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.

 

[Wolverine2349]

Lets be fair here. Intel 10nm was on par or even better in performance and density metrics than the 7nm TSMC process. AMD had use of tiles/chiplets design (which I will point out that Intel has now finally adopted as well). I have to imagine that they gave up some latency and throughput in order to gain the advantages in yield and cost that the chiplets provide.

In the same generation, AMD started using stacked cache which pretty much took and still today retains the gaming performance crown away from Intel. I note that prior to this, much was always made of gaming benchmarks being the hallmark of processor performance (not sure why myself).

I will agree that Intel processor designs have had an edge in single threaded performance for quite some time; however, they have done so increasingly by sacrificing thermal performance...... so much so that 2 generations of these processors quite literally eat themselves due to Intel flying WAY too close to the design limits.

I'll give you the IPC argument; however, the clock speed was a mistake. It resulted in higher thermal density and the eventual failure of chips in the field, but Intel needed this clock speed in order to compete with the AMD chips that were running at a small fraction of the thermal load.

I'll also give it to you that Luner Lake leaks look good; however, lets not get too carried away here. If it turns out to be true that LLK has better performance than Zen 5, it will have done it due to Intel paying much higher prices for a better process node for the CPU die than AMD is paying for on Zen 5. FWIW, I remain skeptical that LLK is going to "spank" Zen 5 at all. Sure, it will have better multi-thread performance since it has access to more full cores than does Zen 5. Since AMD already has a larger core count version of their Zen 5c CCX on N3E, I wouldn't get too excited about the laptop and desktop LLK besting Zen 5 on a more expensive and more dense process. I certainly wouldn't go all the way to saying Intel has a "better CPU design".

I also find it interesting that suddenly multi-core benchmarks are so important to Intel on the laptop and desktop where before, these were de-emphasized by many (I assume because AMD did better in them). I suspect that AMD will be focusing on their single threaded performance (which still looks good against LLK I believe) and gaming performance once Zen 5 X3D chips come out.

Intel did a great job introducing asymmetric processing with the "big-little" concept. Caught AMD off guard there for sure. Still, one could easily argue that a Zen 5c core is superior to a LLK efficiency core. Intel simply put more E cores on a die than AMD is currently willing to do as it would require AMD pay for a more dense process node as well.

In the server market, AMD will still hold the core advantage with Turin .... and each of the E cores in Turin will have SMT providing an additional 30% performance. I am guessing that Turn will actually out perform Intel's planned 288 core Xeon, but I could be mistaken as many other factors will be in play other than individual core performance.

Does anyone know the relative die size of a N3E Zen 5c to an intel E core on N3E? I would love to look at the performance per die space comparison.

Intel 10nm on par in density to 7nm TSMC. I mean 10nm is bigger. It had better performance due to monolothic die. And Golden Cove whipped Zen 3 on 7nm by 17% better IPC and faster clocks. But more space efficient 10nm was not. And certainly not more power efficient

Well the Zen 5C cores will whip Gracemont e-cores and the only 6% improved Crestmonts. Gracemont e-cores are Skylake level with low 4.X GHz clocks

But Skymont is a massive 37% INT and 68% FP over Crestmont which is already better Gracemont by like 6% and Skymont has Raptor Cove IPC.

SO Skymont will spank Zen 5C easily. Zen 5C is reduced clocks gimped cache and weak Zen 5 IPC which is hardly improvement over Zen 4. And Raptor Cove has better IPC than Zen 4 and 5. And Skymont is supposed to have 2% better IPC than Raptor Cove, It will whip Zen 5C having equal or better clocks in addition to superior IPC.

 

[Wolverine2349]

Much like Intel because AMD hasnt done it doesnt mean they can't. Much like intel they have most likely decided not to for thier own reasons. There is nothing magicaly about doing it with ZEN5c vs Zen5. Their decision is likely all about die size desired for maximizing yields.

The rumored 12 core intel is supposedly for embedded designs from what I read, so likely will be available to the consumer. It may not be the crazy high performance version like RPL is, we will have to wait and see... When was the realease date rumored for this?

WHere is proof that AMD can?

Intel already has actually done it, just not on modern IPC consumer platform nor on a ring bus.

