Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads

[Tigerick]

Wildcat Lake (WCL) Specs

Intel Wildcat Lake (WCL) is upcoming mobile SoC replacing Raptor Lake-U. 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 Q1 2026.

		Intel Raptor Lake U	Intel Wildcat Lake 15W?	Intel Lunar Lake		Intel Panther Lake 4+0+4
Launch Date		Q1-2024	Q2-2026	Q3-2024		Q1-2026
Model		Intel 150U	Intel Core 7	Core Ultra 7 268V		Core Ultra 7 365
Dies		2	2	2		3
Node		Intel 7 + ?	Intel 18-A + TSMC N6	TSMC N3B + N6		Intel 18-A + Intel 3 + TSMC N6
CPU		2 P-core + 8 E-cores	2 P-core + 4 LP E-cores	4 P-core + 4 LP E-cores		4 P-core + 4 LP E-cores
Threads		12	6	8		8
Max Clock		5.4 GHz	?	5 GHz		4.8 GHz
L3 Cache		12 MB		12 MB		12 MB
TDP		15 - 55 W	15 W ?	17 - 37 W		25 - 55 W
Memory		128-bit LPDDR5-5200	64-bit LPDDR5	128-bit LPDDR5x-8533		128-bit LPDDR5x-7467
Size		96 GB		32 GB		128 GB
Bandwidth				136 GB/s
GPU		Intel Graphics	Intel Graphics	Arc 140V		Intel Graphics
RT		No	No	YES		YES
EU / Xe		96 EU	2 Xe	8 Xe		4 Xe
Max Clock		1.3 GHz	?	2 GHz		2.5 GHz
NPU		GNA 3.0	18 TOPS	48 TOPS		49 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.

 

[coercitiv]

100% slower to prevgens

It was at this moment he achieved perfect stillness.

 

[Hulk]

Wouldn't Panther Lake pricing be based on what the system integrators think the market will bear? If they are charging a lot then they probably think they can sell it for a lot?
I'm a simple man, just trying to think nuts and bolts.

PTL has higher IPC than Zen5/5c:

https://wccftech.com/intel-panther-lake-powered-by-the-new-cougar-cove-p-cores-and-darkmont-e-cores-takes-a-lead-over-amd/

View attachment 137542

Why does Zen 5c have lower IPC than 5? In relation to it's core frequency memory access is faster for 5c. They should be closer I think?

Why is Darkmont LPE so much behind Darkmont E? I realize LPE is not on the ring but I don't think that matters much in this bench?

 

[coercitiv]

Wouldn't Panther Lake pricing be based on what the system integrators think the market will bear? If they are charging a lot then they probably think they can sell it for a lot?

System integrators will price relative to BoM and how much they believe they can differentiate their devices (so not much).

I expect sales for premium laptops and premium CPU models will suffer greatly this year. The word on the street is prices for laptops are going up 20%+ from RAM/NAND alone, so expensive CPUs are going to have a very rought time. PTL better have decent manufacturing costs or else.

 

[RTX2080]

Wouldn't Panther Lake pricing be based on what the system integrators think the market will bear? If they are charging a lot then they probably think they can sell it for a lot?
I'm a simple man, just trying to think nuts and bolts.

It's too expensive. Some low to mid tier OEMs cannot afford the PantherLake pricing, they give up and are going to do StrixPoint or ArrowLake-HX instead. The PantherLake cost is about equals to ArrowLake-HX+GB203 dGPUs. They have to lower the price, maybe just need to wait until Q4.

 

[511]

PTL has higher IPC than Zen5/5c:

https://wccftech.com/intel-panther-lake-powered-by-the-new-cougar-cove-p-cores-and-darkmont-e-cores-takes-a-lead-over-amd/

View attachment 137542

This is mobile Zen 5 though

 

[511]

It's too expensive. Some low to mid tier OEMs cannot afford the PantherLake pricing, they give up and are going to do StrixPoint or ArrowLake-HX instead. The PantherLake cost is about equals to ArrowLake-HX+GB203 dGPUs. They have to lower the price, maybe just need to wait until Q4.

ARL-HX is just way expensive compared to Panther Lake if you look at the die size and you are adding like 200-300+mm2 of N4 Silicon I Don't think that price is correct.
I think the non X SKU with 4+8+4 should be similar price to Arrow Lake H.

 

[Kepler_L2]

Why does Zen 5c have lower IPC than 5? In relation to it's core frequency memory access is faster for 5c. They should be closer I think?

The Zen5c CCX has less L3 than the Zen5 CCX. FPU on Zen5c is also slightly nerfed.

Why is Darkmont LPE so much behind Darkmont E? I realize LPE is not on the ring but I don't think that matters much in this bench?

No L3 cache

 

[Joe NYC]

System integrators will price relative to BoM and how much they believe they can differentiate their devices (so not much).

