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.

 

[DavidC1]

They gotta win over AMD at some metric at least.

Please tell me you are joking. Both AMD and Intel knows graphics are actually important, and it is. And your statement supports my statement. It does sell.

I bought 2600K(I was a fool). If I didn't need to I would have got a 2500K and saved $100. So the iGPU cost me $100 more. And I saw people having highest Iris configs, which were extra $100 over non-Iris chips but exactly the same CPU. And Intel said segmentation allowed record revenues. Their revenues were stagnant for 5-6 years until Core branding and increased 20% per year for 1st Gen and 2nd Gen. They said back then Core i3 was their biggest seller. Not Pentium, not Celeron.

If it has "miserable performance" would you want one for gaming at all? The lowest AMD could go is an 8 core with the full Halo iGPU. That would save them a few bucks because they could use a dummy CCD but it would still be an expensive die with very little volume, which means it would still cost a lot. Until peoples mindsets change and green sticker important are no longer things this won't change.

That's why you want it paired with a low end CPU and save money, because it would be justified.. Core Ultra 3 with 12 Xe3 graphics. Having only Ultra 9 with 12 Xe3 are marketing tricks. Deep down people know this. The iGPUs are in reality not free because of this reason. They are fleecing you like sheep and every time they come back and say "Do it again, PLEASE! Fleece me over and over and over again!".

Sandy Bridge with 12 EU only for i7. Same with Ivy Bridge. Haswell with Iris Pro only for i7. That trend continues today. At least until Raptorlake you got your best non-halo iGPU on lowest parts. With Meteorlake forwards you need to spend money on Ultra 7 or 9 for the miserly iGPU.

The lowest AMD could go is an 8 core with the full Halo iGPU.

Then it would at least be $100 cheaper than today, which is not an insignificant thing, because if your primary concern is the graphics then it's just $100 savings for you. This would actually propagate halo iGPUs faster rather than relegating it to Apple pricing all the time.

You wouldn't be saving $100 because Intel/AMD saves $20 on BoM like @LightningZ71 is trying to say. My point is, because now they have to brand 3 or 5 instead of 7/9 branding, they have to lower it by $100.

 

[LightningZ71]

They don't HAVE to do anything! They want to cover their costs and then turn a profit. Your proposal for PTL makes a more expensive product that they can only charge LESS for! It just doesn't work. At least for Strix Halo, AMD could make a single CCD with max iGPU config and be able to charge marginally less for it because all it would need is a dummy die with no change in production step change.

 

[Khato]

There should be an alternate configuration where a low end CPU is paired with the best GPU. It's not like even 12 Xe3 is going to play anything much more than low-med settings for most people anyway.

It is interesting that even Intel seems to be acknowledging that fact with respect to gaming. In that part of the improved PTL gaming performance comes from changing scheduling priority from the P cores to the E cores when GPU bound regardless. So yeah, for gaming purposes they certainly could offer a lower spec CPU with the 12Xe iGPU... but they won't. Better to tie them together to maximize profits.

 

[511]

IIt is interesting that even Intel seems to be acknowledging that fact with respect to gaming. In that part of the improved PTL gaming performance comes from changing scheduling priority from the P cores to the E cores when GPU bound regardless. So yeah, for gaming purposes they certainly could offer a lower spec CPU with the 12Xe iGPU... but they won't. Better to tie them together to maximize profits.

the scheduling priority is simple reason at low power levels E cores provide better perf

 

[Doug S]

I speak for the market.

No, you don't.

The market OVERWHELMINGLY wants iGPUs not dGPUs. Why are you continuing to beat a dead horse insisting that the "no one wants iGPUs" when that's 1) clearly not true, 2) dGPUs are a niche market, and 3) a shrinking niche at that.

dGPUs will never go away, but the overwhelming majority of people are not doing anything that a only a dGPU can handle. I'm not sure why you continue spouting bs that you and everyone else reading it knows is absolutely false. Are you trolling, or do you have a brain worm?

Now there are reasonable arguments to be made about the level of demand for "high performance" iGPUs. You don't need AMD Halo levels of iGPU to satisfy that overwhelming majority of people who aren't doing anything that requires a dGPU. The number of people who go from "I need a dGPU" to "an iGPU is fine" with a doubling of iGPU performance isn't that large, so it is questionable whether such a market is viable enough to be profitable for AMD/Intel. The customers who have always used dGPUs in the past, even if a halo level iGPU could handle their CURRENT needs, realize they may not have the same level of needs a few years down the road so a "better" iGPU is not a good option for them.

