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.)

 

[SiliconFly]

I wouldn't put full trust in MLID videos.

He claimed 19-28% for Lion Cove or something? And something also ridiculous for Zen 5? Him and many others also lavished praise for RDNA3. And what was the joke? "38% for Int and 68% on Skymont for FP is a decent improvement according to MLID sources"? I can't say he's totally wrong in everything either but when it comes to details he has been. RDNA3 misunderstanding tells me that him, Adored, and others do not have a full understanding of architecture.

This is also if Intel lives through their financial situation, which I believe less and less every day.

He's the same dimwit who kept saying LNC will have 30% IPC uplift. And ironically, he was saying the same about Zen 5 too!

Yea that’s fair. I certainly don’t hold his words as 100 pct true. But I’d hope he actually did discuss with this person what he said he did. I mean over on Reddit we’ve seen my images that I posted where the info stated about Royal Core sounded very similar.

I get you on the improvement with Skymont though, if there is any silver lining it’s that IPC improvement is massive. The e-core side has clearly innovated more than the p-core recently and if given the room they will probably surpass them; and is probably why p-core is trying to block it.

I wouldn't write off P cores yet! The final IPC uplift of LNC in ARL is still not known. And I still stand by my claims of 15% to 20% (or maybe more) IPC uplift and an overall ST performance uplift of 10% to ~20% over previous gen equivalent 14900K. It's still possible. And we'll know soon enough.

He is doing right thing. Intel is not in good shape and there´s no time to waste money on complicated exotic projects (e.g. Ponte Vecchio/Rialto Bridge/Rentable Unit/ etc.) History proves that this was the case with others as well ( Bulldozer with speculative multithreading / Nvidia Denver core /DEC Alpha 21464 )

Intel has always been a microprocessor-first company. Thats their core competency which they've already lost to competition. They're not the best microprocessor company in the world anymore (already lagging behind Apple & AMD). They need to prioritize projects like royal core first instead of prioritizing stuff like AI. Pat is turning out to be Intel's 3rd knucklehead.

Keep in mind this project started several years ago. It wasn’t just an embryonic idea here, it was well established and moving apace.

In this instance: Jim Keller left after two years, 4 top senior established architects in their Advanced Architecture Development Group left to start their own company (all of them had over 15 years at Intel). It’s pretty understandable if you invested years of your time and intellect on a project for it to be abruptly cancelled, to then be willing to leave.

Regarding the rumors; usually when there is smoke there is fire, and this is unlikely to just be a coincidence.

Exactly!

Maybe but the fool couldn't figure out if Jim Keller was the one to listen to rather than the imposter Renduchintala so Keller left out of exasperation. With Keller advising Pat, things could have turned out a LOT different.

Sadly, I think Keller left before Pat came in.

Pat is a bad manager 🤨.. why camcle royal core smh

Actually I too get the same feeling. He's an excellent engineer. But it appears he doesn't know how to count/manage money.

 

[AcrosTinus]

I know many here aren’t fans here of MILD, but he has commented on the Royal Core cancellation theme. If what he’s saying in the video is legit, it sounded pretty cool:
Dynamic splitting of threads per core. 2x IPC of RPL. Says this was from a face-to-face meeting with a verified Intel employee aware of the project.

AGAIN, Royal Core is not real, it is not a product. You cannot cancel something that never was. This is a normal R&D exercise in the labs. After things are deemed viable or not, aspects are extracted and used in other products.

How can people be this gullible?
He is basically extracting you for views, feeding you stories that meet your interests, he could basically make things up and people will listen to him.

In reality this video if from a guy that is known to fabricate stuff, claiming to allegedly be talking to a unverified Intel source that somehow spilled the beans to a anti Intel youtuber.

No wonder media literacy is low and idiots can influence stock markets. If fairy tails are given the same weight as researched outlets we are doomed, RIP AnandTech. You inspired me to go through electrical engineering much love.

 

[Kocicak]

AGAIN, Royal Core is not real, it is not a product. You cannot cancel something that never was. This is a normal R&D exercise in the labs.

What are you talking about? You consider "a real product" only what has been officially anounced by the manufacturer?

So you REALLY KNOW future planned Intel CPUs are you are sure that Royal core (the 6 core 6-24 thread rumored CPU) has never been seriously planned as one? Are you Intel employee?

May I ask you to explain typology of projects in Intel depending on how seriously is the outcome of the project considered to become a real product in the future and how are these projects being approved, financed and managed?

Thank you.

 

[Magio]

The primary driver for better battery life from those products is their ST power, so I don't understand how LnL will achieve these expectations.

