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+4+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.

 

[511]

compared to Lunar Lake, Panther offers basically no improvements in ST.

Intel graphs also show this though

Did you actually see the presentation? It's like a Tick+.

I did it's nearly a copy past many things were already present in Lion Cove the only thing changed were BP/TLB/memory disambiguation this is not a change that would drive Much ST improvement the main improvement were due to the uncore changes in PTL vs ARL.

 

[Hulk]

@Hulk

It's not 2-3%, that's what we get with minor changes like Crestmont. The Cougar Cove got bigger changes. 5-10% is believable.

It's clocking lower, hence cancelling out the advantages. It needs to be 5% faster just to cancel out the loss in clocks.

It's losing frequency compared to Lunar Lake? That's the only Lion Cove/Skymont to compare, right?

 

[jdubs03]

I know it’s just Geeekbench but:


Ultra 9 285H


Pts per clock (single core):
UX9 388H: 601
U9 285H: 533

That’s not terrible in terms of IPC.

 

[DavidC1]

Intel graphs also show this though

I did it's nearly a copy past many things were already present in Lion Cove the only thing changed were BP/TLB/memory disambiguation

Branch prediction changes are a significant change. It never happens in a pure shrink/Tick core.

 

[511]

It's losing frequency compared to Lunar Lake? That's the only Lion Cove/Skymont to compare, right?

It didn't loose Frequency vs Lunar it's 5.1 GHz for Lunar and 5.1 For Panther Lake

 

[poke01]

Intel graphs also show this though

Sigh do they?



This one shows they regressed in ST

 

[511]

Branch prediction changes are a significant change. It never happens in a pure shrink/Tick core.

Depends on changes but Lion cove has a worse branch predictor than Skymont

 

[DavidC1]

I know it’s just Geeekbench but:

Ultra 9 285H

Pts per clock (single core):
UX9 388H: 601
U9 285H: 533

That’s not terrible in terms of IPC.

There's nothing wrong with Geekbench. But it suffers from all user-submitted benchmark problems: lack of consistency. That's why sorting is a big deal, because you need to compare peak non-OC results using same OS, and then they are comparable. The peak scores are those free from OS, driver, configuration, and random fluke variations. That goes same with @poke01 GB results.

 

[Hulk]

Depends on changes but Cougar cove has a worse branch predictor than Lion Cove

Please explain?

 

[511]

This one shows they regressed in ST

By 1-2% margin of error? I was more looking at the CB 24 graph now can't find it

 

[DavidC1]

It didn't loose Frequency vs Lunar it's 5.1 GHz for Lunar and 5.1 For Panther Lake

It's 5.4GHz for 285H, which is also a predecessor of. We can't conveniently exclude Arrowlake when Intel themselves have said it replaces both Lunar and Arrow, and it makes sense.

By 1-2% margin of error? I was more looking at the CB 24 graph now can't find it

That's why Pantherlake is Icelake v2.

 

[511]

Please explain?

I meant Lion Cove has worse branch predictor than skymont I got confused between the cores
😅

 

[DavidC1]

I meant Lion Cove has worse branch predictor than skymont I got confused between the cores
😅

We're talking about Cougar Cove though. And BP change is a significant change. So are TLB and memory disambiguation changes. A tick would get maybe TLB and leave it at that.

 

[511]

It's 5.4GHz for 285H, which is also a predecessor of. We can't conveniently exclude Arrowlake when Intel themselves have said it replaces both Lunar and Arrow, and it makes sense.

It is for sure a regression vs arl

 

[DavidC1]

From Techpowerup.

Cougar Cove keeps the same pipeline depth as Lion Cove but is more efficient at keeping its execution units busy. The branch predictor and instruction fusion logic have been refined, and prefetch behavior is more adaptive to mixed workloads. Engineers explained that the design team relied on performance data gathered from earlier client chips to identify where short, bursty workloads were leaving execution units idle. These small refinements make the new P-core more consistent in real-world client workloads, without changing its overall structure.

The reorder buffer and allocation windows have been resized to improve balance between integer and vector operations. The front end is now better at sustaining fetch and decode bandwidth in small instruction loops, reducing wasted cycles in everyday tasks such as web rendering or office applications. Intel described these changes as incremental but meaningful, designed to cut down on "wasted wakeups" and improve the core's steady-state efficiency.

These are significant changes. That's why the argument is 50/50 on whether Pantherlake 5.1 is faster than Arrowlake 5.4. The loss in clocks mean there will inevitably be applications where Pantherlake is slower than Arrowlake. That's what averages mean.

 

[511]

We're talking about Cougar Cove though. And BP change is a significant change. So are TLB and memory disambiguation changes. A tick would get maybe TLB and leave it at that.

RWC was a tick and it Got lot more than that

 

[DavidC1]

RWC was a tick and it Got lot more than that

RWC got very minor changes. You can allude to it by the wording differences, where RWC presentation had nothing worth pointing out, while Cougar Cove did.

If what you are saying is true that it's like RWC, then Pantherlake 5.1 will be in average 2-3% slower than Arrowlake 5.4. That doesn't seem to be the case.

 

[Hulk]

Sigh do they?
View attachment 136556


This one shows they regressed in ST

Hold on. Let's have a good look at this.
First of all I see Panther Lake beating the HX370 while using 8 Watts less. It's not huge, but well, there it is.

