Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads

[Tigerick]

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.



Comparison of upcoming Intel's U-series CPU: Core Ultra 100U, Lunar Lake and Panther Lake

Model	Code-Name	Date	TDP	Node	Tiles	Main Tile	CPU	LP E-Core	LLC	GPU	Xe-cores
Core Ultra 100U	Meteor Lake	Q4 2023	15 - 57 W	Intel 4 + N5 + N6	4	tCPU	2P + 8E	2	12 MB	Intel Graphics	4
?	Lunar Lake	Q4 2024	17 - 30 W	N3B + N6	2	CPU + GPU & IMC	4P + 4E	0	12 MB	Arc	8
?	Panther Lake	Q1 2026 ?	?	Intel 18A + N3E	3	CPU + MC	4P + 8E	4	?	Arc	12

Comparison of die size of Each Tile of Meteor Lake, Arrow Lake, Lunar Lake and Panther Lake

	Meteor Lake	Arrow Lake (20A)	Arrow Lake (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	4P + 4E	4P + 8E
LLC	24 MB	24 MB ?	36 MB ?	12 MB	?
tCPU	66.48
tGPU	44.45
SoC	96.77
IOE	44.45
Total	252.15

Intel Core Ultra 100 - Meteor Lake



As mentioned by Tomshardware, TSMC will manufacture the I/O, SoC, and GPU tiles. That means Intel will manufacture only the CPU and Foveros tiles. (Notably, Intel calls the I/O tile an 'I/O Expander,' hence the IOE moniker.)

 

[majord]

It's a bit hard to hit pause buttons on CPU Architectures. Unpausing 1, 2 3 yrs later and your once innovative, advanced , ambitious Ideas are getting old, and tired. The competitors of this industry don't wait around.

If you don't have the funds to inject , at the pace required to continue a project, you have no choice but to Cancel it. Make do with what you have, and start fresh with a new target when possible. Which basically means years of pain in the "current market" during this gap period.

Same for failed architectures , i.e P4 and Bulldozer.. you just live in that pain for ~5yrs (bit less these days, but same basic principles apply)

 

[Henry swagger]

Your assumption is wrong. RU does not need massively wide cores. When conditions are favorable (like idling), it can create a massively wide virtual core by combining regular smaller cores of today. This one is from Ian:

"if each physical core is a 4-wide out-of-order design, if a thread running on a virtual core can utilize the resources of all four cores essentially making a giant 16-wide design".


Exactly. They combine regular cores.

And Intel didn't cancel RU because of any tech issues. It's a proven technology. They cancelled it because of lack of money & Pat is a dimwit.


It appears he's hit the pause button due to severe lack of money. They may resume it once they're in a better position. I don't think it should hurt Nova Lake plans. But anything after that, that uses RU might not get decent upgrades or might get delayed. Also, maybe NVL never had any RU to begin with. Not sure (just guessing), cos a X post recently said, P & E & Royal cores are all *different* cores.

Nova lake has cougar cove core not ru .. lets hope ru idea are implemented on cougar cove or future coves

 

[AcrosTinus]

See, this is how you know the whole thing was fantasy, or if it ever really existed that it was nothing like this.

No chance in the world that anyone takes a step from the 6/8/10 issue of today all the way to 24. Its completely ridiculous, there is not remotely that much parallelism to extract, you'd execute 24 instructions at once so rarely that the level of complexity could never pay back the huge increase in power consumption.

This is all a bunch of bs, if there ever was a "Royal Core" it sure as hell wasn't this. This is like saying that VW had some secret EV project that would get 1500 mile range out a battery half the cost of today's but still the same size and weight. You might want to believe it, but advancements don't happen like that.

This!
And that is the GAP in reality CPUs are a black box to maybe 99.995% of society, so someone like MILD and people that love to attach narratives to things, will generate claims and no one really knows if there is even an attachment to reality.
I would love for a channel to exist where an actual engineer goes through these videos and fact checks them...

That is the type of content I would watch.

 

[Nothingness]

Exactly. They combine regular cores.

And Intel didn't cancel RU because of any tech issues. It's a proven technology. They cancelled it because of lack of money & Pat is a dimwit.

What uarch proved that approach was worth it?

 

[Nothingness]

It's a bit hard to hit pause buttons on CPU Architectures. Unpausing 1, 2 3 yrs later and your once innovative, advanced , ambitious Ideas are getting old, and tired. The competitors of this industry don't wait around.