The SoC may provide good idle , good HW accelerated video consumption numbers, but to get these into the real world battery life advantage will take very disciplined platform engineering, and without a program like Evo on steroids , and focusing their marketing on specific devices I don't think there'll be much to show for it. but we'll see.

The key for battery life on LNL (besides SoC and media engine improvements) will be how many loads they'll be able to confine to the Skymont cluster (and of course how efficient that cluster is). LNC seems like a marked improvement in ST efficiency but it still doesn't quite compare to Oryon or especially to Apple Silicon in terms of PPW.

But if you can avoid having the LNC cores turning on as much as possible, and limit them mostly to heavy loads (where, let's be real, battery life is mostly a factor of TDP/battery size), then that's not too big a problem as far as battery life is concerned.

 

[cannedlake240]

Intel has always been a microprocessor-first company

Intel was once a memory company. Then their founding fathers managed to (very successfully) pivot to Intel to the CPU business. I guess Pat, inspired by this past glory and the Nvidia's profits, now wants Intel to become an AI + Foundry company selling CPUs on the side. The problem is, just this week he admitted that Intel isn't competing with the best of Nvidia, Google and AMD anytime soon. Hope Falcon Shores is really promising, otherwise dissolving a design team after half a decade of work on an innovative uarch is just bizarre... The AI hype might end any time now while Client & DC CPUs aren't going anywhere

 

[511]

And Bob Swan should have fixed it. But he didn't care about the future of Intel foundries. Instead, the knucklehead was more interested in Intel stock prices buy backs.

Agreed

Like I mentioned before, both the biggest knuckleheads in Intel's history, Brian Krzanich & Bob Swan, worked real hard to run Intel into the ground and they succeeded. Now, Pat Gelsinger is trying out a lot of new tricks, but unfortunately, the results have been subpar.

At least Pat is trying he is not being successful

For example:
Intel 4/3 don't have enough capacity.
Intel 20A is delayed.
They're bleeding server market share.

Intel 3 is a very good node it's just they lack machines it's PPW is around N3B/E and density similar to N4 but the lack of capacity is an issue
20A is like Intel 4 a public beta node for few sku for learnings

Their desktop client product quality is subpar.
Their mobile client MTL is a colossal failure (imho).

Meteor lake is a failure it sold 15 million+ units 😂
Tiger lake was a success sold 100Million+

This isn't the Intel I knew. If Pat doesn't pull a rabbit out of his hat real soon, he too might join the list as the third biggest knucklehead in Intel's history following Brian Krzanich & Bob Swan.

His rabbit is named 18A/PTL/CLF

Imagine, we can call them *3 idiots*!

4 we already have paul simply refused Apple iphone SOC

 

[Thunder 57]

And Bob Swan should have fixed it. But he didn't care about the future of Intel foundries. Instead, the knucklehead was more interested in Intel stock prices buy backs.

Like I mentioned before, both the biggest knuckleheads in Intel's history, Brian Krzanich & Bob Swan, worked real hard to run Intel into the ground and they succeeded. Now, Pat Gelsinger is trying out a lot of new tricks, but unfortunately, the results have been subpar.

For example:
Intel 4/3 don't have enough capacity.
Intel 20A is delayed.
Intel 18A not enough customers.
They're bleeding server market share.
Their desktop client product quality is subpar.
Their mobile client MTL is a colossal failure (imho).

This isn't the Intel I knew. If Pat doesn't pull a rabbit out of his hat real soon, he too might join the list as the third biggest knucklehead in Intel's history following Brian Krzanich & Bob Swan.

Imagine, we can call them *3 idiots*!

That's the problem with an interim CEO.

Let's play a game, I'm the golden trickster fish and you're Pat Lucky. You found me while fishing, and you get one wish to save Intel from the ones below:

• Intel gets best SoC of 2025 with Lunar Lake and Arrow Lake wins vs. Zen 5 in performance and efficiency (ARL refresh included)
• Intel matches or beats TSMC with their fifth node in 4 years, and is set to win foundry contracts for short and mid term

Why am I the trickster golden fish, you ask? Because when you pick one, the other is automatically excluded. In a way this matches reality, it would take a miracle for both of the above to happen at the same time. So which one do you pick to save Intel? 🐡

I'd say A is more likely but B would be far better for Intel in the long run.

 

[AcrosTinus]

What are you talking about? You consider "a real product" only what has been officially anounced by the manufacturer?

So you REALLY KNOW future planned Intel CPUs are you are sure that Royal core (the 6 core 6-24 thread rumored CPU) has never been seriously planned as one? Are you Intel employee?

