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.

 

[techjunkie123]

Geekwans review of arrow lake is up. Chadmont PPC is impressive but doesn't do well in gaming.

 

[511]

Just for reference

 

[GTracing]

Just for reference
View attachment 110997

I was just about to post that, lol. I'm surprised to see Skymont score higher than raptor lake ISO-clock. That's actually insane.

 

[511]

I was just about to post that, lol. I'm surprised to see Skymont score higher than raptor lake ISO-clock. That's actually insane.

Saved you the effort 🙂
Imagine Clearwater forest with 5-7% more powerful core than Skymont core that is not held by fabric and 288 of them on 18A

 

[GTracing]

Saved you the effort 🙂
Imagine Clearwater forest with 5-7% more powerful core than Skymont core that is not held by fabric and 288 of them on 18A

Yeah, Intel might actually beat AMD in server performance again.

I also found this interesting. We've seen other reviewers report on the high memory latency for the p-cores, but I haven't seen memory latency numbers for the e-cores. 194 ns is so high that I wonder if they're doing something wrong with the testing. If it's accurate then it makes the e-cores that much more impressive.

 

[Markfw]

Yeah, Intel might actually beat AMD in server performance again.

I also found this interesting. We've seen other reviewers report on the high memory latency for the p-cores, but I haven't seen memory latency numbers for the e-cores. 194 ns is so high that I wonder if they're doing something wrong with the testing. If it's accurate then it makes the e-cores that much more impressive.
View attachment 111000

Actually, I doubt it. Turin is faster than Intel best desktop, but Turin is another 20-30% faster than its predecessor, due to clock speed PLUS other things.

 

[MS_AT]

I was just about to post that, lol. I'm surprised to see Skymont score higher than raptor lake ISO-clock. That's actually insane.

That is expected, the advantage that Raptor had, wider registers, was taken away, with those compiler settings. For int it is less relevant as x264 will be the only subtest to benefit, but from FP point of view it's a different story.

 

[511]

Actually, I doubt it. Turin is faster than Intel best desktop, but Turin is another 20-30% faster than its predecessor, due to clock speed PLUS other things.

Turin is a Server CPU based on zen 5/5C.It is faster by 20% in phoronix bench for a single CPU vs GNR .
Desktop and Server share the die Geeker wan gave us number for per core performance.Zen 5 is 7% faster than Skymony in SpecInt and Darkmont is 5-7% faster than Skymont with the slow L3 so you can do the math 192C/384T vs 288C/288T both cores having similar SpecInt 1T performance
Also HT increases AMD MT Performance by 30%

 

[MS_AT]

Desktop and Server share the die Geeker wan gave us number for per core performance.Zen 5 is 7% faster than Skymony in SpecInt and Darkmont is 5-7% faster than Skymont with the slow L3 so you can do the math 192C/384T vs 288C/288T both cores having similar SpecInt 1T performance

Which SpecInt? Compiled with gcc or clang, icx or aocc? With or without AVX512? With fancy optimization options enabled or disabled? As a curiosity, from M1 review by Andrei

Since our A14 results, we’ve been able to track down Apple’s compiler setting which increases the 456.hmmer by such a dramatic amount – Apple defaults the “-mllvm -enable-loop-distribute=true” in their newest compiler toolchain whilst it needs to be enabled on third-party LLVM compilers. A 5950X with the flag enabled increases its score to 91.64, but also while seeing some regressions in other tests. We haven’t had time to re-test further platforms.

The baseline otherwise for this subtest was 48.00 for 5950x

 

[Markfw]

Turin is a Server CPU based on zen 5/5C.It is faster by 20% in phoronix bench for a single CPU vs GNR .
Desktop and Server share the die Geeker wan gave us number for per core performance.Zen 5 is 7% faster than Skymony in SpecInt and Darkmont is 5-7% faster than Skymont with the slow L3 so you can do the math 192C/384T vs 288C/288T both cores having similar SpecInt 1T performance
Also HT increases AMD MT Performance by 30%

Where is the benchmark you are referring to ? looking at one small number (specint 1T) does not make a faster cpu overall.

 

[511]

Which SpecInt? Compiled with gcc or clang, icx or aocc? With or without AVX512? With fancy optimization options enabled or disabled? As a curiosity, from M1 review by Andrei

The baseline otherwise for this subtest was 48.00 for 5950x

For this one i am assuming same settings as done by Geekerwan Clang -g -Ofast no fancy optimization.Hope this answers your question 🙂

 

[511]

Where is the benchmark you are referring to ? looking at one small number (specint 1T) does not make a faster cpu overall.

SpecInt Rate is not a single Number tbh it is an aggregate of multiple test as Industry standard as possible it has many substests.
I agree one small number won't make a CPU faster otherwise people won't be buying Sierra Forest it will in the end depends on the workload

 

[gdansk]

Turin is a Server CPU based on zen 5/5C.It is faster by 20% in phoronix bench for a single CPU vs GNR .
Desktop and Server share the die Geeker wan gave us number for per core performance.Zen 5 is 7% faster than Skymony in SpecInt and Darkmont is 5-7% faster than Skymont with the slow L3 so you can do the math 192C/384T vs 288C/288T both cores having similar SpecInt 1T performance
Also HT increases AMD MT Performance by 30%

You're ignoring clock rates
Considerable disappointment follows those extrapolating from only IPC.