I expect sales for premium laptops and premium CPU models will suffer greatly this year. The word on the street is prices for laptops are going up 20%+ from RAM/NAND alone, so expensive CPUs are going to have a very rought time. PTL better have decent manufacturing costs or else.

There is some funny math that Intel is using and planning to use. Numbers are made up just to illustrate a point:

Let's say Raptor has Foundry cost of $100, but Foundry sells it to Products for $50.
The big foundry turnaround will be to make 18A products that will cost $150 to make and Foundry will sell it to Products for $200.

Great turnaround for Foundry, but then how is Products going to sell Panther Lake that has a cost of $200?

As part of this re-alignment, Intel Products is in middle of several rounds of price increases. Let's see how it works out, how market is going to absorb the price increases.

 

[511]

There is some funny math that Intel is using and planning to use. Numbers are made up just to illustrate a point:

Let's say Raptor has Foundry cost of $100, but Foundry sells it to Products for $50.
The big foundry turnaround will be to make 18A products that will cost $150 to make and Foundry will sell it to Products for $200.

No? The problem is foundry was charging market prices but the cost to make is high but with 18A/Intel 3 it's going to be fixed they will charge the market price and Their cost structure is not that horrendous for these node but they have to pay ramp cost of the node which is expensive for every new node.

You can see DCAI Having better margins part of is due to Intel 3.

Great turnaround for Foundry, but then how is Products going to sell Panther Lake that has a cost of $200?

As part of this re-alignment, Intel Products is in middle of several rounds of price increases. Let's see how it works out, how market is going to absorb the price increases.

It's upto product to make sure they can actually make it good to sell it ARL/MTL had design issue and IFS Only handled the packing part if the design F*** up it's up to them not IFS to make competitive product why do you think NVL is N2 cause Intel believes that going external is better than 18AP for the top SKU and now the mid SKU.

 

[jdubs03]

It's upto product to make sure they can actually make it good to sell it ARL/MTL had design issue and IFS Only handled the packing part if the design F*** up it's up to them not IFS to make competitive product why do you think NVL is N2 cause Intel believes that going external is better than 18AP for the top SKU and now the mid SKU.

Doesn’t inspire a lot of confidence. Particularly for 18AP, which was supposed to unlock the higher frequency. Seems like IFS integrity and future prospects relies on 14A.

 

[DavidC1]

Wouldn't Panther Lake pricing be based on what the system integrators think the market will bear? If they are charging a lot then they probably think they can sell it for a lot?
I'm a simple man, just trying to think nuts and bolts.

Yes, but they have to bear the increased storage and RAM pricing anyway. They have little room to adjust in that regard, other than doing the usual offer low amount of RAM and storage and overcharge on upgrades on them.

Prices have a nonlinear impact with volume. 20% increase in price is greater than 20% difference in volume. A laptop at $500 probably sells at 5x the volume of a $1000 one.

It's too expensive. Some low to mid tier OEMs cannot afford the PantherLake pricing, they give up and are going to do StrixPoint or ArrowLake-HX instead. The PantherLake cost is about equals to ArrowLake-HX+GB203 dGPUs. They have to lower the price, maybe just need to wait until Q4.

Prices are pretty much same between Gorgon Point and Pantherlake going by Lenovo Legion pricing.

Also, another bad timing by Intel discontinuing 12th Gen CPUs(DDR4), unless they believe the AI bubble will pop soon.

No L3 cache

Yea Darkmont has to rely on only L2 and the uselessly slow Memory Side Cache(which isn't so useless on the power efficiency part).

 

[511]

Doesn’t inspire a lot of confidence. Particularly for 18AP, which was supposed to unlock the higher frequency. Seems like IFS integrity and future prospects relies on 14A.

It's a product decision and don't know when that was made but 18AP was planned for 4+8 SKU but it got moved to N2 some time

 

[Hulk]

Prices have a nonlinear impact with volume. 20% increase in price is greater than 20% difference in volume. A laptop at $500 probably sells at 5x the volume of a $1000 one.

That is a great point. Thanks for taking the time to respond.

 

[DavidC1]

You know they were talking about prioritizing using the E cores so the iGPU can have more power allocated to it.

It doesn't look like it's doing that so far? Potentially hints at improvements coming.

 

[fastandfurious6]

PTL has higher IPC than Zen5/5c:

https://wccftech.com/intel-panther-lake-powered-by-the-new-cougar-cove-p-cores-and-darkmont-e-cores-takes-a-lead-over-amd/

View attachment 137542

there is no universe where the above is true

ARL-H is a terrible product, slower than zen5 in everything. How does it have higher IPC than Zen5 😂

It was at this moment he achieved perfect stillness.

when in market prices HALO 392 has same price to cheapest model of flagship PTL, there is zero reason to buy PTL

if you present valid counterargument to this I'll agree with you

 

[MoistOintment]

ARL-H is a terrible product, slower than zen5 in everything. How does it have higher IPC than Zen5 😂

Because IPC is not PPC. You can have other issues, like high latency, impacting latency sensitive tasks

 

[Josh128]

Wouldn't Panther Lake pricing be based on what the system integrators think the market will bear? If they are charging a lot then they probably think they can sell it for a lot?
I'm a simple man, just trying to think nuts and bolts.

