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	8 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)	Arrow Lake Refresh (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Desktop Only	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2025 ?	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	8P + 32E	4P + 4E	4P + 8E
LLC	24 MB	24 MB ?	36 MB ?	?	8 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.)

 

[Timorous]

Intel Ultra 5 125H and Ryzen 7 7840HS tested at 65W: Intel with higher iGPU score but falls short in CPU tests - VideoCardz.com

Intel Core Ultra 5 125H at 65W shows lower CPU score and higher iGPU score than AMD Ryzen 7 7840HS Intel’s upcoming Core Ultra 100H series is just around the corner. Intel is going to show off its new laptop series called Meteor Lake in two days. This will introduce a new name for the […]

videocardz.com

Seems Intel is back in the game, at least for the iGPU.

One primary reason for getting a Radeon laptop just went whoosh!

My condolences, @adroc_thurston

View attachment 90294

That single core score is not great, a regression from RPL?

Also let's see how the iGPU manages in actual games since Intel seems to do relatively better in 3D Mark.

 

[inf64]

I still vividly remember MLID claiming MTLS having (supposedly) some insane IPC targets of 10-25%, while having much better power efficiency. By now we all know that IPC is pretty much the same, while power efficiency is not impressive at all. MTL is turning out to be a turd, except maybe the iGPU part, which finally can challenge Zen 4 APUs. Battery life is also abysmal, so I don't see how this thing is a better laptop part than RPL or Zen 4. Makes you wonder how screwed up intel execution is, with all those resources they have.

 

[Hitman928]

I still vividly remember MLID claiming MTLS having (supposedly) some insane IPC targets of 10-25%, while having much better power efficiency. By now we all know that IPC is pretty much the same, while power efficiency is not impressive at all. MTL is turning out to be a turd, except maybe the iGPU part, which finally can challenge Zen 4 APUs. Battery life is also abysmal, so I don't see how this thing is a better laptop part than RPL or Zen 4. Makes you wonder how screwed up intel execution is, with all those resources they have.

Battery life numbers (if true) weren't good but I will say that battery life is probably one of the last "performance" metrics that get dialed in with laptops and can be absolutely crushed by immature drivers/firmware. I'm still not that hopeful for MTL at this point, but battery life is the area that I would say has the highest chance of coming out better than what the leaks have indicated, don't have long to wait now.

 

[FlameTail]

I still vividly remember MLID claiming MTLS having (supposedly) some insane IPC targets of 10-25%, while having much better power efficiency. By now we all know that IPC is pretty much the same, while power efficiency is not impressive at all. MTL is turning out to be a turd, except maybe the iGPU part, which finally can challenge Zen 4 APUs. Battery life is also abysmal, so I don't see how this thing is a better laptop part than RPL or Zen 4. Makes you wonder how screwed up intel execution is, with all those resources they have.

MLID has been heavily saying in his recent videos that MTL efficiency is not good, and that laptop makers will be using AI as the main selling point for MTL laptops, not performance or efficiency.

 

[igor_kavinski]

Also let's see how the iGPU manages in actual games since Intel seems to do relatively better in 3D Mark.

I'm thinking it will be close in games even if it can't outright beat 780M. Raytracing might also be better though waiting for benchmarks to confirm that. But it's a good development in general. Games finally move a bit forward in the lowest setting graphical fidelity. Hopefully AMD fires back with a more powerful APU and not content with just keeping parity with Intel's graphics performance.

 

[cebri1]

MLID has been heavily saying in his recent videos that MTL efficiency is not good, and that laptop makers will be using AI as the main selling point for MTL laptops, not performance or efficiency.

MTL was never about delivering better performance on CPU side, back in September Intel was asked directly about the p-core and they said that Redwood Cove was a port from Raptor/Golden Cove with no IPC gains, with some minor IPC gains on the e-cores.

The 12500H boost to 4.7Ghz while the 125H only delivers 4.5Ghz. Normalize for frequency and the result difference in CB20 is under 3%, so within margin of error. If MLID said that MTL was going to provide a huge boost he was lying, as always.