Second, we have to look at the core configuration for Panther and Arrow H.
4+8+4 for Panther and 6+8+2 for Arrow

Here's the thing. At Iso-Power P cores are always going to be more efficient than E cores.
Panther has traded 2 P's for LPE's and still comes out about even in 4 benchmarks and far ahead in another. Just talking CPU here.

Also, honestly the one where it dominates is the most important one, video editing, the one that gets you waiting the most.

So full power is going to be the toughest comparison for Panther to Arrow due to that 2P for 2LPE swap. But I would suspect battery life is going to be great because that LPE island might actually be quite able to keep the P's sleeping much of the time.

Now to play devil's advocate. If Intel has kept the core configuration the same for Panther, that being 6+8+2, Panther would have looked better and we would have gotten a better CPU.

But Intel is Intel and they saved quite a bit of die area by swapping those P's for E's. Even if the ratio is now 3:1, losing 2 P's give you enough area for 6 E's. They only need to add 2 E's, not 6, so they saved the die space of an entire Darkmont cluster!

To quote Newman from Seinfeld in Jurasic Park, "They got cheap on us."

 

[511]

RWC got very minor changes. You can allude to it by the wording differences, where RWC presentation had nothing worth pointing out, while Cougar Cove did.

If what you are saying is true that it's like RWC, then Pantherlake 5.1 will be in average 2-3% slower than Arrowlake 5.4. That doesn't seem to be the case.

Cause ARL had bade memory latency and poor L3? It got fixed with Panther remember how coooed ARL uncore is

 

[DavidC1]

Hold on. Let's have a good look at this.
First of all I see Panther Lake beating the HX370 while using 8 Watts less. It's not huge, but well, there it is.

He's basing it on the Speedometer result which is non-GPU, non-media single core.

Cause ARL had bade memory latency and poor L3? It got fixed with Panther remember how coooed ARL uncore is

Intel didn't say anything about Redwood client changes on their slide other than bigger L1. They have for Cougar Cove. Another thing is I believe the claims that Lion Cove missed targets and Cougar fixes them.

 

[Hulk]

RWC got very minor changes. You can allude to it by the wording differences, where RWC presentation had nothing worth pointing out, while Cougar Cove did.

If what you are saying is true that it's like RWC, then Pantherlake 5.1 will be in average 2-3% slower than Arrowlake 5.4. That doesn't seem to be the case.

Intel has been fighting the move from monolithic since Redwood cove from a latency point-of-view. Redwood Cove's cache improvements were directly aimed at reducing the tile penalty.

Arrow Lake continued the trend with the L0 cache and gigantic L1 data cache. I think Intel is still figuring out the tiled front end.

 

[511]

Here's the thing. At Iso-Power P cores are always going to be more efficient than E cores.

Depends on power

 

[Hulk]

I meant Lion Cove has worse branch predictor than skymont I got confused between the cores
😅
View attachment 136557

Geez, I wish Raptor Cove was in that chart!

 

[511]

Intel didn't say anything about Redwood client changes on their slide other than bigger L1. They have for Cougar Cove. Another thing is I believe the claims that Lion Cove missed targets and Cougar fixes them.

Well Intel cooked the L3/Uncore in meteor lake also this

Intel’s Redwood Cove: Baby Steps are Still Steps

Intel’s Meteor Lake chip signaled a change in Intel’s mobile strategy, moving away from the monolithic designs that had characterized Intel’s client designs for more than a decade.

chipsandcheese.com

 

[DavidC1]

Hold on. Let's have a good look at this.
First of all I see Panther Lake beating the HX370 while using 8 Watts less. It's not huge, but well, there it is.

Second, we have to look at the core configuration for Panther and Arrow H.
4+8+4 for Panther and 6+8+2 for Arrow

Here's the thing. At Iso-Power P cores are always going to be more efficient than E cores.
Panther has traded 2 P's for LPE's and still comes out about even in 4 benchmarks and far ahead in another. Just talking CPU here.

Also, honestly the one where it dominates is the most important one, video editing, the one that gets you waiting the most.

So full power is going to be the toughest comparison for Panther to Arrow due to that 2P for 2LPE swap. But I would suspect battery life is going to be great because that LPE island might actually be quite able to keep the P's sleeping much of the time.

Battery life is going to be great except in pure idle, which isn't realistic as that means not using computer. Icelake also had great idle. We know it works from Lunarlake. Meteor/Arrow's LPE was too slow to be useful.

"LPE" in Lunar and Panther is more like a power efficient version of the E, rather than the useless LPE in Meteor/Arrow. The LPE in Arrowlake is an insignificant contributor to MT performance. So for MT it's 6P+8E vs 4P+8E+4LPE.

Now to play devil's advocate. If Intel has kept the core configuration the same for Panther, that being 6+8+2, Panther would have looked better and we would have gotten a better CPU.

If they kept LPE like on Arrow, then the battery life would have been poor as Arrow.

Well Intel cooked the L3/Uncore in meteor lake also this

Intel’s Redwood Cove: Baby Steps are Still Steps

Intel’s Meteor Lake chip signaled a change in Intel’s mobile strategy, moving away from the monolithic designs that had characterized Intel’s client designs for more than a decade.

chipsandcheese.com

Yes, but Intel themselves didn't point that out, because it didn't do anything really. In Pantherlake they are. If they were insignificant as Pantherlake, they couldn't claim equivalence or in some cases better than Arrowlake.