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.

 

[Magio]

There is only PPW uplift from the node Intel never gave up Tick Tock model they still use it for derisking.

If the node was a huge jump in transistor performance over N3B (far from TSMC's best node) one would assume it could have at least achieved better clocks at same power and that Intel would have preferred that to lower clocks (than ARL-H) at lower power. And of course to this point we don't even know if it actually is more efficient.

 

[511]

If the node was a huge jump in transistor performance over N3B (far from TSMC's best node) one would assume it could have at least achieved better clocks at same power and that Intel would have preferred that to lower clocks (than ARL-H) at lower power. And of course to this point we don't even know if it actually is more efficient.

Increased xtor performance doesn't necessarily means Fmax increase.
For N3B and 18A comparison LNL is a solid baseline.

 

[DrMrLordX]

No way man. Is this thing going to be Cannon Lake / Ice Lake 2.0?? ie an "investor requirement" launch??

Nah, we were warned about parametric yields, so no surprises there.

Probally going to be meme level volume. So they are being conservative on clocks.

Initial launch will probably be like that, yes. Who knows how long it will be before they can push more volume? Also parametric yields (see above). Personally I was guessing 5.1-5.3 Ghz but it looks like they didn't get enough to bin that high.

 

[511]

Initial launch will probably be like that, yes. Who knows how long it will be before they can push more volume? Also parametric yields (see above). Personally I was guessing 5.1-5.3 Ghz but it looks like they didn't get enough to bin that high.

at least it's the same as LNL if it was <5.0 Ghz it would have been comedy 🤣 🤣. If you believe raichu it's 40-50% total yield in August since the event is going on it means the QS has been finalized.

 

[DrMrLordX]

at least it's the same as LNL if it was <5.0 Ghz it would have been comedy 🤣 🤣

Comedy for us, tragedy for IFS.

 

[Joe NYC]

at least it's the same as LNL if it was <5.0 Ghz it would have been comedy 🤣 🤣. If you believe raichu it's 40-50% total yield in August since the event is going on it means the QS has been finalized.
View attachment 131243

It explains why so many potential foundry customers lost interest.

 

[511]

It explains why so many potential foundry customers lost interest.

Should be improved as they ramp up but yeah this and the fact that 18A PDK was botched for external

 

[igor_kavinski]

They should start selling the defective dies as ornaments to recoup some of the cost.

A sneak peak at my marketing genius: 18A Nova Lake 52-core package keychain. It's real! Use a SEM to look at it and marvel at modern technology miracles! Maybe you find the secret code to claim an actual CPU! Get yours today while supplies last!

 

[CouncilorIrissa]

at least it's the same as LNL if it was <5.0 Ghz it would have been comedy 🤣 🤣. If you believe raichu it's 40-50% total yield in August since the event is going on it means the QS has been finalized.
View attachment 131243

tbf LNL for the longest time had very few 5.0+ bins available (268v almost didn't exist whatsoever).

 

[511]

tbf LNL for the longest time had very few 5.0+ bins available (268v almost didn't exist whatsoever).

that was on N3B that was very mature during LNL launch

They should start selling the defective dies as ornaments to recoup some of the cost.

A sneak peak at my marketing genius: 18A Nova Lake 52-core package keychain. It's real! Use a SEM to look at it and marvel at modern technology miracles! Maybe you find the secret code to claim an actual CPU! Get yours today while supplies last!

These exists for that reason

 

[regen1]

Parametric yield can improve with time. In past with "plus" node frequencies got significant boost(esp. for same u-arch and design targets).
Intel 4 and Intel 3 could be considered same family of nodes. MTL-H and MTL-U got max of 5.1GHz and 4.9GHz respectively at launch with Intel 4.
ARL-U(Redwood cove P-cores as in MTL-U) on Intel 3 got max of 5.3GHz, 0.4GHz extra.
During Tigerlake's launch i7 1185G7(max 4.8GHz) was the top SKU but ~9months later i7 1195G7(max 5GHz) was launched on the same node(probably a new stepping iirc). Subsequent releases(12th,13th and refreshes) on Intel 10nm/7 family came with increased max frequencies.
Other factors could be if the design is focused on outright performance or perf/W, area efficiency.

tbf LNL for the longest time had very few 5.0+ bins available (268v almost didn't exist whatsoever).

Yeah, but to an extent that has been generally the case(for a long time) with top end i7/ultra 7 vPro SKUs(12700H v 12800h, 13700 v 13800H, 155H v 265H, 255H v 265H, 258V v 268v), they are clocked a little higher than the more mainstream i7/ultra 7and mostly available in business and some premium notebook designs.

 

[Saylick]

View attachment 131196

screenshot

lol, that e-beggar Prakhar is trying to pull a MLID by backtracking his statement by arguing his "maybe" implied he was just guessing. Something tells me had his guess been right, he would've paraded his xeet around as if he knows something others don't.

 

[Meteor Late]

Parametric yield can improve with time. In past with "plus" node frequencies got significant boost(esp. for same u-arch and design targets).