Intel has it on modern server and Workstation Xeon platforms being Saphhire Rapids and Emerald Rapids way more than 8 big cores on a singe tile or mesh which has consistent core to core latency unlike multiple CCDs beyond 8 cores. Yes mesh is not the best for gaming, though maybe mesh overclocked it can be good. But other things that need or desire consistent core to core latency beyond 8 cores. Plus no scheduling issues

AMD does not have more than 8 big cores on a single CCD/node even on the most enterprise EPYC and Threadripper unlike Intel with SPR and Emerald Rapids and beyond.

 

[cannedlake240]

massive 37% INT

Lol no ARL vs RPL it's more like 20-25%. 37% is over LP e Crestmont which massively gimped with half the L2, no SLC and slow memory

 

[SiliconFly]

.

 

[cannedlake240]

Diamond Rapids is LNC

Nah it's renamed PNC or something like that. It's coming out in late 26 why would it use the outdated LNC lol

 

[CouncilorIrissa]

But Skymont is a massive 37% INT and 68% FP over Crestmont which is already better Gracemont by like 6% and Gracemont has Raptor Cove IPC.

Gracemont is nowhere near RPC, what are you smoking lol

According to different sources, it's anywhere from -20% to -17% behind RPC.





Even according to Intel's own slides, it's +38% over LPE core that has (1.26/1.76) = 72% of RPC's PPC)

 

[Wolverine2349]

Gracemont is nowhere near RPC, what are you smoking lol

According to different sources, it's anywhere from -20% to -17% behind RPC.

View attachment 107876
View attachment 107877


Even according to Intel's own slides, it's +38% over LPE core that has (1.26/1.76) = 72% of RPC's PPC)

Typo meant Skymont not Gracemont. Gracemont is way way way behind Raptor Cove big time. Skymont supposed to be equal to Raptor Cove.

 

[511]

Diamond Rapids is LNC. Only Granite Rapid is RWC+.

DMR is Panther/Cougar Cove on 18A that is why it is delayed LNL is designed mainly for N3B LNL/ARL

 

[511]

Gracemont is nowhere near RPC, what are you smoking lol

According to different sources, it's anywhere from -20% to -17% behind RPC.

View attachment 107876
View attachment 107877


Even according to Intel's own slides, it's +38% over LPE core that has (1.26/1.76) = 72% of RPC's PPC)

It is also gimped E core without access to the ring and optimized for low power they showed it is +2% faster than RPC anyways on a bad day it will be slightly slower than RPC which is still plenty fast

 

[OneEng2]

In the laptop and desktop markets, I believe that cost of production is critically important in providing (what I would call) a "good product" for the company.

Just making a good performing processor for these markets isn't the design challenge. It is making a good performing processor that can meet a price point and make a profit.

This is quite a bit easier in the server market in many respects. Where I think the challenge exists in a server processor design is that you can only fit so much processing power into an existing power envelope. While there is certainly still a cost aspect to these products, I think there is much more wiggle room on profit margin in this market.

It is my thinking that Intel has bridged the gap with "tiles". Intel still appears to lag in 3D cache stacking. Both AMD and Intel appear to have equivalent "glue" technology to tie all these things together.

AMD has the lead in graphics GPU design, and also AI processors (although they still lag NVIDIA in the latter).

It appears (from discussions here) that in 2025, Turin (and variants) will go head-to-head with Granite Rapids and Sierra Forest. Since both AMD and Intel will be using (I believe) TSMC's process nodes for this, I would expect that they have the same potential yield capabilities (although I am sure that processor design likely has a hand in determining yield as well).

Depending on how this all pans out with respect to performance and availability, I would expect Intel to at least staunch the bleeding in this market.

I am a little less certain of the laptop and desktop market at this time though. I will have to wait for the actual reviews (vs leaks) to have a better feel for it.

 

[9949asd]

 

[SiliconFly]

.

 

[jdubs03]

Then whats the Intel server CPU thats coming out in 2025 H2?

Pretty sure Clearwater Forest-Darkmont is planned to release then.

 

[cannedlake240]

Then whats the Intel server CPU thats coming out in 2025 H2?