 

[adroc_thurston]

The market OVERWHELMINGLY wants iGPUs not dGPUs.

lol

and 3) a shrinking niche at that.

lol dawg watch out for dem attach rates for GeForce in laptop

You don't need AMD Halo levels of iGPU to satisfy that overwhelming majority of people who aren't doing anything that requires a dGPU

yeah you do, -halo (x60/70 laptop tier parts basically) is the baseline for 'good' gaming experience.

 

[Doug S]

lol dawg watch out for dem attach rates for GeForce in laptop

Those numbers are small.

You're probably using something stupid for your numbers like STEAM. Yeah using statistics grabbed from gamers is going to show a crazy high attach rate for dGPUs, but that's not real world. That's just the numbers for people within the gaming niche.

Corporations, and typical consumers, aren't buying laptops with a dGPU. They aren't buying laptops that even have the OPTION for a dGPU.

 

[adroc_thurston]

Those numbers are small.

They sure as hell are not.

Corporations, and typical consumers, aren't buying laptops with a dGPU.

Commercial is a separate market we're not talking about, and also a market which will never, ever touch the fat iGP SKUs for obvious reasons.
In any case, GeForce laptop attach rate is great and gaming laptop market is very healthy.

 

[Thunder 57]

They sure as hell are not.

Commercial is a separate market we're not talking about, and also a market which will never, ever touch the fat iGP SKUs for obvious reasons.
In any case, GeForce laptop attach rate is great and gaming laptop market is very healthy.

So you went from "the market" to "gaming market"? Talk about moving goalposts. And according to Amazon data (supposedly) laptops that include an Nvidia GPU are piss poor. As in 6.7% YTD. Even a healthy gaming market is peanuts compared to the market as a whole.

 

[Tup3x]

The amount of people who use integrated Intel GPUs to play their favourite gacha game is definitely not small. Also it definitely doesn't look like discrete graphics for laptops is the future considering what AMD is going to do (and NVIDIA with Intel) in the future.

 

[DrMrLordX]

Add to that the fact that the IEM doesn't get to bleed you for a dGPU markup, and you have the perfect storm for a moribund product.

This isn't even an AMD thread. Let's not go throwing stones at Strix Halo when it's selling out over and over again in miniPC format. AMD did very well with it.

If AMD was serious about Strix Halo, they would have partnered with an OEM to develop a reference platform to launch the product with. From what I've read, they didn't really do that beyond the uSFF developer and production boards. It's got the basics to be a great product, but they refused to fund developing an ecosystem for it.

Why would that have helped anything? They're moving boards, everything's good. For them anyway.

lol dawg watch out for dem attach rates for GeForce in laptop

Which isn't changing anytime soon since NV just inked a deal with Intel.

@Tupex

It wouldn't be too hard for NV to just ask Intel to integrate GeForce cards soldered onto Intel reference mobos and call it a day. Whether that makes it discrete or not is largely-irrelevant.

 

[poke01]

It wouldn't be too hard for NV to just ask Intel to integrate GeForce cards soldered onto Intel reference mobos and call it a day. Whether that makes it discrete or not is largely-irrelevant

they won’t be soldered onto to the mobo, it’s part of the SoC. It’s better like that, dGPUs are power hungry and the end game for laptops is using NVIDIA IP in an Intel SoC.

 

[511]

I seriously hope the display and Media IP used is Intel's not Nvidia

 

[ToTTenTranz]

People don't use iGPUs so forcing them into the full iGP config pricing at i3 is lol.

no because larger iGP is a cost adder that no one wants.

It's super weird that you're insisting so much on this flex when we know:

- AMD's Medusa range goes so hard on iGPUs that they basically converted their entire consumer RDNA5 dGPU range into chiplets that can go bolt onto APUs to improve iGPU performance.

- Intel's Nova Lake AX has up to 48 Xe3 cores, over twice as large as the B580 dGPU, with a whopping 144MB LLC and 256bit LP5X.

- Nvidia just gave $10B to Intel so they could put their iGPU chiplets into Intel's x86 APUs and be able to enter the big iGPU market.


Nvidia, Intel and AMD are all putting serious R&D money into these larger iGPUs you insist no one wants and no one uses.
Either all these 100billion/trillion companies (who undoubtedly spend millions on market research and user metrics) are collectively delirious at the same time on this specific subject, or you're just completely wrong.

 

[regen1]

I seriously hope the display and Media IP used is Intel's not Nvidia

That would be a good way to have less complications in design.
Nvidia's media engine support matrix and quality is very good especially 50 series(Blackwell) onwards.

I hope whatever they do, those remain separate lines and don't hamper Intel's own roadmaps(at least not too much). Intel has really decent GPU IP along with an extremely top class Media engine for their own usage.