May I ask you to explain typology of projects in Intel depending on how seriously is the outcome of the project considered to become a real product in the future and how are these projects being approved, financed and managed?

Thank you.

We have no real insight, what MLID is offering is conjecture and is worth nothing, he has no ethics.
And the first sentence is on the money, a real product is what is announced by the manufacturer and can be bought by consumers or businesses.
Everything else is internal development that is not meant for the public, meaning when an R&D exercise does not pan out you cannot speak about a cancelled product, closer would be a potential idea that was deemed not viable. Even then the confidence in such information should be extremely low.

But in the age of fake news and misinformation you decide your media diet.... good luck.

 

[cannedlake240]

already have paul simply refused

They somehow had a relatively smooth transition to 22nm, but the signs of trouble brewing were already there, it's said. This new GPU/AI push better not be a repeat of the mobile debacle they had a decade ago

 

[Kocicak]

what MLID is offering is conjecture and is worth nothing, he has no ethics.

A person delivering a message and the content of the message and two different things and even a pathological liar sometimes tells the truth.

So you automatically presuming that the information coming from the source you do not like is false is simply an incorrect inference.

 

[majord]

The key for battery life on LNL (besides SoC and media engine improvements) will be how many loads they'll be able to confine to the Skymont cluster (and of course how efficient that cluster is). LNC seems like a marked improvement in ST efficiency but it still doesn't quite compare to Oryon or especially to Apple Silicon in terms of PPW.

But if you can avoid having the LNC cores turning on as much as possible, and limit them mostly to heavy loads (where, let's be real, battery life is mostly a factor of TDP/battery size), then that's not too big a problem as far as battery life is concerned.

I agree, but if you can confine a decent amount to E cores, then begs the questions why 4+4? , why not 2+6 or something. This is the reason I find big Little architectures..(and that includes AMD APU's C core arrangement ) a bit of an enigma at times . They make sense fundamentally, but how does one decide what balance is appropriate? It's really trickly in the wonderfully unpredictable world of x86 ,. Even if it's not optimal, it won't show in ST MT benchmarks

 

[Magio]

I agree, but if you can confine a decent amount to E cores, then begs the questions why 4+4? , why not 2+6 or something.

You probably need the quad core LNC cluster for gaming, for example. And of course, LNL won't be a gaming powerhouse but it does have a beefy GPU for its TDP so clearly that was part of the focus.

Personally I feel like LNL is a very well balanced chip on paper, looking to strike a balance between efficiency and breadth of viable use cases. Of course, now it has to confirm that "in the field".

 

[511]

Personally I feel like LNL is a very well balanced chip on paper, looking to strike a balance between efficiency and breadth of viable use cases. Of course, now it has to confirm that "in the field".

3 days before the NDA is lifted

 

[SiliconFly]

AGAIN, Royal Core is not real, it is not a product.

How can people be this gullible?

I beg to differ. I think the name royal core might even be imaginary. And the term rentable units might have been cooked-up by one of the so called leakers. But the concept is real & Intel working on it was real. It was even confirmed by Dr. Ian Cutress. And comparing timelines, it's evident it was Jim Keller's baby during his time at Intel. Not sure about it's current state.

It's even clear from MLID's video where he says he asked keller "how to combine multiple threads to make a single thread to boost IPC". Stupid question. He says keller replied it's actually the opposite which is very true. I'm actually appalled by MLID's ignorance. If he can't grasp such basic concepts, we need to question each & every assessment he makes based on his leaks.

Just to clarify:

Hyper-threading (HT) is running multiple threads in a single core. Generally speaking, HT makes a single core appear as 2 cores to the OS to facilitate efficient multi-threading.

Rentable Units (RU) on the other hand is simply running a single thread simultaneously on multiple cores. A single thread is sliced into multiple smaller threadlets, and all threadlets are simultaneously executed using multiple physical RU cores (and/or virtual RU cores according to implementation).

It was supposed to be the holy grail of ST performance uplift. And Intel clearly had it within its grasp without a doubt. And for heaven's sake, when they acquired it, the proof-of-concept implementations was into 3rd generation. Meaning, very mature and very real. And now thanks to the 3rd knucklehead Pat, it appears it's cancelled or postponed or split into pieces and added incrementally into products. I'm starting to getting a feeling we may never see proper RU in the near future.

 

[SiliconFly]

His rabbit is named 18A/PTL/CLF

Lets hope it pans out well. With this kinda execution, keeping hopes up is becoming more and more difficult.

