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

Page 314 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Tigerick

Senior member
Apr 1, 2022
686
576
106
PPT1.jpg
PPT2.jpg
PPT3.jpg



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

ModelCode-NameDateTDPNodeTilesMain TileCPULP E-CoreLLCGPUXe-cores
Core Ultra 100UMeteor LakeQ4 202315 - 57 WIntel 4 + N5 + N64tCPU2P + 8E212 MBIntel Graphics4
?Lunar LakeQ4 202417 - 30 WN3B + N62CPU + GPU & IMC4P + 4E08 MBArc8
?Panther LakeQ1 2026 ??Intel 18A + N3E3CPU + MC4P + 8E4?Arc12



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

Meteor LakeArrow Lake (20A)Arrow Lake (N3B)Arrow Lake Refresh (N3B)Lunar LakePanther Lake
PlatformMobile H/U OnlyDesktop OnlyDesktop & Mobile H&HXDesktop OnlyMobile U OnlyMobile H
Process NodeIntel 4Intel 20ATSMC N3BTSMC N3BTSMC N3BIntel 18A
DateQ4 2023Q1 2025 ?Desktop-Q4-2024
H&HX-Q1-2025
Q4 2025 ?Q4 2024Q1 2026 ?
Full Die6P + 8P6P + 8E ?8P + 16E8P + 32E4P + 4E4P + 8E
LLC24 MB24 MB ?36 MB ??8 MB?
tCPU66.48
tGPU44.45
SoC96.77
IOE44.45
Total252.15



Intel Core Ultra 100 - Meteor Lake

INTEL-CORE-100-ULTRA-METEOR-LAKE-OFFCIAL-SLIDE-2.jpg

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.)

Clockspeed.png
 

Attachments

  • PantherLake.png
    PantherLake.png
    283.5 KB · Views: 23,984
  • LNL.png
    LNL.png
    881.8 KB · Views: 25,456
Last edited:

Ghostsonplanets

Senior member
Mar 1, 2024
568
962
96
LNL BMG is 1024 ALUs at up to 1.8GHz@17 - 30W. And the comparison point, MTL-U7 165U, is 512 ALUs@2GHz. For Lunar Lake to pack double the ALUs and yet only be 50% faster is quite disappointing.

Perhaps this result might not even be due to BMG IP underperforming but rather from it not scaling well at low power levels and this performance being the best Intel could get over MTL-U (MTL-U at <20W just stops existing). Or maybe the software stack is still immature due to the changes BMG bring to the GPU design.

The most worrying thing though is that Intel performance figures for iGPU improvement Gen over Gen are usually exaggerated and specifically pushing the best result rather than the average improvement. MTL-H, for example, is said by Intel to be 2x over Xe LP 96EU. But in reality, it's around 1.5 to 2x.

So it's possible that LNL improvement in performance with real games is even lower than what is being said here.

How much faster do you think Strix iGPU will be than LNL @ 17W?
@uzzi38 said that LNL being 1.5x over MTL-U mean that it's not even beating Phoenix@15W. So certainly STX will be even faster @15W and much, much, much faster at 30W+.
 

DavidC1

Senior member
Dec 29, 2023
426
614
96
How much faster do you think Strix iGPU will be than LNL @ 17W?
According to gaming benchmarks, the 8 Xe core ARC is 50-60% faster in gaming compared to the 4 Xe core ones in the U parts, with the higher settings extending to 70-80%. It'll surely be more competitive than today, where compared to competition it's basically useless.

@Ghostsonplanets It is partly due to BMG IP underperforming. This is the part where many years ago they were aiming for "leadership performance per watt". The leaked roadmaps also show they wanted the top end G10 Battlemage to be <225W while positioning higher than A770 did. That implies something close to 2x perf/watt improvement.

The RTX 4060 Ti is 25% faster than A770 while having greater than 50% difference in power use, so it's already close to 2x the perf/watt. 2x perf watt even in late 2024/early 2025 would have been competitive. 2x perf/watt now would have been very competitive.

The fact that Intel is not talking much about GPU performance is unfortunately a sign. They are boasting about "AI performance", because that's a metric that is like a machine gun drawn on paper - it seems to matter but no one really cares.

Maybe they are sandbagging but such events are unfortunately very rare. Similar to strange lack of AMD's mention of Strix's iGPU. This unfortunately jives with the recent rumors of notebook Battlemage parts being cancelled.

At least I agree on an SoC level it looks decent. Battery life looks promising. It'll put a big dent again to those with lack of knowledge that believes lack of battery life is due to the ISA. If it wasn't for the "AI" which is basically an industry push to make censorship and spying/tracking easier then it would actually be a good chip for a personal computer too.
 

DavidC1

Senior member
Dec 29, 2023
426
614
96
20-30% lower power consumption vs. Snapdragon Elite and Phoenix for CPU purpose built for power consumption.

Well, it's not bad, but I was expecting bigger differences.
For an x86 part where people argue endlessly that it's due to the ISA and they can never get there, getting even with ARM will be revolutionary. Beating it will have their heads roll.

