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.)

 

[SiliconFly]

He made the comparison, not me. I said you can't compare different chips like that and he disagreed. If he didn't intend to make the comparison, he had multiple opportunities to clarify. Stop trying to gaslight people.

Like I mentioned, this conversation has gone too far. A lot more has been said than whats actually necessary. Time to close it.

 

[Abwx]

Like I mentioned, this conversation has gone too far. A lot more has been said than whats actually necessary. Time to close it.

Right, let s go back to the essentials.
I once said that at normal powers AMD HPoint has 30-40% better efficency at similar perfs than MTL, seems that a handheld manufacturer just confirmed the numbers :

According to the manufacturer itself, the AMD Ryzen 7 8840U APU at 15-20W matches the Intel Core Ultra 155H in terms of CPU and gaming performance with the latter running at a higher 28 Watt TDP

OneXPlayer 2 Pro Handheld Launched: Ryzen 7 8840U Matches Intel 155H At Just 15W, $1049 US Price

OneXPlayer 2 Pro handheld has now gone official with the console featuring the AMD Ryzen 7 8840U and a starting price of $1049 US.

wccftech.com

 

[SiliconFly]

Right, let s go back to the essentials.
I once said that at normal powers AMD HPoint has 30-40% better efficency at similar perfs than MTL, seems that a handheld manufacturer just confirmed the numbers :

OneXPlayer 2 Pro Handheld Launched: Ryzen 7 8840U Matches Intel 155H At Just 15W, $1049 US Price

OneXPlayer 2 Pro handheld has now gone official with the console featuring the AMD Ryzen 7 8840U and a starting price of $1049 US.

wccftech.com

Well, at the heart of MTL is RWC which is seriously outdated and power hungry if you ask me. Trying to compare Zen4 with RWC at this point is not really worth it anymore (both performance & efficiency). I already feel MTL is a dud. I maybe wrong as Intel might come up with some future bios patches, µcode updates and/or better windows device drivers to address existing issues and improve performance and efficiency. But I'm sure MTL has lost the mindshare already, and thats not a good sign.

 

[Hitman928]

Right, let s go back to the essentials.
I once said that at normal powers AMD HPoint has 30-40% better efficency at similar perfs than MTL, seems that a handheld manufacturer just confirmed the numbers :

OneXPlayer 2 Pro Handheld Launched: Ryzen 7 8840U Matches Intel 155H At Just 15W, $1049 US Price

OneXPlayer 2 Pro handheld has now gone official with the console featuring the AMD Ryzen 7 8840U and a starting price of $1049 US.

wccftech.com

I am glad you are stating it this way now, it is more clear for everyone else and will be less controversial.

With that said, MTL does catch up to Hawk Point in efficiency but not until you get to significantly higher power levels where Hawk Point plateaus much harder in performance than MTL with increasing power. I think this just shows again the difference in design approaches and how Intel will need to change up how they approach CPU design to get back to be truly competitive. Hopefully Lunar Lake shows that they understand this and are making steps in that direction.

 

[dullard]

I think this just shows again the difference in design approaches and how Intel will need to change up how they approach CPU design to get back to be truly competitive. Hopefully Lunar Lake shows that they understand this and are making steps in that direction.

If the leaks are correct, that is what Intel is doing. No claims leaked so far directly from Intel indicate strong performance increases, all the claims indicate efficiency increases.

Page 142 - Discussion - Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads

Page 142 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[Anhiel]

Well, the new efficiency isn't surprising considering Intel had published node improvement since before 2020. So, moving from Intel 4 to Intel 3 to Intel 20A we are supposed to expect
1.18*1.15=1.357
I'ven't seen anything news worth looking at until the recent Intel 10.