Clearwater is next year that slots into the big socket used by both granite and SRF. Pat G recently said the next P core Xeon still hasn't even taped in on 18A yet, clearly it isn't coming out in 2025. Plus DMR was likely upgraded to a newer core from LNC, no wonder it's launching later

 

[SiliconFly]

.

 

[511]

In the laptop and desktop markets, I believe that cost of production is critically important in providing (what I would call) a "good product" for the company.

Just making a good performing processor for these markets isn't the design challenge. It is making a good performing processor that can meet a price point and make a profit.

This is quite a bit easier in the server market in many respects. Where I think the challenge exists in a server processor design is that you can only fit so much processing power into an existing power envelope. While there is certainly still a cost aspect to these products, I think there is much more wiggle room on profit margin in this market.

It is my thinking that Intel has bridged the gap with "tiles". Intel still appears to lag in 3D cache stacking. Both AMD and Intel appear to have equivalent "glue" technology to tie all these things together.

AMD doesn't have Glue it's all TSMC while Intel's glue is different just to be clear Intel basically have two glues they can pick whichever one

AMD has the lead in graphics GPU design, and also AI processors (although they still lag NVIDIA in the latter).

AMD has good GPU designs but the thing is they lack in Software Something Intel and Nvidia has coutside of game drivers intel arc is superior in many cases to AMD Intel has extensive Software support in terms of OneAPI

It appears (from discussions here) that in 2025, Turin (and variants) will go head-to-head with Granite Rapids and Sierra Forest. Since both AMD and Intel will be using (I believe) TSMC's process nodes for this, I would expect that they have the same potential yield capabilities (although I am sure that processor design likely has a hand in determining yield as well).

Intel won't use TSMC for datacenter for majority for the foreseeable future all DC EMR SPR SRF GNR CLWF DMR is Intel yields depends on Process and die size packing also has yield if you don't do packing correctly you are done for

Depending on how this all pans out with respect to performance and availability, I would expect Intel to at least staunch the bleeding in this market.

I am a little less certain of the laptop and desktop market at this time though. I will have to wait for the actual reviews (vs leaks) to have a better feel for it.

Yeah we need to see more

 

[Wolverine2349]

AMD doesn't have Glue it's all TSMC while Intel's glue is different just to be clear Intel basically have two glues they can pick whichever one.

Referring to glue? Like how are things glued and how intel does it.

I thought a process node was all an SOC and no glue?

I mean technically didn't Intel glue two Pentium 4 chips into one package to make the Pentium D and thus they were not a true dual core but rather a dual socket Pentium 4s in one package. Also kind of like how AMD currently glues 2 5800X/7700X/5600X/7600X/9700X/9600X CCDs into one paclkage chip to make the current Ryzen 9 parts??

 

[jdubs03]

Figured I’d update the thread with a couple recent GB6 scores for the 288V. Looks like it’ll be about the same as the HX 370 (Zenbook S 16) in ST.

Pretty much confirms 2800-2880s ST, and right around 11000 MT.

 

[Wolverine2349]

ARL can turn off p core now, there is a 12 e core only cpuz score.View attachment 107654

I have a spare 12600K in a backup system.

I just ran CPUZ 2.11 benchmark the default one with default settings just e-cores off in BIOS

Single core score was 791 and clock speed was 4.85GHz.

Ran it a couple other times and very close to same score of 790 to 793 with 4.85GHz clock speed single core.

So given that the Core Ultra 265K Skymont e-cores go up to 4.6GHz per its specs: https://www.techpowerup.com/cpu-specs/core-ultra-7-265k.c3776

Assuming in this screenshot it was running at 4.6GHz for the single core score, it has an IPC of about 4-6% lower than Golden Cove.

That is still dang impressive. Not quite Raptor Cove IPC, but Chips and Cheese details and states its IPC is very close to Redwood Cove under Final Words: https://chipsandcheese.com/2024/06/15/intel-details-skymont/

And Redwood COve has like 8-9% worse IPC than Raptor Cove per this:

Intel Meteor Lake P-cores Show IPC Regression Over Raptor Lake?

