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

Tigerick · Aug 22, 2022

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.)

Joe NYC · Sep 4, 2024

DavidC1 said:
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.

You could be onto something. Renaming 20A to 18A, cancelling 20A sounds better than announcing 20A is a year late.

Doug S · Sep 4, 2024

poke01 said:
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.

It isn't the OS alone. Apple was specifically choosing lower power Intel CPUs - sometimes they got Intel to make them special SKUs when they didn't like the specs of the ones Intel offered others.

It would be difficult to figure out how much of the difference was due to the chip, and how much was due to the OS. Though I think everyone knows that Windows is kind of a resource pig that always seems to have random stuff happening in the background even when there are no user tasks running.

DavidC1 · Sep 4, 2024

Doug S said:
It would be difficult to figure out how much of the difference was due to the chip, and how much was due to the OS. Though I think everyone knows that Windows is kind of a resource pig that always seems to have random stuff happening in the background even when there are no user tasks running.

That's Wintel for ya. They blame each other for their problems, but they are a perfect couple because both of them having problems.

9949asd · Sep 4, 2024

9949asd said:
P95 237w only 75c, and can oc to p5.5.

also，idle only 9w.

jdubs03 · Sep 4, 2024

9949asd said:
also，idle only 9w.

Hmm those are certainly better results than what HXL on twitter was showing. Was showing 885 and 15709. Since it’s an Intel to Intel comparison I’ll give it more benefit of the doubt than not.

9949asd · Sep 4, 2024

jdubs03 said:
Hmm those are certainly better results than what HXL on twitter was showing. Was showing 885 and 15709. Since it’s an Intel to Intel comparison I’ll give it more benefit of the doubt than not.

919 is p5.5，stock is 5.2. ARL now just like zen5 doing in cpuz. In the r23 r24 will look way better.

jdubs03 · Sep 4, 2024

9949asd said:
919 is p5.5，stock is 5.2. ARL now just like zen5 doing in cpuz. In the r23 r24 will look way better.

I think you mean all-core is 5.2Ghz.

That score is decent though. There is a small bump in single thread and an okayish bump in multi thread.

237W in P95 is a massive improvement though. Hopefully that bares out.

9949asd · Sep 5, 2024

jdubs03 said:
I think you mean all-core is 5.2Ghz.
View attachment 106859
That score is decent though. There is a small bump in single thread and an okayish bump in multi thread.

237W in P95 is a massive improvement though. Hopefully that bares out.

yes, I’m talking about all core.

cebri1 · Sep 5, 2024

9949asd said:
P95 237w only 75c, and can oc to p5.5.

That’s a massive improvement over the 14700K. Around 100 watts less?

jdubs03 · Sep 5, 2024

cebri1 said:
That’s a massive improvement over the 14700K. Around 100 watts less?

Notebookcheck has it at 400W for P95 with an external monitor. Anandtech has peak power at 397W.

cebri1 · Sep 5, 2024

jdubs03 said:
Notebookcheck has it at 400W for P95 with an external monitor. Anandtech has peak power at 397W.

Mmmm the one I checked got to around 350W.

jdubs03 · Sep 5, 2024

cebri1 said:
Mmmm the one I checked got to around 350W.

Would make more sense tbh. Notebookcheck is system power and it’s close and value to Anandtech, so my assumption is it’s also system power.

Just caught this:

https://twitter.com/x/status/1831488980979986685

Wasn’t the 4P-8E-4LPE-12Xe sku PTL-P? And the 4P-0E-4LPE, PTL-U?
Looks like a couple of typos.

511 · Sep 5, 2024

jdubs03 said:
Would make more sense tbh. Notebookcheck is system power and it’s close and value to Anandtech, so my assumption is it’s also system power.

Just caught this:

https://twitter.com/x/status/1831488980979986685

Wasn’t the 4P-8E-4LPE-12Xe sku PTL-P? And the 4P-0E-4LPE, PTL-U?
Looks like a couple of typos.

From this 4Xe3 core on Intel 3 i guess it is for system with dGPU there is no reason to have 12Xe3 ina system with dGPU and tiling makes it flexible for them to do it

PTL-H 4+8+4+4Xe 25W
PTL-H 4+8+4+12Xe 25W
PTL-H 4+0+4+4Xe 25W

DavidC1 · Sep 5, 2024

jdubs03 said:
Wasn’t the 4P-8E-4LPE-12Xe sku PTL-P? And the 4P-0E-4LPE, PTL-U?
Looks like a couple of typos.

Well it says 25W, so it indeed is PTL-H.

511 · Sep 5, 2024

https://twitter.com/x/status/1831571711931904103

igor_kavinski · Sep 5, 2024

Since it's a TSMC node, I think it will give maybe 5% extra overclocked performance at 300W, if the dial can even be turned all the way up to that?

naukkis · Sep 5, 2024

majord said:
This really makes no sense.

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

ST is important, and LNL has that in droves it seems, but this strength will drop off pretty quickly beyond 4 threads , and be completely DOA byond 8 obviously compared to other offerings. That's a bit lacklustre in 2024. And it means it gets flogged by other U class chips from a generation ago . Matched by some 2 gen's ago. That's a problem that's hard to shake. I'm certainly not forking out for something that can't do more real work than my 2+yr old machine, just because it can chomp through a ST workload faster.

