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.

 

[regen1]

Intel has made many sizable mistakes over the last couple of decades. But to me, by far, the biggest mistake was authorizing over $150 billion in stock buybacks. Note: not all of that was spent (a quick search shows $108 billion to $128 billion was spent). But that is money that has vanished for Intel with nothing to show for it. Think of where they would be if they invested even 1/3rd of that into their fabs instead (or like Apple saved it for rainy days--like now).

Yeah, it's not just that did buybacks but also they didn't go to EUV route(one of the many major issues) or invest in it soon enough.

 

[511]

Yeah, it's not just that did buybacks but also they didn't go to EUV route(one of the many major issues) or invest in it soon enough.

10nm was soon for EUV it was supposed to be a 2016 node and EUV was not ready even TSMC inserted EUV 2018-19 s

 

[regen1]

10nm was soon for EUV it was supposed to be a 2016 node and EUV was not ready even TSMC inserted EUV 2018-19 sView attachment 132208

Yeah, for 10nm esp. its original timeline it wasn't probable.
Thing is even subsequently the internal development and investment for EUV didn't catch pace fast enough. The development timelines for subsequent nodes could be faster if it did.

It's not just those stuff, the way they ignored their own networking section and simultaneously also fumbled Mellanox acquisition among many other stuff. When 10nm was borked/delayed should have given a lot freer hand to Altera to design in external fabs(something similar to Mobileye which was free to fab wherever it wanted)

 

[jpiniero]

Yeah, for 10nm esp. its original timeline it wasn't probable.
Thing is even subsequently the internal development and investment for EUV didn't catch pace fast enough. The development timelines for subsequent nodes could be faster if it did.

They were probably busy trying to fix 10 nm first. Given that they had already spent the money on acquiring the tools and such, 10 nm was always going to get 2 years no matter what. Remember they had remote sites all set up, waiting for the final go. Then had to take them down.

 

[511]

It was Intel basically stuck on Cobalt and their Hyper Scaling when scaling became hard

 

[regen1]

They were probably busy trying to fix 10 nm first. Given that they had already spent the money on acquiring the tools and such, 10 nm was always going to get 2 years no matter what. Remember they had remote sites all set up, waiting for the final go. Then had to take them down.

They were trying to fix 10nm among its many revisions but its the timeline of subsequent nodes where the failure got extended and they are still paying for this. 2017-2021 is a period where Intel has quite stagnated in foundry progress but they are still raking in a lot of money, AMD is still in the process of getting consumer confidence and Milan moment was little farther in the line and ChatGPT GPT moment is still a long way away.
Ian Cutress is of the opinion that had Intel been more aggressive in their EUV game earlier the present nodes(Intel 4/3, 18A family) would have been significantly earlier and deeper in their cycle which I agree with. (EUV machine do take a lot of time to arrive from the date of order)

It was Intel basically stuck on Cobalt and their Hyper Scaling when scaling became hard

And there went its collaboration with ARM and others that time(trying to be in foundry game then). One could say they had cultural and other issues as well for foundry game but broken 10nm meant no progress at all.

 

[511]

Ian Cutress is of the opinion that had Intel been more aggressive in their EUV game earlier the present nodes(Intel 4/3, 18A family) would have been significantly earlier and deeper in their cycle which I agree with. (EUV machine do take a lot of time to arrive from the date of order)

this was all due to decisions made pre 2020

 

[regen1]

this was all due to decisions made pre 2020

Yeah, if Pat got the role in 2019(he and Board didn't agree to terms then) foundry progress likely would have been well ahead of where it is today(may be the critical TowerSemi deal would also have happened well earlier without China blocking). Also Networking(spinning Cornelis/Omnipath) and Mellanox fumble probably won't have happened among many other misses. Buybacks are not always that bad but even in 2020 they were doing it, that wasn't the time to do buybacks.

 

[regen1]

[News] Intel Reportedly Raises Prices Up to 20% for Intel 7-Based Raptor Lake, Alder Lake; Asia Market Hit Hard

While Intel has just unveiled Panther Lake, its first major product built on the 18A process, the company is reportedly raising prices for Alder Lake ...

www.trendforce.com

Preissteigerung bei Core-CPUs: Intel-Prozessoren werden auch in Deutschland teurer

Vor Wochen erstmals berichtet sind Preissteigerungen bei Intel-CPUs nun auch in Deutschland klar sichtbar. Sie gelten aber nicht überall.

www.computerbase.de

Seems further price increase, more visibly experienced in more regions.