That's all fine and dandy but not when frequencies don't improve over the previous base node. So if MTL-H gets 5.1GHz with base Intel 4 and then Panther Lake gets 5.1GHz with 18A base node, what's that if not a colossal failure?

 

[regen1]

That's all fine and dandy but not when frequencies don't improve over the previous base node. So if MTL-H gets 5.1GHz with base Intel 4 and then Panther Lake gets 5.1GHz with 18A base node, what's that if not a colossal failure?

For sure it could be somewhat underwhelming to some. Not ruling out that it could be underwhelming relatively even after 18A-P.

Still a little too early to call failure though, different designs and peak frequency has not regressed. Many factors even aside from the node could be at play, even the design targets for PTL or a possible risk-averse first release on a node with 2 new tech(GAAFET and BSPDN). We'll know how bad or good 18A family is after reviews drop(esp. on perf/W and other factors plus we have to see 18A-P).

 

[poke01]

thats really good

 

[regen1]

View attachment 131289
thats really good

Don't know what "Fmax" he is referring to.

PTL-H(at least in the initial launch SKUs) have max frequency of 5.1GHz.
There's another SKU with max frequency reaching 5GHz.

 

[Kepler_L2]

View attachment 131289
thats really good

100% nonsense

 

[poke01]

100% nonsense

Just 10 more days till they can’t bs anymore I suppose

 

[adroc_thurston]

Just 10 more days till they can’t bs anymore I suppose

yeah.
In any case, you get 5.1 top bin and it launches next year.

 

[511]

That's all fine and dandy but not when frequencies don't improve over the previous base node. So if MTL-H gets 5.1GHz with base Intel 4 and then Panther Lake gets 5.1GHz with 18A base node, what's that if not a colossal failure?

Different uArch?

Also why are we talking MLIDs sensei seriously

 

[ToTTenTranz]

View attachment 131289
thats really good

This would make 18A a total winner.

 

[511]

This would make 18A a total winner.

Hold your horses that's not true it's Memg doing meng things. The power efficiency part can be true that it can deliver ARL-H perf at LNL TDP but not the max clock.

 

[ToTTenTranz]

Hold your horses that's not true it's Memg doing meng things. The power efficiency part can be true that it can deliver ARL-H perf at LNL TDP but not the max clock.

Yes, I don't think PTL-H will have those clocks. But I also don't get why it's so important for people that a laptop CPU reaches 5.5GHz or has massive single-core performance.


Are people spending that much time running Geekbench 6 on their laptops? Or are they going to pair a 45W Panther Lake H with a RTX 5090 and they're afraid of hitting a CPU bottleneck? Were they planning on using PTL-H laptops to run servers for multiplayer games?

Power efficiency at lower clocks should be much more important for end users, and transistor density at those efficiency levels should be much more important for Intel. I don't know if Intel's 18A has performance libraries and then power-efficiency + transistor density libraries like TSMC does with their processes, but if they do then they probably went with the latter for Panther Lake.



And after power efficiency at general loads, what matters more for people will be how fast the Xe3 iGPU is for games or AI workloads. Being able to load up different LLMs or image generation models on AI Playground and running those with decent performance is bound to be a lot more important than reaching a 4600 ST score in geekbench 6.


It's not that I don't get the enthusiast PoV here. Hitting 5.7GHz would be cool. But if PTL-H is exclusively a mobile chip and Arrow Lake's desktop successor is only coming with Nova Lake, I'm not following on this "OMG 18A is horrible Intel is lost because the CPU cores in their mobile chips can't hit 5.5GHz" meme.

 

[511]

And after power efficiency at general loads, what matters more for people will be how fast the Xe3 iGPU is for games or AI workloads. Being able to load up different LLMs or image generation models on AI Playground and running those with decent performance is bound to be a lot more important than reaching a 4600 ST score in geekbench 6.

~50% if it doesn't get botched by memory bandwidth in the game

 

[LightningZ71]

Parametric yield can improve with time. In past with "plus" node frequencies got significant boost(esp. for same u-arch and design targets).
Intel 4 and Intel 3 could be considered same family of nodes. MTL-H and MTL-U got max of 5.1GHz and 4.9GHz respectively at launch with Intel 4.
ARL-U(Redwood cove P-cores as in MTL-U) on Intel 3 got max of 5.3GHz, 0.4GHz extra.
During Tigerlake's launch i7 1185G7(max 4.8GHz) was the top SKU but ~9months later i7 1195G7(max 5GHz) was launched on the same node(probably a new stepping iirc). Subsequent releases(12th,13th and refreshes) on Intel 10nm/7 family came with increased max frequencies.
Other factors could be if the design is focused on outright performance or perf/W, area efficiency.


Yeah, but to an extent that has been generally the case(for a long time) with top end i7/ultra 7 vPro SKUs(12700H v 12800h, 13700 v 13800H, 155H v 265H, 255H v 265H, 258V v 268v), they are clocked a little higher than the more mainstream i7/ultra 7and mostly available in business and some premium notebook designs.

For 14+... And for 10nm/Intel 7, they improved frequencies through successively reducing logic density by small increments (excluding super fin, which was an actual change in the XTOR shape).