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.)

 

[dullard]

If I could do all the work I can do my desktop, on a thin and light laptop , why wouldn't I want that..

Because it life just doesn't work that way. A workstation CPU will blow away the thin and light laptop in MT.

• You might get close to 1000 points in CineBench MT with the very best thin and light laptop. https://images.anandtech.com/graphs/graph21485/136438.png
• But even a relatively subpar 13900K desktop CPU can get over 2000 points: https://www.anandtech.com/bench/CPU-2024/3382
• And the Threadripper 7995WX gets almost 6000. https://www.anandtech.com/bench/CPU-2024/3382

If you want productivity, it isn't in a thin and light laptop. Even with a great CPU, there just isn't the power. And if there was the power, it is no longer thin and light.

 

[majord]

Because it life just doesn't work that way. A workstation CPU will blow away the thin and light laptop in MT.

• You might get close to 1000 points in CineBench MT with the very best thin and light laptop. https://images.anandtech.com/graphs/graph21485/136438.png
• But even a relatively subpar 13900K desktop CPU can get over 2000 points: https://www.anandtech.com/bench/CPU-2024/3382
• And the Threadripper 7995WX gets almost 6000. https://www.anandtech.com/bench/CPU-2024/3382

If you want productivity, it isn't in a thin and light laptop. Even with a great CPU, there just isn't the power. And if there was the power, it is no longer thin and light.

That's not the point at all.. It was a Theoretical Question.. IF an SoC existed that could make it possible, why wouldn't I want it ? It was a response to the notion that MT is 'useless' in a thin and light..

Of course it doesn't exist , but anything that can get closer to this 'nirvana' is equally desirable isn't it. I'm not going to say.. "Oh well, it's a thin and light laptop.. I'll just accept it can't do these sort of workloads" .

 

[poke01]

Apple uses their own custom hardware designs and their choice of select/premium components when compared to OEMs like HP, Dell, Lenovo, etc. All vertically integrated. Intel+OEM laptops can never match Apple's perfection.

It’s not Apples hardware only, when Apple used Intel they also got better standby than Windows. It’s just macOS being better optimised at such things.

Windows can be better but Microsoft needs to put in some work

 

[Hulk]

Dell Lunar buy page

Dell has their buy page up, @Hulk you buying?

I'm not seeing any customization options for display or SSD. I'm looking for a better display (with touch) and 1TB SSD.

Don't worry I will be buying an LL system but I do want to see what else will be available!

 

[SiliconFly]

I'm not seeing any customization options for display or SSD. I'm looking for a better display (with touch) and 1TB SSD.

Don't worry I will be buying an LL system but I do want to see what else will be available!

I think it's better to wait for more reviews before making a decision.

 

[StinkyPinky]

Dell Lunar buy page

Dell has their buy page up, @Hulk you buying?

16GB of ram for $1400....yuck. These OEM's always act like ram is some precious resource mined from the martian moons or something.

 

[jpiniero]

16GB of ram for $1400....yuck. These OEM's always act like ram is some precious resource mined from the martian moons or something.

Remember it's on package... so the ram you get is dictated by the chip Dell chooses to put in.

 

[poke01]

16GB of ram for $1400....yuck. These OEM's always act like ram is some precious resource mined from the martian moons or something.

It’s even worse when that price also doesn’t give a greater than FHD+ panel nor its OLED.

 

[ikjadoon]

Two big pieces of news, courtesy of reddit:

Intel cancels 20A Arrow Lake, most (all?) ARL CPUs are TSMC

Intel: "With this decision, the Arrow Lake processor family will be built primarily using external partners and packaged by Intel Foundry."

Intel CFO opens the door to corporate breakup – eventually

Asked at an investment conference Wednesday whether Intel’s semiconductor design and factory businesses might eventually be entirely separate companies, Chief Financial Officer David Zinsner said it’s important that they operate separately. He said each side of the business will have internal systems that maintain that independence.

“Where it goes from there is difficult to say,” he said at the Citi Global Technology Conference. “Our goal here is to drive maximum shareholder value.”

 

[dr1337]

Why would anybody want to buy a high performance part and then increase battery life by decreasing its performance? Strange.