Anyhow, since I've been nagged to look at the latest leaks and it's been a year since I last did some prediction analysis for ARL/LNL etc...
So according to latest leaks LNL (17W) MT is about 1.5x of MTL-U (15W) and boost clock is 2.78 GHz.

https://twitter.com/x/status/1766693553080807723

Intel's next-gen CPU boosts to 2.8 GHz without Hyper-Threading — Lunar Lake chip with eight cores, eight threads has a bigger L2 cache than the L3 cache

Early Lunar Lake ES silicon shows exciting specifications.

www.tomshardware.com

Not much details are given so lets gather the basics first
I'm taking Core Ultra 5 135U as reference for calculations as it has headroom so unlikely to suffer throttling, hence, more reliable test results.
All things equal the difference lies in p-/e-core configuration: MTL-U 2+8+(2) and LNL 4+4
* for relative power consumption LNL/MTL-U ~104/88=1._18_
17/15=1.1_3_ close so difference must be related to clockspeed?
* for relative MT LNL/MTL-U ~7.76/8.62=0.900 (nice)
The power limit is supposed to be 30W. At 2.78 GHz if I'm not wrong that's about 26.57W on the shifted (node difference) MTL curve. So about 10% is left... might be related to HT.
2.78 GHz over 1.9 GHz of Core Ultra 5 135U => 1.463x

Assumptions:
based on past gamer tests: HT off 1.24x IPC and (1/1.1) power
onboard cache estimating: 1.05x IPC (for apps not games)
tentative actual improvement: 1.05 IPC

adjusting baseline
1.463*0.900*1.24*1.05*1.05=1.8
adjusting LNL relative MT(1.8)=12.8917563

LNL/MTL-U: 12.8917563/8.62=1.49556337587
way too perfect fit, makes me wonder if the leak isn't another faked calculation leak
no need to adjust any of the above IPCs
so max raw IPC improvement 1.05*1.05*1.24=1.3671
I suppose this is the reason for past claimed improvement of up to 40%?

as for power consumption (ST) : LNL @2.78 GHz MTL-U @4.4GHz
(2.78/4.4)/1.357=0.46559924968178468546928384806056
So half as claimed.

as for Benchmark numbers
based on https://nanoreview.net/en/cpu-compare/intel-core-ultra-7-155u-vs-intel-core-ultra-5-135u
Core Ultra 5 135U has Cinebench R23 ST:1689 MT:10667
So LNL
ST: 1689*(2.78/4.4)*1.8=1921
MT: 10667*1.5=16000.5 (nice)
There's still this (possibly) 10% power head room left. Looking at the power curve at best another 5% difference for ST. But might be just consumed by iGPU, NPU or the SOC cores.

Anyhow, as expect there's huge potential for ARL/LNL ST given high clockspeed. Not really significantly different from my prediction from a year ago, though, but clearly more reliable numbers unless the leak is fake.

 

[Heartbreaker]

Intel's next-gen CPU boosts to 2.8 GHz without Hyper-Threading — Lunar Lake chip with eight cores, eight threads has a bigger L2 cache than the L3 cache

Early Lunar Lake ES silicon shows exciting specifications.

www.tomshardware.com

I wondered if anyone would be talking about that engineering sample.

I started seeing a lot of things indicating Lion Cove wouldn't have HT/SMT, and this seems to be the source. An engineering sample. I think it's a bit early to assume that HT is gone based on an early engineering sample, where it just might be disabled.

HT is a big benefit on advanced cores with many parallel sub units, that can be difficult to keep utilized with one thread.

 

[Henry swagger]

Intel's Global Desktop & Notebook PC CPU Shipments Reached 50 Million Units, More Than 3 Times Higher Than AMD & Apple Combined

Q4 2023 was strong for Intel as the chipmaker shipped 50 million CPUs in the desktop & notebook space, far exceeding AMD & Apple.

wccftech.com

Lunar lake will ship even more units.. intel is a different league compared to amd and appple when it comes to market share.. 80% total dominance 😀😂

 

[Hitman928]

Intel's Global Desktop & Notebook PC CPU Shipments Reached 50 Million Units, More Than 3 Times Higher Than AMD & Apple Combined

Q4 2023 was strong for Intel as the chipmaker shipped 50 million CPUs in the desktop & notebook space, far exceeding AMD & Apple.

wccftech.com

Lunar lake will ship even more units.. intel is a different league compared to amd and appple when it comes to market share.. 80% total dominance 😀😂

Where are they getting their numbers from? Their revenue numbers don't line up at all with Intel and AMD's earnings reports. Is this supposed to be full year revenue? Or is it quarterly revenue for the system the chips go in?