Why does Zen 5c have lower IPC than 5? In relation to it's core frequency memory access is faster for 5c. They should be closer I think?

Why is Darkmont LPE so much behind Darkmont E? I realize LPE is not on the ring but I don't think that matters much in this bench?

This IPC figure is for integer only. Zen 5 loses handily there now, but likely still has the edge in floating point ops, at least against Darkmont. Surprised we havent seen SPECfp figures yet, but maybe thats by design...

 

[Hulk]

Here is the SPEC FP numbers I grabbed from the video.

 

[Josh128]

Here is the SPEC FP numbers I grabbed from the video.
View attachment 137566

Looks like I guessed right. So in fp Cougar Cove is equal to Zen5, slightly behind Zen 5c, and Darkmont is still far behind both. Strange that Zen 5c is showing better fp/GHz than Zen 5 though, that makes no sense.

 

[gdansk]

Here is the SPEC FP numbers I grabbed from the video.
View attachment 137566

Also a good example why rate/GHz is idiotic. Behold Zen 5c, the best FP monster!!

 

[Geddagod]

Looks like I guessed right. So in fp Cougar Cove is equal to Zen5, slightly behind Zen 5c, and Darkmont is still far behind both. Strange that Zen 5c is showing better fp/GHz than Zen 5 though, that makes no sense.

It's because Zen 5C is clocked lower, and ipc generally decreases as clock speed increases, esp if u are memory bound.

 

[Hulk]

Looks like I guessed right. So in fp Cougar Cove is equal to Zen5, slightly behind Zen 5c, and Darkmont is still far behind both. Strange that Zen 5c is showing better fp/GHz than Zen 5 though, that makes no sense.

If I had to take a shot at this I'd explain it like this.

5>5c INT is latency bound, access to L3 is very important. The smaller L3 of 5c is a problem as the working set does not fit in the 5c L3 as well as the Zen 5 L3. Even though 5c can get to main memory in less clock cycles (lower core frequency) it's not enough to offset Zen 5 have more of the data "right there" in the L3.

5c>5 FP is bandwidth bound, access to main memory is very important. The working set doesn't fit well in the L3 for Zen 5 or 5c and there are a lot of calls to main memory. 5c with it's lower core frequency doesn't waste as many cycles waiting on main memory.

That's my intuition and it is completely unsupported, I'm just trying to explain the data!

Now the Big Brains here and take me apart!

BTW, that plus or minus 5% performance is about what we used to see when Intel would improve memory access back in the tick/tock days when the process shrink might come with some memory subsytem improvements, or the difference in performance between desktop and laptop CPU's due to desktop having faster memory subsystem.

 

[Hulk]

Napkin math time.
Converted FP and INT cores to an equal weight basis where "1.0" is the top score. Added those two values to find overall "throughput/cycle."

Could we, for the most part, agree on these numbers? Or should there be a different weighting between FP and INT?

Zen 5 and Lion Cove being within 1% of one another is pretty realistic overall, I mean outside of gaming, which I realize is a huge caveat.

Also this calculation shows both Cougar Cove and Darkmont gaining ~4% on the predecessors.

Finally the bandwidth (for FP) vs. latency (for INT) explanation seems to be supported by the fact that Darkmont LPE holds up much better compared to Darkmont E in FP.

Overall Throughput/Cycle
Cougar Cove	100%	1.95	0.0%
Zen 5c	96%	1.88	-3.8%
Zen 5	95%	1.86	-4.6%
Lion Cove	94%	1.84	-5.9%
Darkmont E	92%	1.80	-7.6%
Skymont	88%	1.72	-11.6%
Darkmont LPE	86%	1.68	-14.0%

 

[DrMrLordX]

when in market prices HALO 392 has same price to cheapest model of flagship PTL, there is zero reason to buy PTL

if you present valid counterargument to this I'll agree with you

Battery life might be pretty good on Panther Lake. That's about it. Intel shouldn't be charging THAT MUCH just for some extra battery life though.

 

[OldGamersNeverDie]

I am not seeing Halo at the same price as Panther Lake.

The Asus Tuf laptop with the 392 is $1800

MSI Prestige with Panther Lake 385h is $1300.

Both are 14 inch, same ram, same SSD. Asus has a higher res screreen (2560x1600) with high framerate, but MSI prestige is OLED. Prestige is also lighter (1.36 kg vs 1.48)