Though I agree pausing is quite a bad thing, not everything is lost when a project is cancelled: the company still owns the patents that protect the innovations made during the design development.

 

[jdubs03]

Though I agree pausing is quite a bad thing, not everything is lost when a project is cancelled: the company still owns the patents that protect the innovations made during the design development.

Exactly. Then it’s all about how they allocate the resources to implement those patents on their future designs. And indeed as of now they don’t have the ability to provide those resources. Because they’ve mismanaged the past decade+.

 

[Thunder 57]

It's a soft machines patented technology that Intel acquired long back. You can't just copy. Although, they can create derivatives as long as the derivatives don't blatantly infringe on the original patents. Like usual.

Don't they still cross license everything? Hence AMD being able to use AVX, etc.

...And Intel didn't cancel RU because of any tech issues. It's a proven technology. They cancelled it because of lack of money & Pat is a dimwit...

Is it now? There is some working prototype somewhere? Are you sure it isn't just fundamentaly broken in some way? Or is it like nuclear fusion where it is always "20 years away", or in this case say 5 years away? Maybe Pat is a dimwit, but there is no way you know more about the subject then him. When they do force him out, put your name in for CEO!

 

[SiliconFly]

Don't they still cross license everything? Hence AMD being able to use AVX, etc.

That’s only for the ISA. This is an entirely different technology. An abstraction layer that directly runs on top of the cores.

Is it now? There is some working prototype somewhere? Are you sure it isn't just fundamentaly broken in some way? Or is it like nuclear fusion where it is always "20 years away", or in this case say 5 years away? Maybe Pat is a dimwit, but there is no way you know more about the subject then him. When they do force him out, put your name in for CEO!

They bought it cos it had a 2nd gen proof of concept working perfectly fine as expected. Not nuclear fusion.

And kindly refrain from personal attacks. If you don’t agree with my posts, kindly ignore. And yes, I’ll apply for CEO. And you can be my assistant to fetch coffee.

 

[FlameTail]

Panther Lake leak:

Intel's next-gen Panther Lake listed with up to 16 CPU cores, 12 Xe3 GPU cores in Coreboot files - VideoCardz.com

Intel Core Ultra “Panther Lake” in four configurations? Coreboot files reveal up to 45W TDP, 240W PL4 limit. Intel’s next-generation consumer platform, known as Panther Lake, can already be found in Coreboot files. Coreboot is open-source software intended to replace proprietary firmware used...

videocardz.com

I would like to see a Panther Lake-V that combines the 4P+4LPE tile and 12 Xe3 tile. But that's not happening I guess.

 

[Thunder 57]

That’s only for the ISA. This is an entirely different technology. An abstraction layer that directly runs on top of the cores.

They bought it cos it had a 2nd gen proof of concept working perfectly fine as expected. Not nuclear fusion.

And kindly refrain from personal attacks. If you don’t agree with my posts, kindly ignore. And yes, I’ll apply for CEO. And you can be my assistant to fetch coffee.

What personal attack? I was just stating the facts. There is no way you can know more about why Pat canceled any project then he does himself. Your answer is just that he is a "dmiwit". OK. maybe, maybe not.

Where can I read about this 2nd generation proof of concept? There doesn't seem to much out there that I can see. Latest I see is that they were still "working on it". I don't see anything about it being "proven".

Intel is still working on VISC architecture 5 years after acquisition of Soft machines

Not many of you remember, but in 2016 Intel quietly bought a small startup : Soft Machines . That startup was developing a new type of processor architecture called VISC. New patent are surfacing showing that Intel is still actively developing the concept.

www.blopeur.com

 

[Nothingness]

They bought it cos it had a 2nd gen proof of concept working perfectly fine as expected. Not nuclear fusion.

My understanding is that nuclear fusion supposedly has had several proofs of concept recently, but nothing that makes it viable.

I bet Soft Machines was the same. It's easy to make simulations and FPGA that show great things. And then real implementation starts and you realize the frequency will go to the ground and power through the roof. With no real implementation, this is completely unproven technology.

That doesn't mean Intel got nothing out of the deal. They have at least engineers with new ideas (provided they're not fired...) and likely some patents.

 

[Philste]

Oh, nice. Finally something to look forward to!

