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.

 

[OneEng2]

No, the $20B NRE would be amortized over a period of time. Intel expects 18A to be used in some capacity into the 2030's, so I would expect the amortization period to be of a similar time frame. Of course more volume lowers the per unit amortization cost and lack of external customers hurts that volume, but in no way is PTL alone expected to bear the entire NRE expense.

I disagree on your timeframe. The move to High NA alone will be a significant process cost and they are already spending money on it (lots of it).

I agree that PTL will not bear the entire expense; however, what else is on the plate between now and the next big process change (and $$$ expenditure?)?

My point is that it is difficult for Intel to go head to head with TSMC at a fraction of the capacity. It's a bad business strategy IMO.

Intel also seems dedicated to regaining the process lead through insanely risky (and costly) endeavors.

Perhaps 18A will show itself to be a match and then some to N2. There just isn't much evidence to show such a claim. Worse, even if it is, the juice may not be worth the squeeze.

 

[MoistOintment]

I disagree on your timeframe. The move to High NA alone will be a significant process cost and they are already spending money on it (lots of it).

I agree that PTL will not bear the entire expense; however, what else is on the plate between now and the next big process change (and $$$ expenditure?)?

My point is that it is difficult for Intel to go head to head with TSMC at a fraction of the capacity. It's a bad business strategy IMO.

Intel also seems dedicated to regaining the process lead through insanely risky (and costly) endeavors.

Perhaps 18A will show itself to be a match and then some to N2. There just isn't much evidence to show such a claim. Worse, even if it is, the juice may not be worth the squeeze.

How much of that $20B NRE includes 18AP? I imagine 18A and 18AP would be a part of the same amortization. This is an older pic from Q1 2024, but Intel was predicting 18A volume for a while.



There's also DMR, Clearwater Forest, Strong chance of Xe3P, likely some NVL dies. Then there's the likelihood that Intel 3 gets retired as the "cheap" tile in the future and 18A gets used for SoC / PCH tiles for future generations. All of these products bare bear 18A NRE costs.

Failing to secure external customers for 18A increased Intel's costs, yes. And there are questions around the ability to finance future nodes without getting external customers and the whole business model, but now that's starting to get outside the scope of PTL SoC pricing estimates.

 

[OneEng2]

Failing to secure external customers for 18A increased Intel's costs, yes. And there are questions around the ability to finance future nodes without getting external customers and the whole business model, but now that's starting to get outside the scope of PTL SoC pricing estimates.

Very fair point.

PTL will succeed or fail based on its product competitiveness in the market if the factors considered steer clear of the financial impact of the product and its supporting technologies on Intel.

 

[MoistOintment]

Very fair point.

PTL will succeed or fail based on its product competitiveness in the market if the factors considered steer clear of the financial impact of the product and its supporting technologies on Intel.

No, because the factors surrounding the costs of future nodes, like 14A, etc. don't impact the cost competitiveness of PTL.

The point being is that Intel has years of chips and production for which to amortize 18A NRE costs across. If a client SoC years from now is using 18A as its SoC chiplet, then a portion of 18A's development costs would be added to the price of that chiplet too. 18A does not need to pay itself off within a year or two. Intel has an amortization timeline in mind, and it extends past PTL and DMR

 

[DavidC1]

What makes these forums "unfun" is when someone has to argue when someone comments, "wow, the sky is really blue today" and someone else, for some reason has to respond, "that's not really blue, it's more bluish, magenta, violet, purple, it is objectively wrong to call a color blue when if you analyze the spectrum of the emitted light, it will not be monochromatic light in the center of spectrum scientists have established as being "blue," therefore your statement is facturally incorrect and without millions of dollars of equipment it is impossible to validate the exact composition of the emitted spectrum by percentage therefor NO comment on color can be made." Then 20 senseless back and forth posts over what constitutes a blue sky.

Buddy, I very much enjoy your contribution to the forums. There's some misunderstanding here however. And I apologize too. I do go about it that way often.

 

[Doug S]

I don't write opinions, I write facts and speculation based on facts. Just because many want to prove me wrong doesn't mean I'm wrong. Give a detailed example where I was wrong.