MTL is about performance/watt. So far the only thing we know is that there are no efficiency gains at 65W but we are comparing a 45W TDP chip with a 28W TDP chip. Let's wait for official reviews at different power levels, if MTL performs the same as RPL at 30W then it's very bad news for Intel, if it beast RPL by at least 10% perf/watt then they are in the right direction.

 

[TESKATLIPOKA]

I'm thinking it will be close in games even if it can't outright beat 780M. Raytracing might also be better though waiting for benchmarks to confirm that. But it's a good development in general. Games finally move a bit forward in the lowest setting graphical fidelity. Hopefully AMD fires back with a more powerful APU and not content with just keeping parity with Intel's graphics performance.

Raytracing should be better, but there is not enough RAW power.
Doom Eternal should be playable >25FPS even with RT on.

 

[mikk]

9000 points in Firestrike with a 7 Xe core Arc (125H 65W), 8 Xe Arc in the Lenovo leak got 10K points, roughly 10% difference between 7 and 8 Xe cores.

 

[cebri1]

Raytracing should be better, but there is not enough RAW power.
Doom Eternal should be playable >25FPS even with RT on.

XeSS/FSR test is what I'm waiting for. I think that's where Xe-LPG can shine and save also on battery.

 

[Geddagod]

I still vividly remember MLID claiming MTLS having (supposedly) some insane IPC targets of 10-25%, while having much better power efficiency.

If you listened to MLID, that's on you lol

MTL is turning out to be a turd, except maybe the iGPU part, which finally can challenge Zen 4 APUs.

I thought it was better than Zen 4 iGPUs?

Battery life is also abysmal,

Are we sure it's just abysmal? Not apocalyptic? Horrendous? I mean if we are going to use hyperbole, why not go all the way?

so I don't see how this thing is a better laptop part than RPL or Zen 4.

Better battery life, AI capabilities, much stronger iGPU. It doesn't look like it's better than Zen 4 tho.

Makes you wonder how screwed up intel execution is, with all those resources they have.

Ice lake just vibing in the corner.
I think Intel's physical design team is cooked or something, idk.

if MTL performs the same as RPL at 30W then it's very bad news for Intel, if it beast RPL by at least 10% perf/watt then they are in the right direction.

It's a terrible sign for GNR if RWC in MTL is no more efficient than RPC.

 

[rtxtwt]

It's a terrible sign for GNR if RWC in MTL is no more efficient than RPC.

GNR might not be bad, all I've heard one of the reason why MTL efficiency miss target is all between the Tiles, which is a very early design and complexity is out of control, and it needs to be fixed in ARL.

Heh, and how much more expensive is it? I have my popcorn ready and waiting.

It's an old rumor. Intel4 is 80% more expensive than TSMC N6. But we don't know how much it would cost lately when compared to N5/N4.

Page 59 - Discussion - Leading Edge Foundry Node advances (TSMC, Samsung Foundry, Intel)

Page 59 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

some gossip suggested Intel4's manufacturing cost is ~1.8x times of TSMC N6's cost

 

[Tigerick]

GNR might not be bad, all I've heard one of the reason why MTL efficiency miss target is all between the Tiles, which is a very early design and complexity is out of control, and it needs to be fixed in ARL.

Yep, that's what I suspected. Each tile needs their own power tail for current and SoC tile needs 3 D2D to maintain constant connectivity. The power penalty would be pretty high, that's why Intel try to put Atom LPE into SoC and hope that the system would shut-down tCPU and tGPU tiles. But I am not sure OS would work with the design and as soon as you wake up the system, all tiles need to wake up again....

That's why I am not sure ArrowLake would solve the design issues. And that's the reason Lunar Lake, Panther Lake and Nova Lake are all reducing the tiles from 4 to 2 only...

 

[Geddagod]

GNR might not be bad, all I've heard one of the reason why MTL efficiency miss target is all between the Tiles, which is a very early design and complexity is out of control, and it needs to be fixed in ARL.

Should be easy to figure out, if MTL reports core vs package power, I think?