A reminder: https://www.anandtech.com/show/5770/lava-xolo-x900-review-the-first-intel-medfield-phone/6
45987.png

Oh what's that? x86 phone competitive with ARM phones? I thought they were hopelessly uncompetitive due to the ISA?

Battery life is a matter of managing power so they can stay idle as much as possible. TDP does not matter. Just because they aren't able to make good battery life parts most of the time is different from whether it's possible to do it. It certainly is possible to do it.
 
Last edited:

DavidC1

Senior member
Dec 29, 2023
426
614
96
Could be the info released so far is battery powered figures only. I'm sure that LNL gets better at 30W with the charger. It's not a big nuisance to play on a handheld with the charger attached.
If even their GPUs have a problem scaling down in power, it gives you an idea the problem with them reaching lower power envelopes is due to the mindset of the company.

Remember the process division saying they won't get a mobile competitive part until 18A-P and 14A?

Well, now apply that logic for every part of the company. CPUs, GPUs, NPUs. Every block on those 'PUs. FPUs, ALUs, branch prediction, circuit design, layout, etc etc.

The thought process is at the heart of what is wrong when it's wrong and what is right when it's right. Nothing else.

Another example: My 45nm Atom based handheld barely managed to get 5 hours of battery life(sometimes 3.5) with a tiny 6 inch screen and 24WHr battery. And it was slow as hell because of In-order architecture and PATA HDD.

The OoO Bay Trail based Dell Tablet I had got 6-8 hour battery on a 8 inch screen and 17WHr battery. And it was very responsive. Both on Windows.

-4.8-6.8W platform power use best in case with 6-inch screen versus
-2.1-2.8W platform power use with 8-inch screen and much faster.

The biggest difference is the company took an actual serious look and changed their focus. And voila, a real improvement. The Dell Bay Trail Tablet idled at 1.4W with screen-on!
 
Last edited:
  • Like
Reactions: lightmanek

Ghostsonplanets

Senior member
Mar 1, 2024
568
962
96
Lunar lake will be a great chip. I think most will be surprised especially the x86 is dead people.

On top of that, Lunar will support all games and apps. Do people not consider their Steam library/finance apps working with no issues a big selling point?

Intel will truly have a lot of wins with Lunar.
I think nobody denies this. It will be a fantastic SoC. It's just that some (like me) were expecting better in some areas. And x86 + proper DX12.2 GPU + Long battery life are big selling points when compared to Snap X Elite.

As for design wins, I think it was rumored that LNL has 3x design wins over MTL. Considering the amount of leaked Lunar Lake designs already, it's very likely that it's being widely adopted by the industry. Although we'll only see proper rollout at CES 25.
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,437
2,919
136
It has double the ALUs tho, Phoenix has the same number of ALUs as Rembrandt.
No, this would be the equivalent of 12 CU Phoenix vs 6 CU Rembrandt.
It's twice the shader cores and each is twice the shaders in itself (I think? BMG IP inherits PVC's SIMD16's).
Pretty bad result.
And how much did you expect to say this is a bad result?
It has 17W TDP, that's only 2W more than 165U while having 2x shader cores.
Do you really think that's enough to feed It at max clocks?
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,437
2,919
136
Well that's the thing, harsh power limits are nicer to wider configs.
True, and that's why you have a 50% increase.

BTW, Increasing CU count in case of AMD provides limited performance gains at 15W or even at 30W TDP.
Only 58% gain despite 200% more CU.
rog-ally-z1-3dmark-performance.jpg

Even at 30W you gain only 106% in TimeSpy.
rog-ally-z1-3dmark-turbo.jpg
So I don't see any reason why It should be different with Intel.
 
Last edited:

LightningZ71

Golden Member
Mar 10, 2017
1,675
1,955
136
We're looking at an expanded iGPU with the ability for a ton more throughput, but, the memory subsystem hasn't really increased that dramatically to go along with it. So far as we know there is no L4/MALL/system cache in place to provide very high throughput data streams.

The limited gains, despite the hefty improvements to the iGPU, are perfectly understandable. This is also why rational people don't expect a lot if improvement from Strix Point, even though it will have 33% more CUs with an at least somewhat improved RAND 3.5 implementation. It only gets a marginal increase in main memory throughput over Phoenix/Hawk, which was already demonstrated to be rather memory throughput limited in many benchmarks.
 

adroc_thurston

Diamond Member
Jul 2, 2023
3,337
4,817
96
the memory subsystem hasn't really increased that dramatically to go along with it.
?
So far as we know there is no L4/MALL/system cache in place to provide very high throughput data streams.
LNL has an 8MB SLC and bumps the mem setup to L5x@8533. lmao.
which was already demonstrated to be rather memory throughput limited in many benchmarks.
No?
None of them are. They're limited by power, that's the defining trait of every RDNA3 part out there.
 