 

[511]

Intel 7 is capacity constrained and the demand is very high so it was kind of expected

 

[fastandfurious6]

the market is already saturated with intel 12/13/14 machines, who keeps buying them 🤣🤣

 

[511]

the market is already saturated with intel 12/13/14 machines, who keeps buying them 🤣🤣

70% of the client market 😛

 

[regen1]

Intel 7 is capacity constrained and the demand is very high so it was kind of expected

Yeah but may be 20% is a little too much.

 

[jpiniero]

Ian Cutress is of the opinion that had Intel been more aggressive in their EUV game earlier the present nodes(Intel 4/3, 18A family) would have been significantly earlier and deeper in their cycle which I agree with. (EUV machine do take a lot of time to arrive from the date of order)

I'm sure Intel had it's far share of problems with Intel 4.

 

[511]

Yeah but may be 20% is a little too much.

MMaybe they want consumer to buy newer products or something anyway whenever there is high demand prices get raised.

 

[DZero]

[News] Intel Reportedly Raises Prices Up to 20% for Intel 7-Based Raptor Lake, Alder Lake; Asia Market Hit Hard

While Intel has just unveiled Panther Lake, its first major product built on the 18A process, the company is reportedly raising prices for Alder Lake ...

www.trendforce.com

Preissteigerung bei Core-CPUs: Intel-Prozessoren werden auch in Deutschland teurer

Vor Wochen erstmals berichtet sind Preissteigerungen bei Intel-CPUs nun auch in Deutschland klar sichtbar. Sie gelten aber nicht überall.

www.computerbase.de

Seems further price increase, more visibly experienced in more regions.

Increasing prices to CRASHtor lake?
how to make a failed process a total dissaster? going that way.

 

[511]

Increasing prices to CRASHtor lake?
how to make a failed process a total dissaster? going that way.

It's still selling more than anything don't forget the RPL-U/RPL-H were not affected and neither ADL ADL repurposed RPL Dies the only issues were the 8+16 Die and it's bin.

 

[oak8292]

I'm sure Intel had it's far share of problems with Intel 4.

Yes, they did have problems. It was called 7 nm at the time and the expectation was it would be used for Ponte Vecchio in the Aurora Supercomputer.

The Intel 7 nm process was delayed and based on what I interpreted from the ASML conference calls the EUV machines were redirected to TSMC to produce the Aurora parts. Intel was paying penalties for the delay in Aurora.

During the pandemic both TSMC and Intel cancelled EUV orders. TSMC turned around and wanted their equipment. The Intel process was not working and they went to TSMC for Aurora.

This is comments by ASML from 4th Quarter 2020:

“And it's just a reflection of what happened last year in Q2 and Q3 were, you know, clearly our key foundry customer came back and said, listen, our key customer for N3 is now blacklisted. (TSMC and Huawei)

So, we cannot ship. So, we need to adjust our 2021 outlook for EUV systems, which was followed by another customer said, well, we're going to delay the roadmap, (Intel)”

Also from ASML on redirecting EUV machines.

“So, the issue is, is that when we look at 2021, two things are impacting our shipment schedule, to our leading edge customers is, one is, what you refer to is effectively has there been a transition from tools that we originally planned for customer a, potentially to customer b and c, I think yes, that has happened.”

ASML Holding N.V. (ASML) CEO Peter Wennink on Q4 2020 Results - Earnings Call Transcript

ASML Holding N.V. (NASDAQ:NASDAQ:ASML) Q4 2020 Earnings Conference Call January 20, 2021 9:00 AM ETCompany ParticipantsSkip Miller - Vice President,...

seekingalpha.com

There are not a lot of customers for EUV so even if they don’t name the customers it is fairly obvious, and even more so after time has passed to confirm what happened.

 

[LightningZ71]

Retail demand has been high for the last few quarters as corporations and governments raced to complete their Windows 10 purges ahead of the October deadline. They've been buying whatever is available in quantity all year. Expect client demand to fall off a cliff soon.

 

[MoistOintment]

Retail demand has been high for the last few quarters as corporations and governments raced to complete their Windows 10 purges ahead of the October deadline. They've been buying whatever is available in quantity all year. Expect client demand to fall off a cliff soon.