Edit: Looking at it again, I think it’s system revenue based upon the CPU provider. The revenue numbers seem way too low if this is the case, but the ratio is probably roughly correct.

 

[Geddagod]

I started seeing a lot of things indicating Lion Cove wouldn't have HT/SMT, and this seems to be the source. An engineering sample. I think it's a bit early to assume that HT is gone based on an early engineering sample, where it just might be disabled.

No, it's not based of just an ES sample. Otherwise one would think they would see "no HT" rumors for like every new generation lol.

 

[Heartbreaker]

No, it's not based of just an ES sample. Otherwise one would think they would see "no HT" rumors for like every new generation lol.

What else is based on?

 

[dullard]

I wondered if anyone would be talking about that engineering sample.

I started seeing a lot of things indicating Lion Cove wouldn't have HT/SMT, and this seems to be the source. An engineering sample. I think it's a bit early to assume that HT is gone based on an early engineering sample, where it just might be disabled.

HT is a big benefit on advanced cores with many parallel sub units, that can be difficult to keep utilized with one thread.

From what I've seen there are three leaks (although the middle leak could just be referring to the top leak below):

• An engineering sample of Arrow Lake that possibly is without hyper-threading (the wording is debatable on if that is actually what was intended). See the image below where something was disabled in BIOS and there were then only 8 IA cores and 8 threads. If the P cores were disabled as mentioned in the image, then there should have been 16 threads from the 16 E cores. But, this is a bit dubious as it is unclear, is an engineering sample, and twice says some cores are disabled. So nothing crystal clear there.
• Then there is this Arrow Lake leak where the interpretation depends on how you translate Chinese. Google translate either says "There is no hyper-threading and no LP-E cores" or it says "There is no hyper-threading? No LP-E cores" depending on how I hold up my phone to it. I don't speak Chinese to translate better. And it could just be referring to the leak above. https://www.bilibili.com/opus/904913222308462613
• Then there is that Lunar Lake leak you talk about. Since Arrow Lake and Lunar Lake share the same Lion Cove P cores, it would make sense that they would either both have hyper-threading or neither have it.

Possible reasons to drop hyper-threading:

• Its gains are small (often 10% to 30% if there are gains).
• Those gains diminish in real life as more and more cores are added. The gains begin to be mostly seen in benchmarks.
• There are hyper-threading losses of ~10% in some software, especially industrially important software (like simulations)
• There are hyper-threading losses due to spectre and meltdown

So possibly, as the gains diminish and the losses are getting worse, Intel has decided to drop hyper-threading. Also, there are rumors of a replacement, but that is yet to be seen.

 

[igor_kavinski]

What's interesting in that leak is that the ARL CPU can boot with only E-cores. Haven't heard that being possible with ADL/RPL. The life of benchmarkers should get easier without having to use thread affinity tricks to isolate the performance of only E-cores. It also gives one the flexibility to live by only on E-cores and thereby less heat and noise and turning on the P-cores occasionally for heavier and infrequent workloads.

 

[Geddagod]

From what I've seen there are three leaks (although the middle leak could just be referring to the top leak below):

• An engineering sample of Arrow Lake that possibly is without hyper-threading (the wording is debatable on if that is actually what was intended). See the image below where something was disabled in BIOS and there were then only 8 IA cores and 8 threads. If the P cores were disabled as mentioned in the image, then there should have been 16 threads from the 16 E cores. But, this is a bit dubious as it is unclear, is an engineering sample, and twice says some cores are disabled. So nothing crystal clear there.
• Then there is this Arrow Lake leak where the interpretation depends on how you translate Chinese. Google translate either says "There is no hyper-threading and no LP-E cores" or it says "There is no hyper-threading? No LP-E cores" depending on how I hold up my phone to it. I don't speak Chinese to translate better. And it could just be referring to the leak above. https://www.bilibili.com/opus/904913222308462613
• Then there is that Lunar Lake leak you talk about. Since Arrow Lake and Lunar Lake share the same Lion Cove P cores, it would make sense that they would either both have hyper-threading or neither have it.