 

[Thunder 57]

It's even clear from MLID's video where he says he asked keller "how to combine multiple threads to make a single thread to boost IPC". Stupid question. He says keller replied it's actually the opposite which is very true. I'm actually appalled by MLID's ignorance. If he can't grasp such basic concepts, we need to question each & every assessment he makes based on his leaks.

What are you talking about? Reverse Hyperthreading is total a thing! /s

 

[igor_kavinski]

What are you talking about? Reverse Hyperthreading is total a thing! /s

We have enough deep knowledgeable techies on this forum. We should create our own version and patent it and then we can buy AT from Future PLC and do whatever we want with it!

Come on, guys! Let's ASSEMBLE!!!!

 

[coercitiv]

3 days before the NDA is lifted

Oh, nice. Finally something to look forward to!

 

[Magio]

3 days before the NDA is lifted

Well I doubt we'll know *everything* on September 3rd. Mobile platforms can only really be judged by the quality of the designs they'll be integrated in, so it will take until LNL design wins are revealed and released into the hand of reviewers.

But we'll know a lot more as it launches and during IFA in general.

 

[igor_kavinski]

Let's play a game, I'm the golden trickster fish and you're Pat Lucky. You found me while fishing, and you get one wish to save Intel from the ones below:

• Intel gets best SoC of 2025 with Lunar Lake and Arrow Lake wins vs. Zen 5 in performance and efficiency (ARL refresh included)
• Intel matches or beats TSMC with their fifth node in 4 years, and is set to win foundry contracts for short and mid term

Why am I the trickster golden fish, you ask? Because when you pick one, the other is automatically excluded. In a way this matches reality, it would take a miracle for both of the above to happen at the same time. So which one do you pick to save Intel? 🐡

TBH, I don't think any of those will happen.

Lunar Lake will be popular with OEMs the same way MTL was (so nothing out of the ordinary for Intel).

Arrow Lake will sell decently but won't set any performance records. The RAM overclocking ability will satisfy the existing LGA1700 overclockers into upgrading and trying to reach new DDR5 heights.

As for matching TSMC? Maybe but no way large volume customers are gonna bet on them ONLY for their precious product. Their fabs/process may be used as a backup or to try out less ambitious product SKUs to see how well Intel serves them and their needs.

If there is going to be a miracle, it's going to be something secret that Intel has managed to keep hidden from public view so far.

 

[511]

Come on, guys! Let's ASSEMBLE!!!!

ARM assembly or x86 Assembly pick your poison or RISC-V as a hardcore option

 

[511]

Well I doubt we'll know *everything* on September 3rd. Mobile platforms can only really be judged by the quality of the designs they'll be integrated in, so it will take until LNL design wins are revealed and released into the hand of reviewers.

But we'll know a lot more as it launches and during IFA in general.

David Huang already has microbenchmarked everything he will release the data on 3rd

 

[Magio]

As for matching TSMC? Maybe but no way large volume customers are gonna bet on them ONLY for their precious product. Their fabs/process may be used as a backup or to try out less ambitious product SKUs to see how well Intel serves them and their needs.

Capacity will come, or at least the plans for that aren't kept secret:

Intel 18A and Intel 3 should remain attractive nodes for a good while (including their refinements 18A-P, Intel 3T/E/PT), let's not forget that even today the best chips from Qualcomm, AMD and Nvidia are still on TSMC's 4nm processes and those are very lucrative contracts for TSMC. So if they can ramp the capacity on those as planned, there will be contracts to be found aplenty.

If those are rolled out without many issues, despite the initially limited capacity I can only imagine the prospect of uncontested leadership with 14A would attract extremely lucrative contracts. Of course, great nodes would also be a boon for the product side of the company albeit Intel will have to be careful not to withhold too much of their capacity for themselves if they want to attract the biggest guns in the chipmaking industry.

Of course I'm not saying any of that will come to be, I'm no sightseer. But the plan is there and it is extremely promising on paper.

 

[igor_kavinski]

Here's a prediction: Despite lacking hyperthreading, LNL (both P-cores and E-cores) will perform decently in 1080p gaming workloads because Skymont is roughly "Raptor Cove"-powerful and Lion Cove is supposed to be "Raptor Cove++". So any game threads juggling between heterogeneous cores shouldn't result in a wide discrepancy in performance. If this proves to be true and I can find a decent model that I can get with monthly installments, Imma gettin' it!

 

[igor_kavinski]

But the plan is there and it is extremely promising on paper.

Their execution on promised plans so far has been meh. I guess they think their investors are the type who obsess about two birds in the bush rather than the one in their hands.