuri69]

MT is okay. ST is very weird tho.

Check the Phoronix review to rule out Micor$oft bugs.

 

[511]

Foundry will either save Intel in a couple years or completely take the rest of it down with itself. They are sacrificing the whole Intel products side to fund it. Oh they also have to fix the cpu uarch, soc design issues and build a competitive AI offering

Their designs were being carried by foundry tbh till 14nm debacle now design needs to perform Hope Glenn Hilton gives us a repeat of Nehlam

 

[maddie]

AMD employs older style chiplet interconnect. Physical wires connecting die-to-die routed thru the substrate itself. Intel employs more advanced packaging (stacking) where the chiplets sit on the interposer which in turn sits on top of the substrate itself. That is the technical distinction between the two imho.

You are a software guy, correct?

 

[Bouowmx]

Check of derBauer on Youtube, in the last few minutes he talks about that topic with more to come.

Interesting DLVR behavior. If DLVR is bypassed (voltage from VRM is not stepped down by DLVR), CPU power consumption is reduced. But this may reduce VRM efficiency because it has to step down more from 12V.

 

[SiliconFly]

.