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.

 

[SiliconFly]

.

 

[Saylick]

Oh my! Any chance, the upcoming magic fix (0x114) update will fix any of these?

Would be nice to see if it did, but then it also implies ARL's launch was rushed and it came out of the oven half-baked.

 

[igor_kavinski]

it came out of the oven half-baked.

RPL refresh did so someone at Intel kindly increased the voltages in firmware to toast it up nice and crispy

 

[igor_kavinski]

Why did it take so long to beat the 4 minute mile and then so many quickly achieved it afterward?

Excellent point. The mental barrier of believing what's possible changes reality. In my C class, the teacher said whoever creates a diamond formation of asterisks using just a single printf is a real problem solver with logical thinking. I was the only one who managed it because I believed someone had done it before. Everyone else dismissed it as impossible.

 

[ondma]

It's a tradeoff and informed customers will make the decision that makes more sense to them.

9600X:

5% faster in games on average.
AVX-512 support.
Drop-in upgrade to X3D in future, saving on the cost of a whole platform upgrade.
Can't think of any serious cons.

245K:

Higher RAM speeds with CUDIMM in future when DDR5-9600 becomes cheaper.
+10% faster in productivity on average.

I/O speed issues with NVMe drives (https://forums.anandtech.com/thread...akes-discussion-threads.2606448/post-41324708)

No upgrade path as far as we know.
No AVX-512 support.
Unattractive for those interested in playing emulated games.

Ehh... I wouldnt buy a six core cpu now from either manufacturer, but especially from intel w/o hyperthreading. 8 cores or bust. And no, I dont count the E cores.

 

[controlflow]

Update 2.2 is live!

Update 2.2 for Cyberpunk 2077 is being rolled out on PC, PlayStation 5 and Xbox Series X|S!

www.cyberpunk.net

CyberPunk update 2.2:
"Optimized threading system to improve performance by up to 33% on Intel Arrow Lake based CPUs."

 

[maddie]

Update 2.2 is live!

Update 2.2 for Cyberpunk 2077 is being rolled out on PC, PlayStation 5 and Xbox Series X|S!

www.cyberpunk.net

CyberPunk update 2.2:
"Optimized threading system to improve performance by up to 33% on Intel Arrow Lake based CPUs."

What does that mean? Up to? Certain sections are 33% better?

 

[controlflow]

What does that mean? Up to? Certain sections are 33% better?

Need to wait for more benchmarks but a user on Overclock.net tried it with the in game benchmark and saw a 13% increase in average FPS and 32% improvement in minimum FPS. Seems to improve FPS in the most compute intensive scenes in the game.

Will be interesting if other games get similar updates to tune for ARL and if the upcoming microcode update further improves gaming performance.

 

[igor_kavinski]

Certain sections are 33% better?

Goes up and down as the threads ping pong between P and E cores

I think the optimization would be that the game engine is now aware which core is P and which one is E and making sure to target E-cores for sustained MT tasks.

 

[OneEng2]

Goes up and down as the threads ping pong between P and E cores

I think the optimization would be that the game engine is now aware which core is P and which one is E and making sure to target E-cores for sustained MT tasks.

I would agree.

This also provides evidence that Lion Cove isn't as useless as some have believed it to be and that Skymont isn't the one-core-to-rule-them-all as believed as well.

 

[SlimFan]

I would agree.

This also provides evidence that Lion Cove isn't as useless as some have believed it to be and that Skymont isn't the one-core-to-rule-them-all as believed as well.

Or it's a software fix that doesn't have to do with avoiding 2/3 of the cores in the system. That isn't as fun to say, though, is it.

 

[LightningZ71]

More specifically, they likely know the cache footprint of their individual game threads and figured out how to schedule the low cache load threads on the E cores and the higher cache loads on the P cores to minimize L3 usage.

 

[cannedlake240]

Skymont isn't the one-core-to-rule-them-all as believed as well

That was a wrong conclusion to draw in the first place. Skm is a smaller core that makes has to make tradeoffs to be area efficient. It's still an E core

 

[ajsdkflsdjfio]

I would agree.

This also provides evidence that Lion Cove isn't as useless as some have believed it to be and that Skymont isn't the one-core-to-rule-them-all as believed as well.

Turning off P-cores and overclocking E-cores actually outperforms default config in some cases. The fact that overclocked E-core only configuration outperforms the stock core configuration is not just a testament to how broken Arrow Lake was at launch but also how good the E-cores are. The fact that you can essentially have a E-core only cpu and have it perform comparably to current generation cpus, in some cases, is insane.