Which is something we had to do as well. Granted, we only had 200 laptops that didn't support Windows 11, so it wasn't a huge order, but when we looked at what our supplier had, RPL-U Dell Latitudes were  nearly half the price of MTL-U. There wasnt even a discussion: RPL-U was the obvious choice.

I imagine we're not the only ones, because when youre ordering laptops in the hundreds or thousands at a time, a $400 - $800 premium per unit for MTL-U (or ARL-U) over RPL-U is just simply not worth it at all for most enterprises.

 

[LightningZ71]

It's what we saw too. Our earlier batches over the last year were Adl-u/p and RCL-u. MTL prices came down for later batches and we were aiming more for the mobile crowd instead of desk jockeys that needed to walk down the hall to meetings and take an occasional trip. Multiple 4 digit orders throughout the year to beat the deadline.

I gotta say, MTL may get pooped on for absolute performance, but our mobile users really like them. My work laptop recently got refreshed from an 8 core Comet Lake I7 H to a core 285 MTL and while I don't notice the speed being much different, it sips battery compared to the Comet Lake H. I expect to see some execs get LNL tablets/2 in 1 devices soon, so I'll get some first hand with them too.

 

[OneEng2]

Yeah, it's not just that did buybacks but also they didn't go to EUV route(one of the many major issues) or invest in it soon enough.

They were trying to fix 10nm among its many revisions but its the timeline of subsequent nodes where the failure got extended and they are still paying for this. 2017-2021 is a period where Intel has quite stagnated in foundry progress but they are still raking in a lot of money, AMD is still in the process of getting consumer confidence and Milan moment was little farther in the line and ChatGPT GPT moment is still a long way away.
Ian Cutress is of the opinion that had Intel been more aggressive in their EUV game earlier the present nodes(Intel 4/3, 18A family) would have been significantly earlier and deeper in their cycle which I agree with. (EUV machine do take a lot of time to arrive from the date of order)

I completely agree with this.

Intel made its way by maintaining node superiority over the last 30 years. AMD stubbed it's toe on "Real Men Have Fabs", but eventually recovered. They also strategically purchased ATI which at the time people thought was dumb (but was a great long term investment). Once AMD got on TSMC, and Intel stumbled with 10nm there was a delay as you point out as Intel is a mighty lumbering beast (as is the market and the big OEM deals). It is only now that the piper must be paid .

 

[511]

I completely agree with this.

Intel made its way by maintaining node superiority over the last 30 years. AMD stubbed it's toe on "Real Men Have Fabs", but eventually recovered. They also strategically purchased ATI which at the time people thought was dumb (but was a great long term investment). Once AMD got on TSMC, and Intel stumbled with 10nm there was a delay as you point out as Intel is a mighty lumbering beast (as is the market and the big OEM deals). It is only now that the piper must be paid .

OEM market is difficult to take cause you require a lot of efforts not just the best product and than there are other stuff people forget about Intel's Wi-Fi/BT/TB and other connectivity stuff that other vendor doesn't provide.

 

[OneEng2]

OEM market is difficult to take cause you require a lot of efforts not just the best product and than there are other stuff people forget about Intel's Wi-Fi/BT/TB and other connectivity stuff that other vendor doesn't provide.

Which is why Intel is still very strong in OEM IMO. It isn't necessarily about who has the best chip.... still, with laptops, having silly long battery life is a huge plus as it allows the OEM to use a smaller battery reducing the BOM costs.

LNL is actually quite good.

 

[oak8292]

OEM market is difficult to take cause you require a lot of efforts not just the best product and than there are other stuff people forget about Intel's Wi-Fi/BT/TB and other connectivity stuff that other vendor doesn't provide.

I am not an expert on this but it seems that a lot of the required connectivity IP is available from third parties. Some is even available already laid out on specific nodes.

I am not sure that TB is available from third parties but BT and WiFi are.

Here is an announcement from Broadcom from last week.

“Wi-Fi 8 IP for license for Internet of Things (IoT), automotive, and mobile devices as a complement to the access ecosystem.”

Broadcom Introduces Industry's First Wi-Fi 8 Silicon Ecosystem Powering the AI Era

Broadcom Inc. today announced the industry's first Wi-Fi 8 silicon solutions for the broadband wireless edge ecosystem including residential gateways, enterprise access points, and smart mobile clients. Broadcom's Wi-Fi 8 solutions are purpose-built to address the stringent performance...

www.techpowerup.com

TSMC has OIP with a lot of third party IP already laid out on specific nodes.