 

[511]

You're ignoring clock rates
Considerable disappointment follows those extrapolating from only IPC.

I am not lol i took this graph under consideration but i missed the mesh and imc power and other stuff i have too many variables to account for

 

[gdansk]

I am not lol i took this graph under consideration but i missed the mesh and imc power and other stuff i have too many variables to account for
View attachment 111008

1. Turin-D is on a process with a considerably different v/f curve. The N4P version is still intersecting with N3B Skymont at 2W
2. Turin-D has more than 2W per core vs less than 1.7W per core for Spammontforest.

 

[MS_AT]

I am not lol i took this graph under consideration but i missed the mesh and imc power and other stuff i have too many variables to account for

I wouldn't use HX370-E core to make performance comparisons to Turin-D. Different node, different last level cache, different SIMD backend implementation.

 

[OneEng2]

Saved you the effort 🙂
Imagine Clearwater forest with 5-7% more powerful core than Skymont core that is not held by fabric and 288 of them on 18A

Hopefully we wont have to "Imagine" it .

Turin is a Server CPU based on zen 5/5C.It is faster by 20% in phoronix bench for a single CPU vs GNR .
Desktop and Server share the die Geeker wan gave us number for per core performance.Zen 5 is 7% faster than Skymony in SpecInt and Darkmont is 5-7% faster than Skymont with the slow L3 so you can do the math 192C/384T vs 288C/288T both cores having similar SpecInt 1T performance
Also HT increases AMD MT Performance by 30%

I doubt Clearwater is going to retake the throne in DC from Zen 5 in real world application performance, not to mention Zen 6.

Turnin Zen 5 gets 40% MT improvement with SMT per core (which is flat out crazy IMO). Zen 5 also has AVX512. I don't think you can scale SpecInt into general DC workload performance for Skymont. Also, if Intel isn't shipping Clearwater by Q3 2025, they will soon be facing Zen6 with double the number of cores per CCD on 2nm GAA. So a theoretical AMD Zen 6 384C/768T part is what Skymont would be facing for the majority of its existence.

 

[João Bortolace]

Yeah, Intel might actually beat AMD in server performance again.

I also found this interesting. We've seen other reviewers report on the high memory latency for the p-cores, but I haven't seen memory latency numbers for the e-cores. 194 ns is so high that I wonder if they're doing something wrong with the testing. If it's accurate then it makes the e-cores that much more impressive.
View attachment 111000

Chips and cheese when testing the MTL, he noticed that IMC was sometimes in low power mode when using the E-cores, increasing latency

Update on Meteor Lake DRAM Latency Measurements

Chip makers are always looking to save power, especially in mobile devices.

chipsandcheese.com

 

[511]

1. Turin-D is on a process with a considerably different v/f curve.
2. Turin-D has more than 2W per core vs less than 1.7W per core for Spammontforest.

1. I forgot that but 18A is superior to N3E which Turin D is using by ateast 5%(TSMC Claims on par with N3P) and Intel claims it is on par with N2
2. No arguments here as i said many variables to account for we have imc and IO power in TDP as well

 

[OneEng2]

1. I forgot that but 18A is superior to N3E which Turin D is using by ateast 5%(TSMC Claims on par with N3P) and Intel claims it is on par with N2
2. No arguments here as i said many variables to account for we have imc and IO power in TDP as well

I am guessing that the truth lies somewhere in between. N2>18A>N3P

 

[511]

Turnin Zen 5 gets 40% MT improvement with SMT per core (which is flat out crazy IMO). Zen 5 also has AVX512. I don't think you can scale SpecInt into general DC workload performance for Skymont. Also, if Intel isn't shipping Clearwater by Q3 2025, they will soon be facing Zen6 with double the number of cores per CCD on 2nm GAA. So a theoretical AMD Zen 6 384C/768T part is what Skymont would be facing for the majority of its existence.

I thought it was 30% from the benches done by David Huang.I wasn't talking about AVX-512 Performance cause Skymont will loose there Also 2nm GAA is not shipping before Q4 2026 . Intel has 1 year and than after 9 months if Zen6 we have Arctic wolf with 512C/T rouge river forest 🙂 in 2027

 

[511]

I am guessing that the truth lies somewhere in between. N2>18A>N3P

Yes definitely i agree IEDM is this December both Intel and TSMC will present their respective 2nm class process 18A and N2 we can find an analysis on Semiwiki by than

 

[Hitman928]

1. I forgot that but 18A is superior to N3E which Turin D is using by ateast 5%(TSMC Claims on par with N3P) and Intel claims it is on par with N2
2. No arguments here as i said many variables to account for we have imc and IO power in TDP as well

Were those comments for the original 18a specs or the downgraded ones?

 

[511]

Were those comments for the original 18a specs or the downgraded ones?

I don't know but TSMC knows better so I will take their word for it

 

[Hitman928]

I thought it was 30% from the benches done by David Huang.I wasn't talking about AVX-512 Performance cause Skymont will loose there Also 2nm GAA is not shipping before Q4 2026 . Intel has 1 year and than after 1 year we have Arctic wolf with 512C/T rouge river forest 🙂 in 2027

N2 with GAA goes HVM in H2 2025.

I don't know but TSMC knows better so I will take their word for it

The timing of the comment matters though, if it was before 18a was downgraded, then the comment may no longer apply.