 

[regen1]

- Intel's Nova Lake AX has up to 48 Xe3 cores, over twice as large as the B580 dGPU, with a whopping 144MB LLC and 256bit LP5X.

Those Nova Lake-AX specs are wrong apart from use of 256bit LPDDR5x.

 

[511]

Nvidia's media engine support matrix and quality is very good especially 50 series(Blackwell) onwards.

QSV is superior to NVENC

Those Nova Lake-AX specs are wrong apart from use of 256bit LPDDR5x.

yeah not possible that it is 48Xe3P

 

[regen1]

QSV is superior to NVENC

Not in all scenarios though. Especially with Blackwell(50 series), they had upper hand over the media engine in Arrowlake and Arc B580/570 series in a number of use cases.

 

[Meteor Late]

DavidC1 is absolutely right, they only give you the full iGPU in the highest core count config because that allows them to have the highest margins, people will spend more for the top die, so higher margins.
Like for example, in Strix Point, you want 16 CU, you have to buy a 12c/24t HX 370, when 8c/16t is more than enough. That's reality and he is right.

 

[511]

Not in all scenarios though. Especially with Blackwell(50 series), they had upper hand over the media engine in Arrowlake and Arc B580/570 series in a number of use cases.

B580 uses older one the same in ARL/MTL not the newest one.

 

[adroc_thurston]

AMD's Medusa range goes so hard on iGPUs

Two out of like bajillion parts are that?

that they basically converted their entire consumer RDNA5 dGPU range into chiplets that can go bolt onto APUs to improve iGPU performance.

More like they tape CCDs to a desktop dGPU to built-a-bear an APU.
Clever, but also secondary.

Intel's Nova Lake AX has up to 48 Xe3 cores, over twice as large as the B580 dGPU, with a whopping 144MB LLC and 256bit LP5X.

AX is smaller and idk what LSD trip did you get 144M LLC from.

 

[Tup3x]

It wouldn't be too hard for NV to just ask Intel to integrate GeForce cards soldered onto Intel reference mobos and call it a day. Whether that makes it discrete or not is largely-irrelevant.

It wouldn't be hard but it would be bad for power efficiency and obviously no unified memory. Besides, that's what they are doing since MXM lost its popularity.

 

[Fjodor2001]

DavidC1 is absolutely right, they only give you the full iGPU in the highest core count config because that allows them to have the highest margins, people will spend more for the top die, so higher margins

The interesting thing is that they're not doing the same for those that primarily want max CPU perf. I.e. they are not forcing the biggest iGPU upon them.

Meaning you can still get PTL with max CPU spec (4+8+4) but with lowest iGPU spec (4). So you don't have to get the 4+8+4+12 SKU.

 

[regen1]

B580 uses older one the same in ARL/MTL not the newest one.

Yeah, that's known so clubbed them in same group.Panther Lake got the upgrade with new format supports(and we need to see the quality tests).
But still 50-series media engine is pretty good, covered a lot of bases that were earlier missed and few areas where it was better like support for H.264 10 bit 4:2:2, better quality in AV1(40 series was also better in AV1), quicker than QSV(in ARL/MTL/B580) in general.
50 series media engine is kinda a very good all-rounder unlike previous gen(40-series and before).
PTL's media engine should be extremely good though, can't wait to see what they add with NVL's media engine.

 

[regen1]

DavidC1 is absolutely right, they only give you the full iGPU in the highest core count config because that allows them to have the highest margins, people will spend more for the top die, so higher margins.
Like for example, in Strix Point, you want 16 CU, you have to buy a 12c/24t HX 370, when 8c/16t is more than enough. That's reality and he is right.

Well, in MTL-H, ARL-H and in Lunar Lake:
Ultra 7/9 got 8(Xe/Xe+/Xe2) units, ultra 5 got 7(Xe/Xe+/Xe2) units and there's some frequency difference for iGPUs but you get the idea.
It's not that cut down at least in Intel's case.
There has been decent price gaps for model with ultra 5(lowest tier) to ultra 7 and from ultra 7 to ultra 9 in general.

In PTL-H(that's paired with bigger iGPU tile):
Ultra 7/9 will get 12Xe3 units, Ultra 5 will get 10Xe3 units. But we will have to see pricing once its settled down.

Now in Zen 5's case for some more available Strix Point and Krackan Point SKUs though IGPU is cut down a lot more
HX 370/375: (890m) 16units
AI 9 365: (880m) 12 units
AI 7 350: (860m) 8 units
AI 5 340: (840m) 4 units