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.

 

[Saylick]

View attachment 133221

https://twitter.com/x/status/1986086104286597269

They claim it's due to the higher heat density that makes it hard to clock high.

 

[gdansk]

You are forgetting 30-40% reduction in power for the same ST which is pretty big this means you can have LNL st from like 15W to 9W that's bigView attachment 133225

I'm not forgetting anything. That's just not relevant to fmax dropping.
PTL should have had that (great) and also more or the same fmax. You know, like the fruity competition which runs away.

 

[511]

I'm not forgetting anything. That's just not relevant to fmax dropping.
PTL should have had that (great) and also more or the same fmax. You know, like the fruity competition which runs away.

it's part of the node though as for Fruit company they have had the room to spare for a while also above screenshot :rofl: don't want such a Fmax increase OEMs cheap out already

 

[DrMrLordX]

What would a top frequency oriented chip get at 115 W?

Depends on how well individual samples scale above a certain voltage. Possibly not well at all, meaning maybe 5150-5200 Mhz? Once you hit " the wall" on a process, the power/voltage requirements go parabolic, and if the transistor density issues are true then things get even worse.

And you think a fmax regression is unworthy of criticism? Or that it simply isn't real?

There's more going on here than can be described with just a chart. We see the SKUs and their max clocks, but what we don't see is the relative distribution of SKUs that you'll actually see on the market. Will the majority of parts be the 356H and 338H and below? It shouldn't inspire much confidence that the top-end 4+4 part is only 4.7 GHz and 4+0 part is 4.8 GHz. There's also way too many 4+8 SKUs, including some like the 386H and 368H that don't make much sense.

 

[Josh128]

So, how can anyone honestly criticize? Panther Lake may be total crap--but you don't know yet. Somehow you think it is worthy of criticism though.

As for is it real or not, the only Intel mobile chip with higher clock speed requires 2x more power for 300 MHz more. And that is only a couple of Intel mobile chips that do it. Most top out at the exact same 5.1 GHz.

Because 18A was supposed to be the next best thing since sliced bread, and what the company was literally bet on, thats why. Everything is worthy of criticism unless you are a fanbot.

I didnt say Panther Lake was crap. I said 18A was not looking great from what we know so far-- and citing iGPU weekly gains, which are based on TSMC 3nm and reliant on Intels extremely poor GPU drivers, which have historically sucked at launch but allow for impressive gains (precisely because they suck so bad at launch), has basically zero relevance to a CPU they produce on their own node, which is Intels expertise and doesnt rely on drivers basically at all.

Top speeds aside, the sheer numbers of SKUs with the same core count speaks volumes about 18A. 6 different SKUs for the same exact same number of cores? I dont know how you choose to see that but it looks to me like they are having to cherry pick bins just to hit 5.1GHz.

4+8+4
388H 5.1
386H 4.9
368H 5.0
366H 4.8
358H 4.8
356H 4.7

Maybe power efficiency will knock it out of the park? They'd better hope it does because what we have seen so far performance wise isnt impressive.

 

[poke01]

let’s just wait for benchmarks but it’s not looking good.

it's part of the node though as for Fruit company they have had the room to spare for a while also above screenshot :rofl: don't want such a Fmax increase OEMs cheap out already

Do they have room to spare or is it because Apples design is better?

I don’t see the C1 Ultra clocking to 4.6GHz and it’s just as wide as the M5. Apple is the only company with a fat core that’s wider and also maintains its high frequency provided cooling is sufficient. Qualcomm needed N3X to clock to 5GHz and they have 40% smaller core than Applws and yet that’s all they could manage.

It’s looks like N2 CPU cores will show who is the best at boost clocks relative to the size of their core and it for sure won’t be Intel as long as the P. Core exists.

 

[dullard]

Because 18A was supposed to be the next best thing since sliced bread, and what the company was literally bet on, thats why.

18AP is the high performance (high frequency) node. Not 18A. So, why would you criticize that the low power node is reaching basically the same frequency that Intel has hit for years? If 18AP can't do any better, then you have a point. But, until then, you are complaining about the wrong node for high frequency.

 

[Meteor Late]

18AP is the high performance (high frequency) node. Not 18A. So, why would you criticize that the low power node is reaching basically the same frequency that Intel has hit for years? If 18AP can't do any better, then you have a point. But, until then, you are complaining about the wrong node for high frequency.

18AP is the high performance node the same way N3P is the high performance node over N3E.
Besides, Intel 4 was also not the high performance node and still clocked at the same 5.1GHz we see on 18A. So does this mean non-high performance node never ever get fmax improvements? they are garbage, then.

 

[Meteor Late]

Also worth noting you get 4.6GHz in all Apple M5 chips, there is no binning. With binning, I'm willing to bet we get 5GHz or very close like 4.9GHz in some of them. So in reality, the practical difference is even less. But Apple has no need to do that yet. We've seen Qualcomm starting to bin and bam, 5GHz.

 

[poke01]

4.6GHz in all Apple M5 chips

In all Macs it’s 4.6GHz. The iPad M5s are binned.
———
like Josh said we might have to wait till reviews come out, maybe it is better in power efficiency.