IF ST is strong enough to also be performant in MT with a low core count, PLUS it's efficient enough to still have good perf/watt whilde doing it, then your point is valid, but none of these are true.. the ST is basically in line with everything else give or take. and it comes at a consumption we don't know yet.. Intel certainly ain't mentioning it in their slides

Actually Intel is mentioning in their slides. 3x thread performance vs MTL. That means Lunarlake 8 threads have faster max threads MT performance than MTL 22 threads @ 17w.

511 · Sep 5, 2024

naukkis said:
Actually Intel is mentioning in their slides. 3x thread performance vs MTL. That means Lunarlake 8 threads have faster max threads MT performance than MTL 22 threads @ 17w.

Wonder what 14 thread will do in ARL-H 👀 btw 4 threads are not even on ring bus

naukkis · Sep 5, 2024

511 said:
Wonder what 14 thread will do in ARL-H 👀 btw 4 threads are not even on ring bus

To further increase perf/watt on low power mobile they need to get rid of ringbus from p-cores too. They already have acceptaböe cross-latency between clusters.

SiliconFly · Sep 5, 2024

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

These business strategies were introduced by none other than Apple.

SiliconFly · Sep 5, 2024

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

I think they should sell off IFS altogether. They don't need it anymore. Their client/server product divisions are working amazingly well with TSMC!

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

Sad but true!

coercitiv · Sep 5, 2024

naukkis said:
Actually Intel is mentioning in their slides. 3x thread performance vs MTL. That means Lunarlake 8 threads have faster max threads MT performance than MTL 22 threads @ 17w.

Marketing speak aside, LNL is 10% faster than MTL @ 17W and 6% slower @ 23W. It loses steam fast, and that's with a node advantage. For the design target of 9W as declared by Intel the chip is very nice, but for 17W+ it could definitely use a bit more horsepower, either 4P6E or even 4P8E. That's why it's disappointing they're not following it up with another iteration.

Think of it this way, MTL loses to Hawk Point 8P/16T at 23W. LNL loses to MTL at 23W. We'll have independent numbers soon so I'd rather not try to extrapolate what happens at 17W, but it may not be the clear-cut win some were expecting.

511 · Sep 5, 2024

SiliconFly said:
I think they should sell off IFS altogether. They don't need it anymore. Their client/server product divisions are working amazingly well with TSMC!

Nope selling IFS is not an option for them they already invested too much in fabs also it was due to their foundry they have been so dominating cause they can fab it but those things have become inefficient since 14nm they need efficient foundry to drive them the shareholders are now vultures trying to force the seperation for their benifit
The manufacturing part of TSMC/Samsung success is due to their country backing as well look at China funding SMIC either US can do the same or suck off the loss and loose the only leading edge foundry they have manufacturing
requires government support in any country

naukkis · Sep 5, 2024

coercitiv said:
Marketing speak aside, LNL is 10% faster than MTL @ 17W and 6% slower @ 23W. It loses steam fast, and that's with a node advantage. For the design target of 9W as declared by Intel the chip is very nice, but for 17W+ it could definitely use a bit more horsepower, either 4P6E or even 4P8E. That's why it's disappointing they're not following it up with another iteration.

Think of it this way, MTL loses to Hawk Point 8P/16T at 23W. LNL loses to MTL at 23W. We'll have independent numbers soon so I'd rather not try to extrapolate what happens at 17W, but it may not be the clear-cut win some were expecting.

As it's chip for thin&light devices performance at high power is irrelevant. Also trading fast thread speed to more combined low thread speed ain't winning scheme for good user experience. Intel is at least going to right direction, x86 devices might actually be competitive after they got more efficient interconnect to p-cores too. Now even Lunarlake is still about 2-fold behind true mobile devices for sustained low thread performance/watt.

Magio · Sep 5, 2024

SiliconFly said:
I think they should sell off IFS altogether. They don't need it anymore. Their client/server product divisions are working amazingly well with TSMC!

I disagree. Intel's foundries, in the recent past, have dragged Intel down. The whole 10nm debacle lost them their lead, cost them billions upon billions and derailed their product strategy for years. But foundries are a valuable asset, and their foundries' prior success was a big part of why they dominated for so long.

The option to become a fabless chipmaker would have been a tempting one a few years ago but it's too late now. Gelsinger bet too much on their foundries since.

And regardless of if a conversion to fabless chipmaker is even feasible without dooming the whole company, in my opinion it would be a terrible call. Intel is one of three foundries worldwide to even be near the cutting edge and they're crucially the only US company among them. That makes Intel, while they continue to pursue cutting edge nodes, an actual geopolitical asset to the US. You don't give up that sort of status lightly.

I also have to say I'm not quite sure why there's all this doom and gloom about 18A just yet, even in light of the recent Reuters report. If all Broadcom said to Intel is that 18A is not yet ready for HVM (which is the only thing that's being reported)... I get that's not good news and the fact it's being reported suggests it's not the news Intel wanted. But more than a year away from the first product slated to release on 18A I have trouble seeing why HVM readiness is an absolute necessity today. If we were getting that news some time into 2025 then I'd get it but as things stand it sounds like a negative news but hardly a catastrophic one.

Also can't help but feel like it may have been filtered to Reuters by the people at Intel who do want to spin off IFS as a way to pressure the people that don't, but that's entirely speculation from me.

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

Senior member

Attachments

Platinum Member

Platinum Member

Senior member

Member

Senior member

Member

Senior member

Member

Senior member

Senior member

Senior member

Senior member

Senior member

Senior member

Senior member

Lifer

Senior member

Senior member

Senior member

Golden Member

Golden Member

Diamond Member

Senior member

Senior member

Member