  • Love
Reactions: spursindonesia

TESKATLIPOKA

Platinum Member
May 1, 2020
2,437
2,919
136
that's because PHX is scuffed.
RMB 3WGP to 6WGP was a massive-massive bump!
RMB is actually comparable to PHX, If you compare theoretical FLOPs increase vs TimeSpy increase. Basically perf gain is a bit less than 2/3 of FLOP increase in both cases.

Z1 vs Z1 Extreme
Specs: 4CU 2.5GHz vs 12CU 2.7GHz
Flops: 1280 vs 4147 GFlops (+224%)
TimeSpy 15W: 1502 vs 2368 (+58%)
TimeSpy 30W: 1527 vs 3141 (+106%)

660M vs 680M
Specs: 6CU 1.9GHz vs 12CU 2.4GHz
Flops: 1459 vs 3686 GFlops (+153%)
TimeSpy 40W: 1776 vs 2914 (+64%) 660M 680M

Some think that 50% increase in LNL's IGP performance is a FLOP, but I think It's not necessarily bad. It would be more interesting, If we knew the exact TimeSpy score.
 
Last edited:
  • Like
Reactions: buttersdeif

eek2121

Diamond Member
Aug 2, 2005
3,055
4,282
136
Lunar lake will be a great chip. I think most will be surprised especially the x86 is dead people.

On top of that, Lunar will support all games and apps. Do people not consider their Steam library/finance apps working with no issues a big selling point?

Intel will truly have a lot of wins with Lunar.
Types this earlier, but did not have time to finish:

The “x86 is dead” group kills me…with laughter! 🤣

I have said this before in other places, but if x86 were a bottleneck, Intel/AMD would make changes to the spec. Intel recently did this with X86-S, and AMD has done similar things (like x86-64)

Businesses evolve. Apple has changed to a completely new architecture 3 times so far, for example.
 

dullard

Elite Member
May 21, 2001
25,231
3,647
126

poke01

Golden Member
Mar 8, 2022
1,516
1,771
106
I think nobody denies this. It will be a fantastic SoC. It's just that some (like me) were expecting better in some areas. And x86 + proper DX12.2 GPU + Long battery life are big selling points when compared to Snap X Elite.

As for design wins, I think it was rumored that LNL has 3x design wins over MTL. Considering the amount of leaked Lunar Lake designs already, it's very likely that it's being widely adopted by the industry. Although we'll only see proper rollout at CES 25.
Yeah it goes both ways for me. The Apple CPUs are dead, ( ahh yes, because GWIII is the only chip architect in the world) and the X86 is dead ( not as efficient as ARM).

Both are wrong takes. Like Keller says it’s all in design, not the ISA.
 

Ghostsonplanets

Senior member
Mar 1, 2024
568
962
96

DAPUNISHER

Super Moderator CPU Forum Mod and Elite Member
Super Moderator
Aug 22, 2001
28,880
21,728
146
It is good to see more Meteor Lake ARC minis on the way. This AtomMan is pretty darned neat -


Set a frame cap on the games that are a little unbalanced and it will smooth right out. He could have used only rt reflections and XeSS with a 35-40fps cap in cyberpunk and it would have looked much better.
 

Joe NYC

Platinum Member
Jun 26, 2021
2,337
2,955
106
Types this earlier, but did not have time to finish:

The “x86 is dead” group kills me…with laughter! 🤣

I have said this before in other places, but if x86 were a bottleneck, Intel/AMD would make changes to the spec. Intel recently did this with X86-S, and AMD has done similar things (like x86-64)

Businesses evolve. Apple has changed to a completely new architecture 3 times so far, for example.

Exactly, I actually tweeted the same earlier today.

Apple is on the 4th instruction set and some people make it seem like Apple has the crystal ball and can see the future.
 
Last edited:

Ghostsonplanets

Senior member
Mar 1, 2024
568
962
96
It is good to see more Meteor Lake ARC minis on the way. This AtomMan is pretty darned neat -


Set a frame cap on the games that are a little unbalanced and it will smooth right out. He could have used only rt reflections and XeSS with a 35-40fps cap in cyberpunk and it would have looked much better.
Arc Xe 8 is actually really good. It's on par with a GTX 1650 Mobile.
 

John Carmack

Member
Sep 10, 2016
156
248
116
No need for rumors, Intel has stated more than 80 laptop designs. https://www.intel.com/content/www/u...ke-processors-arriving-q3-2024.html#gs.95gqjq

That's a lot lower than the 230 designs they claimed at Meteor Lake launch. Is packaging going to be the "issue" again?

I'm going to settle this nonsense.


That's the official number. Back in full launches such as Skylake, they mentioned over 800 design wins.

I have some doubt that Meteorlake will reach 230 since Raptorlake U takes the low end, but 3x MTL is NOT a small launch to those of you that can't see things unbiased.

To be honest, I don't care as much as I used to. Neither company doing well or badly affects me in any shape or form. It's same endless little-ahead and then little-behind for the past 30+ years. Us caring has same impact as Taylor Swift fans thinking they'll get 1-on-1 time with her.
 
  • Like
Reactions: spursindonesia