Anyone that wants a reasonable battery life in their laptop? The same people that set their Windows laptop to battery saver mode? I mean heck I even use battery saver on my phone.

A big perk of more efficient chips for mobile devices is that they can last longer or be super fast depending on which situation is more practical for the user.

 

[controlflow]

Intel cancels 20A Arrow Lake, most (all?) ARL CPUs are TSMC

I think this makes a lot of sense to do.

ARL was almost entirely TSMC from the beginning. 20A is just early 18A sans some features/libraries. Given that 18A PDK 1.0 is out, Intel is broke and 18A yields are decent enough, it makes sense to just skip 20A and go to 18A. Does not make sense to put anymore money into an inferior 20A at this point if 18A is coming along well enough.

The real purpose of these Intel 4 and 20A half nodes is to be a pipe cleaner and be there for risk mitigation.

 

[jdubs03]

Anyone that wants a reasonable battery life in their laptop? The same people that set their Windows laptop to battery saver mode? I mean heck I even use battery saver on my phone.

A big perk of more efficient chips for mobile devices is that they can last longer or be super fast depending on which situation is more practical for the user.

I do this as well. I always use battery saver on my phone and laptops when unplugged.

 

[DavidC1]

@controlflow Yet Broadcom engineers said 18A is not up to their standards. Something is not right.

It’s even worse when that price also doesn’t give a greater than FHD+ panel nor its OLED.

For some like me, I would pay extra for an IPS FHD panel over 1440p OLED. If you purposely chose Lunarlake for the battery life part, then go OLED, they are bit contrary to each other.

I can't get extra battery life in an easy way while the oohs and aahs you get from a better screen lasts a month, before it fades away. I'm even ok with 1366x768 screen on this Yoga.

 

[Hulk]

Two big pieces of news, courtesy of reddit:

Intel cancels 20A Arrow Lake, most (all?) ARL CPUs are TSMC



Intel CFO opens the door to corporate breakup – eventually

Hmm. 5 nodes in 4 years is looking like pie-in-the-sky.

 

[DavidC1]

Asus Zenbook S 14 OLED UX5406S impressions (2024 series, Intel Lunar Lake)

This article gathers my thoughts on the Asus Zenbook S 14 OLED UX5406S series, the mid-2024 addition to the Zenbook lineup of ultrabooks. This is not a

www.ultrabookreview.com

Having used this early sample over the last few weeks, I have no complaints. This unit felt snappy with light use on battery power, lasted for a long while on a charge, and didn’t lose battery while in sleep mode even for a few days. I haven’t noticed any wifi issues while resuming from sleep either.

You don't have those problems with other versions of Windows when you first reset your laptop or reinstall your Windows. It's when you get endless updates that take forever to install each time and gets little bit wrong each time you get processes that aren't behaving properly.

Speaking of efficiency, though, I will note that this early sample ran notably longer with light use than the Intel Meteor Lake and AMD Strix Point units. 12+ hours of Netflix streaming and 7-10 hours of daily use multitasking should be easily achieved here in actual use. Sure, Asus claims up to 24 hours, but you know how these marketing claims go in real use…

With a honking 72WHr battery, you better get those figures. I can get 5-6 hours on my Yoga with a 35WHr battery. I'd basically be getting Lunarlake figures with a 72WHr one.

Unless your disposable income allows you to get a new $1500 USD laptop every 2 years, battery life is king over any display.

I hope to see devices that are under 2.5lbs, and gets 15+ hour web browsing time. That means 2W screen-on idle and <3W browsing power.

 

[Joe NYC]

Does not make sense to put anymore money into an inferior 20A at this point if 18A is coming along well enough.

It's people also. Meaning Intel can lay off more people.

As far as money, Intel is apparently spending $14 billion per year on wafers from TSMC.

If some of that wafer processing was done inhouse (on 20A), there would be more revenue for Intel foundry and less in expenses to TSMC.

On lower end, Intel could accelerate phase out of Raptor Lake, if 20A is cost competitive. But with ARL being on N3, the cost will be high, higher than Raptor Lake, so Intel will not be able to move away from Raptor Lake fast enough.