 

[adroc_thurston]

Also worth noting you get 4.6GHz in all Apple M5 chips, there is no binning

Oh they are binned.

 

[dullard]

18AP is the high performance node the same way N3P is the high performance node over N3E.
Besides, Intel 4 was also not the high performance node and still clocked at the same 5.1GHz we see on 18A. So does this mean non-high performance node never ever get fmax improvements? they are garbage, then.

I'll ask again, what is the 18A Fmax when power isn't limited? How does that compare to Intel's other nodes?

 

[poke01]

I'll ask again, what is the 18A Fmax when power isn't limited? How does that compare to Intel's other nodes?

The problem is 18A is a GAA node. Having the same Fmax as Lunar Lake which was fabbed on the worst N3 node is not good as that was made for lower TDP use cases.

If we compare to Intels previous nodes, Intel 4 also had a maximum frequency of 5.1GHz.

 

[Doug S]

In all Macs it’s 4.6GHz. The iPad M5s are binned.

Maybe. We can't really be certain. They might be forced to limit the max frequency due to power & thermal limits of the iPad Pro form factor - which might be why there have been rumors the next iPad Pro will get a vapor chamber.

If we see other examples of the M5 family running at different frequencies, or the M6 iPad Pro has a lower frequency even with the vapor chamber, only then could we be pretty certain Apple is binning. Personally I would think Apple would be far more likely to bin on power than on frequency. An M5 that functions at 4.4 GHz but not 4.6 at the M5 power/thermal would obviously be able to operate at 4.6 GHz if you gave it more power. If you had M5s that needed that extra juice, you'd set them aside and use them for the Mini, since it is actively cooled and has no battery life concerns.

 

[Doug S]

They claim it's due to the higher heat density that makes it hard to clock high.

Maybe Intel hasn't solved the thermal issues with BSPDN as well they've claimed?

 

[Khato]

All this talk about process being the determining factor of fmax is just painful.

You design for a certain frequency using the PDK. If the process matches the PDK, then you hit the target frequency. If the process doesn't live up to the PDK, then you don't hit the frequency. We don't know whether PTL was designed to hit 5GHz or 5.5GHz.

 

[OneEng2]

The problem is 18A is a GAA node. Having the same Fmax as Lunar Lake which was fabbed on the worst N3 node is not good as that was made for lower TDP use cases.

If we compare to Intels previous nodes, Intel 4 also had a maximum frequency of 5.1GHz.

N3B is a pretty good node. For performance and PPA, it was best in class. Relative to N2 and 18A it's still pretty darned good. I think anyone expecting big gains from going from N3B to 18A will be disappointed.

Maybe Intel hasn't solved the thermal issues with BSPDN as well they've claimed?

That is my concern. If you talk to guys that work in process engineering, the thermal issues are the topic of discussion with BSPDN.... and indirectly, yields at higher frequencies.

 

[eek2121]

And you think a fmax regression is unworthy of criticism? Or that it simply isn't real?

I've said this before, and I'll say it again: Intel is trying to improve both perf/watt and performance. It is a delicate balancing act. They know their chips got WAY out of hand in terms of power consumption compared to AMD, and AMD has been crushing them in perf/watt and also competitive in total performance. What you are seeing is Intel trying to improve both efficiency and performance, while still being competitive. Meteor Lake and Arrow Lake were also steps in that direction.

Further, Intel 18A/AP are capable of pushing out much more efficient designs, but again, Intel has AMD to deal with. They are hoping to maintain rough clock speed parity with AMD while improving efficiency (on mobile at least).

Neither Intel nor AMD consider ARM or Apple to be a serious competitive threat that this time. 9 out of 10 people buy x86 for a reason, so focusing on that 1 person in 10 is counterproductive...for now.

The thing I'm disappointed about? ALL vendors have seemingly given up on 15W designs.

 

[adroc_thurston]

Neither Intel nor AMD consider ARM or Apple to be a serious competitive threat that this time

?

 

[poke01]

Intel is trying to improve both perf/watt and performance

so is everyone else

 

[Geddagod]

Link to this Reddit guy's post please?

I'm assuming he means this guy, perhaps @DavidC1 can clarify

But based on my previous conversations with him, he doesn't really know sh*t either lol.

 

[poke01]

I'm assuming he means this guy, perhaps @DavidC1 can clarify
View attachment 133249
But based on my previous conversations with him, he doesn't really know sh*t either lol.

That’s fake come on

 

[Fjodor2001]

The thing I'm disappointed about? ALL vendors have seemingly given up on 15W designs.

Yes, and it’s strange. Do we know why that is the case? Has e.g. the cooling in laptops improved so they are able to handle 25W instead of 15W while still being compact and with low noise, or is everyone prioritizing perf?

 

[gdansk]

It seems every laptop company has agreed if they have fans they can go over that. Sadly, we have phones that will dissipate that much or more (briefly). There is one vendor still shipping compact fanless laptops in volume but even they have given up on limiting their SoCs to ~15W in fan-cooled devices.

 

[511]

Do they have room to spare or is it because Apples design is better?

Or AMD/Intel puts a lot of xtor for SIMD Apple uses that budget for Integer Performance