In my experience anyone who claims they are "never wrong" is both a liar and someone uniquely unwilling to ever admit when they are wrong even when it is proven to the satisfaction of nearly everyone else.

 

[poke01]

Interesting. If PTL-S on N3P and clocked 6+GHz came out it would not change anything, market is now expensive.

 

[Geddagod]

View attachment 137980

Interesting. If PTL-S on N3P and clocked 6+GHz came out it would not change anything, market is now expensive.

I doubt there ever was a plan to get PTL compute tiles externally tbh

 

[Hulk]

I sorted and downloaded all of the benchmark data in notebookcheck.com
Focused on 3 benches. Geekbench 5 ST, CB R23 ST, and 7-Zip ST.
Using my known "good" data for CBR23 I only kept the scores that looked valid, ~ +- 5% of CB R23 ST scores I know are solid. This provides some credibility for the 7-Zip and Geekbench scores.

It worked pretty darn well as the resulting scores were tightly grouped by core type.

Here is a summary. Everything lines up quite well with our thoughts I think.

If I were to average these three scores into one "throughput IPC-like score" how would you weight these 3 benches?

I have to say this chart is backed by a lot of data and it took some good work to get it to this point. Without my data collected over the years I could never have sorted the spam out of the Notebookcheck data. The result, the more I look at it, really does line up with what we "know from experience."

Now I just have to find a way to get good data on Gracemont, Skymont, and Darkmont worked into this!
Edit - I was able to pull data I believe is trustworthy from notebookcheck for Gracemont using "n" model cpus. Right at Skylake level performance so this seems right.

 

[Hulk]

As some people around here have been telling us Cougar/Lion Cove don't look as impressive outside of Cinebench.
+8% over Cougar Cove in Cinebench, pretty much flat in 7-Zip and 2% in Geekbench, which is probably most representative of daily used applications.

	CB R23ST		7-Zip ST		Geekbench 6 ST
2026 - Cougar Cove - Panther Lake	431	8%	1314	1%	599	2%
2024 - Zen 5	400		1300		585

When you compare Lion Cove to Zen 5 then Zen 5 comes out on top.

	CB R23ST		7-Zip ST		Geekbench 6 ST
2024 - Lion Cove - Arrow Lake	409	2.4%	1230	-5.4%	572	-2.3%
2024 - Zen 5	400		1300		585

 

[511]

Now I just have to find a way to get good data on Gracemont, Skymont, and Darkmont worked into this!
Edit - I was able to pull data I believe is trustworthy from notebookcheck for Gracemont using "n" model cpus. Right at Skylake level performance so this seems right.

David_Huang's Profile - Geekbench

browser.geekbench.com

Here is the data but on linux

 

[Hulk]

David_Huang's Profile - Geekbench

browser.geekbench.com

Here is the data but on linux

I'm asking Igor if he'll Geekbench test his Arrow Lake Skymont.

Here's how I used to isolate the Gracemonts on my Raptors. Turn off all P's except 1 in the BIOS. Set that one to 900MHz. When you run a ST benchmark Thread Director will go for a more performant E core over that dog slow 900MHz P core. Pretty reliable way to corner those E's!

I have analyzed David's data but couldn't get trends to line up as they should. I worked hard with that data.
Notebookcheck.com data was better because I had my CB data to corroborate their data.

I know that Skymont is about 357 points/GHz for CB R23 ST, which puts its performance right at Zen 4 level and just a hair below Golden/Raptor Cove. I don't want to add it to the chart until I get some actual data though.

Panther Lake's Darkmonts will be hard to isolate since BIOS access is usually limited for mobile but maybe someone can find a way?

 

[Hulk]

Regarding Zen 5, Cougar Cove, and Lion Cove "IPC."

My current thoughts are that Zen 5 is +1.8% over Lion Cove.
Cougar Cove is +3.8% over Zen 5.

Those numbers might be putting too much weight on CB. Meaning Zen 5 is really holding up well against the newcomers.

 

[poke01]

Meaning Zen 5 is really holding up well against the newcomers

considering Zen5 is on N4, it’s doing really well.

I don’t think Intel can hold up against AMD when AMD has node parity. NVL vs Zen6 should be an easy prediction on who will win 1T at least for me.