 

[controlflow]

It's people also. Meaning Intel can lay off more people.

As far as money, Intel is apparently spending $14 billion per year on wafers from TSMC.

If some of that wafer processing was done inhouse (on 20A), there would be more revenue for Intel foundry and less in expenses to TSMC.

On lower end, Intel could accelerate phase out of Raptor Lake, if 20A is cost competitive. But with ARL being on N3, the cost will be high, higher than Raptor Lake, so Intel will not be able to move away from Raptor Lake fast enough.

There is no way Intel has enough 20A capacity to really move the needle and it would cost far too much to really ramp that now.
Going to TSMC isn't cheap but its still much cheaper in the short term than trying to ramp up a process node that you plan to move off by next year anyway. Just take a look at how badly Intel got burned last quarter by being forced to ramp Intel 3 in Ireland earlier than expected. The upfront costs for these things are brutal and Intel can't shoulder it right now.

At least Intel 3 is a foundry node that in theory can be used for other customers, Intel 20A is not. Why spend money on 20A and then have it be superseded right away and then be unable to even sell as a foundry node?

Given their financials, I don't think they can afford to do anything else but to just ramp 18A and hope for the best.

 

[DavidC1]

Given their financials, I don't think they can afford to do anything else but to just ramp 18A and hope for the best.

The whole reason for bad financials is because they are notorious for having super optimistic demand projections, so they overspent and wasted money when it could have been used for the bad times.

Let's say if Pat's team suspended dividend and share buybacks in mid-2021. Then they would have had what, $40-50 billion more?

Remember what they say about experienced vs inexperienced traders. The majority sell on the LOW and buy on the HIGH. They follow the crowd and FOMO.

It is exactly what the management and CEO of the inventor of microprocessors and until recently one of the most successful companies did.

 

[controlflow]

@controlflow Yet Broadcom engineers said 18A is not up to their standards. Something is not right.

Ian Cutress commented on that news piece today, I'm not really sure it is particularly conclusive.

When you are working with test chips from a foundry, it is normal to have a back and forth exchange with them like that. Broadcom allegedly even said they have not concluded their evaluation. Intel says they have a dozen customers evaluating 18A. Companies generally do not announce who their suppliers are so it will be sometime before we know if anyone actually bites.

Gelsinger quoted a defect density of 0.40/cm^2, and since 18A is ~3 quarters or so from ramping, that seems like right where you'd expect it right now. They have a few more quarters to get things in order.

 

[Joe NYC]

There is no way Intel has enough 20A capacity to really move the needle and it would cost far too much to really ramp that now.
Going to TSMC isn't cheap but its still much cheaper in the short term than trying to ramp up a process node that you plan to move off by next year anyway. Just take a look at how badly Intel got burned last quarter by being forced to ramp Intel 3 in Ireland earlier than expected. The upfront costs for these things are brutal and Intel can't shoulder it right now.

At least Intel 3 is a foundry node that in theory can be used for other customers, Intel 20A is not. Why spend money on 20A and then have it be superseded right away and then be unable to even sell as a foundry node?

Given their financials, I don't think they can afford to do anything else but to just ramp 18A and hope for the best.

The tools for 20A and 18A are the same. It is just a different recipe.

Ramping it surely has costs, but it's not like switch to 18A will be instant. Panther Lake is more of a Meteor Lake replacement, so ARL may be shipping for a long time. Years (plural), not just 1 year.

Fab underutilization is a real thing. Intel said in several of the past earnings reports that Intel is incurring underutilization charges.

Prior to "shaking the box" and yet again reshuffling the financials, and reporting categories, "underutilization charges" showed up on the earnings reports.

 

[ikjadoon]

I'm also suspicious of "Idle power". They are using 7W with the screen off really? The Snapdragon uses 4W and 6W? Even the M3 uses 3W?

My 2017 Kabylake Yoga does 3W with screen-on. So my 2017 14nm Intel is superior to Apple M3 right? The problem is both Intel's and NBC's testing methodology is flawed. On battery the power management is better than on AC.