IDK whether this idea would be feasible economically wise/design wise, but I wish Intel would release an E-core only cpu with 12-16 darkmont cores and sell it for like 100-200 bucks and either create a new cheaper arrow lake chipset or better yet somehow engineer the chip to work on older boards like pre-alder lake generations for an easy slot in upgrade. Possibly zen4/alder lake performance at lower power and extremely low die space. If each skymont core is 33% of a lion cove core that means a 12 skymont core cpu would be similar size to a 4-core lion-cove cpu meaning an extremely cheap silicon cost even with darkmont's increased size. It'd be perfect for mini-pcs with crazy levels of power/perf and just in general for people looking for a cheap upgrade from older generations.

Back when the IPC number slides for skymont were first leaked I remember some people on ChipsNCheese's discord doing some very rough calculations on skymont vs golden cove ipc. Apparently their initial calculations put it at like within 5% or something crazy like that using the 1.38x advertised int improvement and using crestmont/gracemont numbers. I'd take this information with a huge grain of salt so i'd say something like 10-20 percent under golden cove to be safe, but still with an ipc improvement with darkmont and clock/performance improvement from 18a. Add improvements from designing a smaller monolithic chip without all the arrowlake/meteor lake chiplet bs and just in general improvements from engineering a chip using only one core and with tiny adjustments to improve darkmont perf for consumer use like how redwood cove in data center is different from redwood cove in meteor lake. Add the fact that skymont can already clock up to 5.2 Ghz stable despite not being designed for pushing performance, it's not crazy to say you could have a 7950x/14900k which uses a third of the power and costs a third of the price(100-200). Even in late 2025 this would be a killer product for a lot of consumers, even me on my 8600k.

https://imgur.com/zn5HsfE

https://imgur.com/SHvyHT7

https://www.bilibili.com/video/BV1tW14YGEmz/

 

[SiliconFly]

.

 

[maddie]

Turning off P-cores and overclocking E-cores actually outperforms default config in some cases. The fact that overclocked E-core only configuration outperforms the stock core configuration is not just a testament to how broken Arrow Lake was at launch but also how good the E-cores are. The fact that you can essentially have a E-core only cpu and have it perform comparably to current generation cpus, in some cases, is insane.

IDK whether this idea would be feasible economically wise/design wise, but I wish Intel would release an E-core only cpu with 12-16 darkmont cores and sell it for like 100-200 bucks and either create a new cheaper arrow lake chipset or better yet somehow engineer the chip to work on older boards like pre-alder lake generations for an easy slot in upgrade. Possibly zen4/alder lake performance at lower power and extremely low die space. If each skymont core is 33% of a lion cove core that means a 12 skymont core cpu would be similar size to a 4-core lion-cove cpu meaning an extremely cheap silicon cost even with darkmont's increased size. It'd be perfect for mini-pcs with crazy levels of power/perf and just in general for people looking for a cheap upgrade from older generations.

https://imgur.com/zn5HsfE

https://imgur.com/SHvyHT7

https://www.bilibili.com/video/BV1tW14YGEmz/

What was the 8P + 0E config? Anyone compared an all P or all E?

 

[ajsdkflsdjfio]

What was the 8P + 0E config? Anyone compared an all P or all E?

It's in the same video. I can't attest to their methodologies or whether this result would be seen in games other than cyberpunk/baldurs gate, but the fact that the result is even attainable at all is crazy.

 

[maddie]

It's in the same video. I can't attest to their methodologies or whether this result would be seen in games other than cyberpunk/baldurs gate, but the fact that the result is even attainable at all is crazy.

Glanced and I didn't see all E. 8P+16E, 1P+16E and 8P. I think the 1P is a hint. Can we solve the puzzle?

 

[csbin]

Where are the performance fixes?

https://twitter.com/x/status/1866793675659022652

 

[511]

Glanced and I didn't see all E. 8P+16E, 1P+16E and 8P. I think the 1P is a hint. Can we solve the puzzle?

1P is required by the Bios you can disable it's use by setting affinity manually if they were smart enough

 

[Hulk]

1P is required by the Bios you can disable it's use by setting affinity manually if they were smart enough

It still manages to sneak in there. When I do testing I set the lone P core to 0.8GHz. It's there but not really having an impact in comparison to the 16 E cores.