Intel Core Ultra "Meteor Lake" mobile processor may be the the company's most efficient, but isn't a generation ahead of the 13th Gen Core "Raptor Lake" mobile processors in terms of performance. This isn't just because it has an overall lower CPU core count in its H-segment of SKUs, but also...

www.techpowerup.com

And since Raptor Cove is 5% better than Golden Cove and Redwood Cove 8-9% behind Raptor and thus a few percent behind Goklden Cove, appears Skymont per my Golden Cove test of 12600K and SKymont screenshot accurate, appears Skymont has within 1-3% of Redwood Cove IPC aftercall.

Once again not quite as good as the +2% better than Raptor Cove that it promises, but still darn impressive and appears Chips and Cheese is right. Though it varies by workload a lot I imagine and maybe best case is 2% better than Raptor Cove with all around performance being Redwood Cove IPC.

I wonder what the clock speeds are running all 12 cores

The Ryzen 5900X also 12 cores scores 9471. It has SMT though. SO wonder how much uplift SMT gives it giving it a higher score than the 8983 12 Skymont scores in the screenshot leaked.

What do you think?

 

[alcoholbob]

I have a spare 12600K in a backup system.

I just ran CPUZ 2.11 benchmark the default one with default settings just e-cores off in BIOS

Single core score was 791 and clock speed was 4.85GHz.

Ran it a couple other times and very close to same score of 790 to 793 with 4.85GHz clock speed single core.

So given that the Core Ultra 265K Skymont e-cores go up to 4.6GHz per its specs: https://www.techpowerup.com/cpu-specs/core-ultra-7-265k.c3776

Assuming in this screenshot it was running at 4.6GHz for the single core score, it has an IPC of about 4-6% lower than Golden Cove.

That is still dang impressive. Not quite Raptor Cove IPC, but Chips and Cheese details and states its IPC is very close to Redwood Cove under Final Words: https://chipsandcheese.com/2024/06/15/intel-details-skymont/

And Redwood COve has like 8-9% worse IPC than Raptor Cove per this:

Intel Meteor Lake P-cores Show IPC Regression Over Raptor Lake?

Intel Core Ultra "Meteor Lake" mobile processor may be the the company's most efficient, but isn't a generation ahead of the 13th Gen Core "Raptor Lake" mobile processors in terms of performance. This isn't just because it has an overall lower CPU core count in its H-segment of SKUs, but also...

www.techpowerup.com

And since Raptor Cove is 5% better than Golden Cove and Redwood Cove 8-9% behind Raptor and thus a few percent behind Goklden Cove, appears Skymont per my Golden Cove test of 12600K and SKymont screenshot accurate, appears Skymont has within 1-3% of Redwood Cove IPC aftercall.

Once again not quite as good as the +2% better than Raptor Cove that it promises, but still darn impressive and appears Chips and Cheese is right. Though it varies by workload a lot I imagine and maybe best case is 2% better than Raptor Cove with all around performance being Redwood Cove IPC.

I wonder what the clock speeds are running all 12 cores

The Ryzen 5900X also 12 cores scores 9471. It has SMT though. SO wonder how much uplift SMT gives it giving it a higher score than the 8983 12 Skymont scores in the screenshot leaked.

What do you think?

The other possibility is Arrow Lake just sucks at CPU-Z for whatever reason, since the P-core leaks have been pretty underwhelming.

 

[511]

Referring to glue? Like how are things glued and how intel does it.

There are multiple ways to glue what AMD use is called Hybrid Bonding for 3D Vcache Intel has Foveros where they take a die and stack multiple on top of it a good article on it

Advanced Packaging Part 2 – Review Of Options/Use From Intel, TSMC, Samsung, AMD, ASE, Sony, Micron, SKHynix, YMTC, Tesla, and Nvidia

Advanced packaging exists on a continuum of cost and throughput vs performance and density. In the first part of the series, we spoke through the need for advanced packaging. Even though the demand…

www.semianalysis.com

I thought a process node was all an SOC and no glue?

I mean technically didn't Intel glue two Pentium 4 chips into one package to make the Pentium D and thus they were not a true dual core but rather a dual socket Pentium 4s in one package. Also kind of like how AMD currently glues 2 5800X/7700X/5600X/7600X/9700X/9600X CCDs into one paclkage chip to make the current Ryzen 9 parts??