 

[Geddagod]

Yep, that's what I suspected. Each tile needs their own power tail for current and SoC tile needs 3 D2D to maintain constant connectivity. The power penalty would be pretty high, that's why Intel try to put Atom LPE into SoC and hope that the system would shut-down tCPU and tGPU tiles. But I am not sure OS would work with the design and as soon as you wake up the system, all tiles need to wake up again....

That's why I am not sure ArrowLake would solve the design issues. And that's the reason Lunar Lake, Panther Lake and Nova Lake are all reducing the tiles from 4 to 2 only...

I don't think it's the foveros power penalty tbh. But we will see soon ig.

 

[cebri1]

It's a terrible sign for GNR if RWC in MTL is no more efficient than RPC.

RWC = RPC

Confirmed several times already. That's why when people talk about regression from RPL to MTL when frequencies do not match, people are either trolling or ignorant.

What To Expect From Meteor Lake With Intel's Robert Hallock And Dan Rogers

Intel unveiled its forthcoming tile-based Meteor Lake processor architecture at its Innovation 2023 event this week, with more cores, a beefed up GPU, and de...

www.youtube.com

 

[Tigerick]

I don't think it's the foveros power penalty tbh. But we will see soon ig.

If monolithic, SoC only needs to maintain one power curve. With 4 tiles, then the system needs to maintain 4 power curves. And I believe there is min V to maintain the connectivity, that's why power penalty would be pretty high....Any electrical enginner want to pitch in?

 

[Geddagod]

RWC = RPC

On a new node.

That's why when people talk about regression from RPL to MTL when frequencies do not match, people are either trolling or ignorant.

What?

If monolithic, SoC only needs to maintain one power curve. With 4 tiles, then the system needs to maintain 4 power curves

Even in a monolithic system, the iGPU might have a different ideal power limit than the cores part of the CPU. Idk what you mean by this exactly.

And I believe there is min V to maintain the connectivity, that's why power penalty would be pretty high...

True, but I wouldn't think it would consume an extreme amount of power or anything. Foveros looks to be pretty efficient, at least that's the impression I got from hotchips

 

[Geddagod]

I love the flip flopping on ARL's cache setup by Xino on Twitter lmao

 

[cebri1]

Exclusive First Look! Intel Meteor Lake Core Ultra With Dan Rogers And Robert Hallock

We're going to be running new Intel Core Ultra Meteor Lake laptops LIVE with very special guests Robert Hallock and Dan Rogers of Intel. Join us at 4:30 ET W...

www.youtube.com

Today in about 4 hours.

 

[Geddagod]

Exclusive First Look! Intel Meteor Lake Core Ultra With Dan Rogers And Robert Hallock

We're going to be running new Intel Core Ultra Meteor Lake laptops LIVE with very special guests Robert Hallock and Dan Rogers of Intel. Join us at 4:30 ET W...

www.youtube.com

Today in about 4 hours.

On reddit it's titled "exclusive performance first look" so hoping we get something interesting

 

[FlameTail]

So when exactly is Meteor Lake dropping?

It's 13th December today.

 

[Geddagod]

So when exactly is Meteor Lake dropping?

It's 13th December today.

Intel's AI day is tmmrw, so I'm expecting tmmrw. Today is just a sneak peek I'm guessing

 

[cebri1]

On reddit it's titled "exclusive performance first look" so hoping we get something interesting

I'm not expecting much, at first I though it was scheduled for tomorrow after they unveil the new processors, which would have been a nice touch from Intel to provide additional transparency. But doing it before the event... I mean I'm not expecting anything mindblowing performance wise, I do want to see the GPU in action and also see what the NPU can do, but perf/watt numbers in the 25-45W range is what I'm most interested in.

 

[dullard]

Intel's AI day is tmmrw, so I'm expecting tmmrw. Today is just a sneak peek I'm guessing

More specifically the AI event is 10 AM EST tomorrow: https://www.intc.com/news-events/pr.../media-alert-join-intels-ai-everywhere-launch

 

[Mopetar]

So when exactly is Meteor Lake dropping?

It's 13th December today.

Supposed to be the 14th, so tomorrow. No idea what the product availability will be like, but there should be enough for people itching for something to stuff their stocking this Christmas.