Completely forgot that Meteor Lake also had independent Reviews on "Launch" on 14th of December. Quite unusual to see the list if actual SKUs and 3rd Party Reviews drop at the same time, but it makes sense for LNL to be the same. So 49 Hours to go. The weird thing is that Asus' presentation is one day later...

 

[maddie]

They bought it cos it had a 2nd gen proof of concept working perfectly fine as expected. Not nuclear fusion.

Not surprising, but you really have to provide better proof than "trust me bro".

 

[coercitiv]

The weird thing is that Asus' presentation is one day later...

This will be a challenge for Asus, how are they going to leak something if their presentation is a day alter? Stay tuned to find out!

 

[511]

This will be a challenge for Asus, how are they going to leak something if their presentation is a day alter? Stay tuned to find out!

Well they already leaked their models months ago

 

[Magio]

Panther Lake leak:

Intel's next-gen Panther Lake listed with up to 16 CPU cores, 12 Xe3 GPU cores in Coreboot files - VideoCardz.com

Intel Core Ultra “Panther Lake” in four configurations? Coreboot files reveal up to 45W TDP, 240W PL4 limit. Intel’s next-generation consumer platform, known as Panther Lake, can already be found in Coreboot files. Coreboot is open-source software intended to replace proprietary firmware used...

videocardz.com

I would like to see a Panther Lake-V that combines the 4P+4LPE tile and 12 Xe3 tile. But that's not happening I guess.

PTL-U gonna be such a bust at this rate. Especially as it's not like it will be dirt chip, being that it will be on 18A.

 

[mikk]

But should be small with only 4+0+4 cores. Other parts are on a different node.

 

[jdubs03]

I’d be curious as to how well PTL does with multithreaded workloads. It does seem like a big jump from Lunar Lake. Double the E cores with Darkmont. So I feel like saying it’s a bust is too pessimistic.

Though I rather there be 6p and 10e instead 4p-8e with the 4 lpe.

 

[511]

PTL-U gonna be such a bust at this rate. Especially as it's not like it will be dirt chip, being that it will be on 18A.

It is gonna be cheaper than LNL while performing slightly better I don't see it as a bust

 

[jdubs03]

It is gonna be cheaper than LNL while performing slightly better I don't see it as a bust

Seems like an understatement per my comment above. 2x the e cores of LNL and those e cores are Darkmont. Sure Panther Cover at most is a 10% IPC increase. But I feel like slightly better is underselling it.

 

[Philste]

Seems like an understatement per my comment above.

They are talking about PTL-U, which is 4+0+4 and 4Xe

 

[jdubs03]

They are talking about PTL-U, which is 4+0+4 and 4Xe

Ah yes my bad. Seems like that’s a focus on ST and battery life. MT isn’t getting much of a bump there.

 

[511]

Seems like an understatement per my comment above. 2x the e cores of LNL and those e cores are Darkmont. Sure Panther Cover at most is a 10% IPC increase. But I feel like slightly better is underselling it.

I meant PTL-U 4P+4LPE+4 Xe3 😅

 

[511]

Ah yes my bad. Seems like that’s a focus on ST and battery life. MT isn’t getting much of a bump there.

For the targeted market it is fine also 10% ipc in p core and 5-6% in e core is already fine for a yearly schedule

 

[mikk]

About PTL they feel confident they can bring up the margins by going back from TSMC to inhous manufacturing. PTL not before late next year (as expected though). Nothing here about Arrow Lake once again, just talking about Panther and Lunar Lake. Panther Lake probably will be even more important than LNL for them simply because it can cover a bigger mobile market segment with 8-16 cores and uses Intels node. This will be a highly significant generation for Intel.

So, the best thing that we're going to be able to do to improve margins is move to the next-generation products, bring those wafers home.

And that's exactly what Panther Lake is as we bring more of those wafers home. But that doesn't affect next year's margins very much, right, since the -- only starts late next year and really affects '26. So '24 and '25 are, I'll say, it's much more of a ground game, right?

And as we get to the 18A product generation that ramp significantly in '26, okay, happy day, bringing wafers home, the cost structure as we've done detailed analysis of 18A cost structure versus competitive nodes, we feel very good about the number of mass layers, the number of EUV layers, days per mass layer, defects and so on.

https://seekingalpha.com/article/4717982-intel-corporation-intc-deutsche-banks-2024-technology-conference-transcript