You can see some NBC tests showing "idle" power measurements at 4W, but the notebook is getting 20 hours on a 55WHr battery. Clearly it's way under 4W.

https://www.notebookcheck.net/Sayonara-Alcantara-Microsoft-Surface-Laptop-3-13-5-inch-Core-i5-Review.454000.0.html

-4.4W Idle minimum.
-1.3W Reader(so not 100% idle) battery consumption. 34 hours on a 45W battery. Actually 1.2W since you can't use 100% of battery.

Now some of Notebookcheck's dedicated efficiency tests are DC side if & when explicitly noted, but "Power Consumption" charts in all reviews (ex. below) are AC side.

Especially at low power draw, conversion losses will be significant (and variable between different power supplies). It will not match up with the battery consumption. Idle Minimum = lowest display brightness, maximum energy saving mode, all radios off, bundled laptop-specific power adapter, AC side

We measure directly between the power supply and the socket and can therefore also make statements about the standard power supply.

It's tough to get anything useful from these charts when AC adapter quality can be variable & neither idle measurement neatly cancels out for normalizing, so it's a crapshoot both ways.

 

[DavidC1]

The tools for 20A and 18A are the same. It is just a different recipe.

Ramping it surely has costs, but it's not like switch to 18A will be instant. Panther Lake is more of a Meteor Lake replacement, so ARL may be shipping for a long time. Years (plural), not just 1 year.

Can we really say this 18A is 20A rebranded with extra features?

Because just based on the naming and the BSPD it should have got the 30% density gains. RibbonFET should have gotten them 15% perf.

Then

I think what they did was they found doing 15% + 10% perf and 30% + x% in density was too optimistic, so they took 20A and added extra libraries and called it 18A.

They may as well be bad as Samsung in lying about the processes now.

 

[controlflow]

Let's say if Pat's team suspended dividend and share buybacks in mid-2021. Then they would have had what, $40-50 billion more?

Gelsinger suspended the stock buy backs almost immediately after becoming CEO. Intel hasn't done a stock buy back since June 2021.

Agree that they should have cut the dividend sooner but even cutting that wouldn't get them anywhere near $40B-50B saved. Not to forget that the main reason Intel has broad institutional ownership was because of the dividend. You remove that dividend and then you might face a massive sell off, delisting from various indices and weaker borrowing capacity.

The reality is that Intel was in deep trouble by 2021 already and things were going to get ugly almost no matter what. That was the debt that Intel built from years of stupid decision making the previous decade. It is easy to talk about what Pat should have done with the benefit of hindsight, but I doubt there are many that could have walked into Intel in 2021 and fixed it right up without a whole lot of pain.

 

[DavidC1]

The reality is that Intel was in deep trouble by 2021 already and things were going to get ugly almost no matter what. That was the debt that Intel built from years of stupid decision making the previous decade. It is easy to talk about what Pat should have done with the benefit of hindsight, but I doubt there are many that could have walked into Intel in 2021 and fixed it right up without a whole lot of pain.

The reason they went overboard with fabs and spending during that time was because of their deeply flawed expectations about demand. They did not expect substantial demand reductions anywhere.

They are cutting out giving out free drinks and fruits in their cafeteria and cancelling Innovation Event. Surely even extra $1 billion in their coffers would have helped.

Look at this. Gelsinger came in Feb 2021 right? This is FY 2022 report:

Intel Reports Fourth-Quarter and Full-Year 2021 Financial Results

www.intc.com

The board declared a quarterly dividend of $0.365 per share on the company’s common stock, which will be payable on March 1 to shareholders of record as of February 7.

Intel increased dividend a full year after Gelsinger.
2021: For the full year, the company generated $30.0 billion of cash from operations, paid dividends of $5.6 billion, and used $2.4 billion to repurchase 39.5 million shares of stock.
2022: For the full year, the company generated $15.4 billion in cash from operations and paid dividends of $6.0 billion.
2023: For the full year, the company generated $11.5 billion in cash from operations and paid dividends of $3.1 billion.
2024: In the first quarter, the company used $1.2 billion in cash from operations and paid dividends of $0.5 billion.

$15.2 billion in dividends that could have been saved. And 2.4 billion in share buy backs. Total = $17.6 billion.

 

[9949asd]

P95 237w only 75c, and can oc to p5.5.