 

[Hulk]

Let's do a differential diagnosis on Arrow Lake. Data from TechPowerup review.
Compared to the 14900K this is where the 285K does significantly better than the 14900K.

The AI apps are over performing obviously due to the AI processor on ARL. But how about the other applications? What do each group (the ones that do better than the 14900K and the ones that do worse) have in common with one another?

AI Photo Photo Enhance +45%
AI Stable Diffusion +31%
AI NLP +18%
Blender +17%
COMSOL +16%
AV1 +15%
SHA3 +14%
NAMD +13%
V-Ray +11%
Y-Cruncher +9%

This is where it is significantly behind.
Gaming - Down about 6% on average at 1080p
Powerpoint -27%
WinRaR -24%
Premiere Pro - 18%
7 Zip decompression -18% ("only" down about 6% for compression)
Speedometer -16%
Outlook -15%
JetStream2 -14%
Avast -13%
AES -13%
MySQL -12%
Word --11%
Altium -10%
Excel -9%

Let's assume L3 latency is a bottleneck for ARL. Do the applications in the second group, which are lagging performance-wise, rely more on L3 than the first group?

Powerpoint, WinRaR,a nd Premiere Pro are showing terrible performance in particular. Are they L3 heavy?

 

[adroc_thurston]

Are they L3 heavy?

Yea.

 

[SiliconFly]

.

 

[OneEng2]

That was a wrong conclusion to draw in the first place. Skm is a smaller core that makes has to make tradeoffs to be area efficient. It's still an E core

Some have imagined that Intel should drop Lion Cove completely, and replace P cores with a better Skymont/Darkmont.

I agree that the MUCH smaller footprint for the E-Core. I do agree that the P cores are not efficient in space (compared to Zen 5 for instance), but that doesn't mean that it is possible for the E Core performance to be made to work as well on single threaded applications across the board as well as a P core does.

Turning off P-cores and overclocking E-cores actually outperforms default config in some cases. The fact that overclocked E-core only configuration outperforms the stock core configuration is not just a testament to how broken Arrow Lake was at launch but also how good the E-cores are. The fact that you can essentially have a E-core only cpu and have it perform comparably to current generation cpus, in some cases, is insane.

IDK whether this idea would be feasible economically wise/design wise, but I wish Intel would release an E-core only cpu with 12-16 darkmont cores and sell it for like 100-200 bucks and either create a new cheaper arrow lake chipset or better yet somehow engineer the chip to work on older boards like pre-alder lake generations for an easy slot in upgrade. Possibly zen4/alder lake performance at lower power and extremely low die space. If each skymont core is 33% of a lion cove core that means a 12 skymont core cpu would be similar size to a 4-core lion-cove cpu meaning an extremely cheap silicon cost even with darkmont's increased size. It'd be perfect for mini-pcs with crazy levels of power/perf and just in general for people looking for a cheap upgrade from older generations.

Back when the IPC number slides for skymont were first leaked I remember some people on ChipsNCheese's discord doing some very rough calculations on skymont vs golden cove ipc. Apparently their initial calculations put it at like within 5% or something crazy like that using the 1.38x advertised int improvement and using crestmont/gracemont numbers. I'd take this information with a huge grain of salt so i'd say something like 10-20 percent under golden cove to be safe, but still with an ipc improvement with darkmont and clock/performance improvement from 18a. Add improvements from designing a smaller monolithic chip without all the arrowlake/meteor lake chiplet bs and just in general improvements from engineering a chip using only one core and with tiny adjustments to improve darkmont perf for consumer use like how redwood cove in data center is different from redwood cove in meteor lake. Add the fact that skymont can already clock up to 5.2 Ghz stable despite not being designed for pushing performance, it's not crazy to say you could have a 7950x/14900k which uses a third of the power and costs a third of the price(100-200). Even in late 2025 this would be a killer product for a lot of consumers, even me on my 8600k.

https://imgur.com/zn5HsfE

https://imgur.com/SHvyHT7

https://www.bilibili.com/video/BV1tW14YGEmz/

I guess I just always assume that Intel engineers are very talented. Who would put useless P Cores in a design where they could have just beefed up E Cores instead?

I just can't buy into the line of thinking where the P core is totally useless and that they could have easily just used all Skymont with a little more special sauce to raise the IPC another 10%.

1P is required by the Bios you can disable it's use by setting affinity manually if they were smart enough

Affinity is only a "suggestion" to the OS. I suspect you can't "disable" it like that.