AMD uses standard Packaging like intel Used for their mobile cpu to put pch on same package a cheap version Foveros is expensive

 

[Wolverine2349]

UpdateL

12600K CPUZ Multi thread score:

With e-cores off and HT on 4.5GHz speed: 5581

With e-cores off and HT off 4.5GHz speed: 4344

So given that half the core count Golden Cove IPC full multi thread test.

I fyou could double it if a 12 P core existed.

So 4344 X 2 is 8688, so less than half of 8983. So in that case maybe full Raptor IPC or better. Not sure what clocks they have or maybe 4.6GHz is all and single e-core clock for Skymont?

 

[Wolverine2349]

There are multiple ways to glue what AMD use is called Hybrid Bonding for 3D Vcache Intel has Foveros where they take a die and stack multiple on top of it a good article on it

Advanced Packaging Part 2 – Review Of Options/Use From Intel, TSMC, Samsung, AMD, ASE, Sony, Micron, SKHynix, YMTC, Tesla, and Nvidia

Advanced packaging exists on a continuum of cost and throughput vs performance and density. In the first part of the series, we spoke through the need for advanced packaging. Even though the demand…

www.semianalysis.com

AMD uses standard Packaging like intel Used for their mobile cpu to put pch on same package a cheap version Foveros is expensive

So is Intel gluing the PCH to Arrow Lake? I think on Alder and Raptor the PCH was on the mobo correct? Its much better for latency that it is on CPU?

Or was the PCH on CPU on Alder and Raptor or at least part of it?

 

[jdubs03]

UpdateL

12600K CPUZ Multi thread score:

With e-cores off and HT on 4.5GHz speed: 5581

With e-cores off and HT off 4.5GHz speed: 4344

So given that half the core count Golden Cove IPC full multi thread test.

I fyou could double it if a 12 P core existed.

So 4344 X 2 is 8688, so less than half of 8983. So in that case maybe full Raptor IPC or better. Not sure what clocks they have or maybe 4.6GHz is all and single e-core clock for Skymont?

This schematic will do, but yup 4.6GHz all-core for Skymont in the unlocked SKUs:

 

[9949asd]

I have a spare 12600K in a backup system.

I just ran CPUZ 2.11 benchmark the default one with default settings just e-cores off in BIOS

Single core score was 791 and clock speed was 4.85GHz.

Ran it a couple other times and very close to same score of 790 to 793 with 4.85GHz clock speed single core.

So given that the Core Ultra 265K Skymont e-cores go up to 4.6GHz per its specs: https://www.techpowerup.com/cpu-specs/core-ultra-7-265k.c3776

Assuming in this screenshot it was running at 4.6GHz for the single core score, it has an IPC of about 4-6% lower than Golden Cove.

That is still dang impressive. Not quite Raptor Cove IPC, but Chips and Cheese details and states its IPC is very close to Redwood Cove under Final Words: https://chipsandcheese.com/2024/06/15/intel-details-skymont/

And Redwood COve has like 8-9% worse IPC than Raptor Cove per this:

Intel Meteor Lake P-cores Show IPC Regression Over Raptor Lake?

Intel Core Ultra "Meteor Lake" mobile processor may be the the company's most efficient, but isn't a generation ahead of the 13th Gen Core "Raptor Lake" mobile processors in terms of performance. This isn't just because it has an overall lower CPU core count in its H-segment of SKUs, but also...

www.techpowerup.com

And since Raptor Cove is 5% better than Golden Cove and Redwood Cove 8-9% behind Raptor and thus a few percent behind Goklden Cove, appears Skymont per my Golden Cove test of 12600K and SKymont screenshot accurate, appears Skymont has within 1-3% of Redwood Cove IPC aftercall.

Once again not quite as good as the +2% better than Raptor Cove that it promises, but still darn impressive and appears Chips and Cheese is right. Though it varies by workload a lot I imagine and maybe best case is 2% better than Raptor Cove with all around performance being Redwood Cove IPC.

I wonder what the clock speeds are running all 12 cores

The Ryzen 5900X also 12 cores scores 9471. It has SMT though. SO wonder how much uplift SMT gives it giving it a higher score than the 8983 12 Skymont scores in the screenshot leaked.

What do you think?

16 skymont almost equals to a 5950x.