 

[511]

considering Zen5 is on N4, it’s doing really well.

I don’t think Intel can hold up against AMD when AMD has node parity. NVL vs Zen6 should be an easy prediction on who will win 1T at least for me.

RPL is holding well as well against modern 4/3 processors as for who would win ST we wold know in like 9 Months

 

[DrMrLordX]

There's also DMR, Clearwater Forest, Strong chance of Xe3P, likely some NVL dies. Then there's the likelihood that Intel 3 gets retired as the "cheap" tile in the future and 18A gets used for SoC / PCH tiles for future generations. All of these products bare bear 18A NRE costs.

Not to beat a dead horse but most Diamond Rapids volume (-SP) is now gone, and LBT didn't even respond to questions about Clearwater Forest during the last earnings report (Clearwater is presumed dead). If there's anything in DCG that's going on 18a/18ap in significant volume, it appears to be Coral Rapids.

 

[511]

Not to beat a dead horse but most Diamond Rapids volume (-SP) is now gone, and LBT didn't even respond to questions about Clearwater Forest during the last earnings report (Clearwater is presumed dead). If there's anything in DCG that's going on 18a/18ap in significant volume, it appears to be Coral Rapids.

Delays basically Killed Clearwater Forest as for 18A volume they still have NVL/PTL/WCL/DMR-AP for the next 2 Years Coral is likely 28 product

 

[DrMrLordX]

DMR-AP

That's not gonna ship many units in all likelihood, unless Intel is expecting it to do the heavy lifting now that Clearwater is gone (and even then . . .)

for the next 2 Years Coral is likely 28 product

It's still going to be 18AP though, innit?

 

[jdubs03]

It's still going to be 18AP though, innit?

Has to be, 14A won’t be in volume production until 2029 unfortunately.

 

[poke01]

RPL is holding well as well against modern 4/3 processors as for who would win ST we wold know in like 9 Months

yeah RPL desktop is good too but power hungry

 

[coercitiv]

No, because the factors surrounding the costs of future nodes, like 14A, etc. don't impact the cost competitiveness of PTL.

The point being is that Intel has years of chips and production for which to amortize 18A NRE costs across.

What you're saying makes sense for a healthy company with cash on hand, the Intel of today can't really afford the luxury of spreading amortization over the entire lifespan of a node.

That being said, the last piece of news I read was a rumor about Intel notifying chinese clients of a 6 month backlog on server chip orders. I think Intel might be doing fine in the short-mid term, one thing they master is to move money around to make all balance sheets look nice. (as long as there's money to move ofc).

 

[deasd]

Intel has reportedly canceled Ultra 9 290k Plus desktop CPU

https://videocardz.com/newz/intel-reportedly-cancels-core-ultra-9-290k-plus-but-keeps-270k-250k-arrow-lake-refresh-skus

 

[DrMrLordX]

Wonder if the 290k was going to be relatively-better than the 9850x3D.

 

[Hulk]

Intel has reportedly canceled Ultra 9 290k Plus desktop CPU

https://videocardz.com/newz/intel-reportedly-cancels-core-ultra-9-290k-plus-but-keeps-270k-250k-arrow-lake-refresh-skus

This is good news. Looks like Intel has learned something from the 14900K release. Meaning don't push clocks beyond a level in which the silicon can remain long-term healthy.

 

[Abwx]

Regarding Zen 5, Cougar Cove, and Lion Cove "IPC."

My current thoughts are that Zen 5 is +1.8% over Lion Cove.
Cougar Cove is +3.8% over Zen 5.

Those numbers might be putting too much weight on CB. Meaning Zen 5 is really holding up well against the newcomers.

Good idea to use 7-Zip, AT used it as representative of server s ST perf since that s an INT app with quite low IPC, and it s more representative for general perf in generic usage of a consumer PC.

Beside we can see that it took ARL to clearly match Zen 3, and also that Zen 5 had an unexpected regression vs Zen 4 while being clearly matched by PTL due to this loss.

You can see that it give an other perspective than Cinebench, FTR Haswell had 60% better ST IPC than C2D in Cinebench but was hardly 13% better in 7-ZIP, that s telling that Cinebench gave a wrong idea of PC perf evolution and why old PCs are still perfectly usable.