Possible reasons to drop hyper-threading:

• Its gains are small (often 10% to 30% if there are gains).
• Those gains diminish in real life as more and more cores are added. The gains begin to be mostly seen in benchmarks.
• There are hyper-threading losses of ~10% in some software, especially industrially important software (like simulations)
• There are hyper-threading losses due to spectre and meltdown

So possibly, as the gains diminish and the losses are getting worse, Intel has decided to drop hyper-threading. Also, there are rumors of a replacement, but that is yet to be seen.

I think the wording for the first one is pretty clear tbh. Also, I think it's important to mention, that pretty much all the major leakers, and even MLID, agree that there is no HT on LNC.

 

[dullard]

I think the wording for the first one is pretty clear tbh. Also, I think it's important to mention, that pretty much all the major leakers, and even MLID, agree that there is no HT on LNC.

The unclear part to me is what was intentionally blanked out:

The pre-Alpha ARL-S CPU that ___________ is configured in BIOS to turn off the performance cores due to a hw issue.
Arrow Lake S 8+16+1, 8 IA Cores/8 threads (Disabled in BIOS)
Allowing 24 CPUs

What is erased there in the blank? If the blank part is "P-cores" (or something that controls P-cores), then why are 24 CPUs allowed with just 16 E cores? If the blank part is "Hyper-Threading" then it makes sense to have 24 CPUs with hyper-threading turned off. But then, if hyper-threading is turned off, it probably exists.

I personally don't like hyper-threading and think it is reaching the end of its useful lifespan. But, even with that, I'm not quite sure the wording is absolutely clear that hyper-threading is gone.

I guess it all depends what an "IA core" is. Has that been officially defined?

 

[Kepler_L2]

The unclear part to me is what was intentionally blanked out:

What is erased there in the blank? If the blank part is "P-cores" (or something that controls P-cores), then why are 24 CPUs allowed with just 16 E cores? If the blank part is "Hyper-Threading" then it makes sense to have 24 CPUs with hyper-threading turned off. But then, if hyper-threading is turned off, it probably exists.

I personally don't like hyper-threading and think it is reaching the end of its useful lifespan. But, even with that, I'm not quite sure the wording is absolutely clear that hyper-threading is gone.

I guess it all depends what an "IA core" is. Has that been officially defined?

Intel Architecture

 

[Hitman928]

The unclear part to me is what was intentionally blanked out:

What is erased there in the blank? If the blank part is "P-cores" (or something that controls P-cores), then why are 24 CPUs allowed with just 16 E cores? If the blank part is "Hyper-Threading" then it makes sense to have 24 CPUs with hyper-threading turned off. But then, if hyper-threading is turned off, it probably exists.

I personally don't like hyper-threading and think it is reaching the end of its useful lifespan. But, even with that, I'm not quite sure the wording is absolutely clear that hyper-threading is gone.

I guess it all depends what an "IA core" is. Has that been officially defined?

IA cores just refer to CPU cores (versus GT cores for graphics). I don't think there's a distinction in this regard between P and E cores. I read the paper that the performance cores are disabled due to a hardware issue and then the next line says that there are 8 P-cores and that 8 of the CPU cores (without HT) are turned off. So between the two lines, you know that all of the P-cores were turned off and that they don't have hyperthreading.

 

[uzzi38]

From what I've seen there are three leaks (although the middle leak could just be referring to the top leak below):

• An engineering sample of Arrow Lake that possibly is without hyper-threading (the wording is debatable on if that is actually what was intended). See the image below where something was disabled in BIOS and there were then only 8 IA cores and 8 threads. If the P cores were disabled as mentioned in the image, then there should have been 16 threads from the 16 E cores. But, this is a bit dubious as it is unclear, is an engineering sample, and twice says some cores are disabled. So nothing crystal clear there.
• Then there is this Arrow Lake leak where the interpretation depends on how you translate Chinese. Google translate either says "There is no hyper-threading and no LP-E cores" or it says "There is no hyper-threading? No LP-E cores" depending on how I hold up my phone to it. I don't speak Chinese to translate better. And it could just be referring to the leak above. https://www.bilibili.com/opus/904913222308462613
• Then there is that Lunar Lake leak you talk about. Since Arrow Lake and Lunar Lake share the same Lion Cove P cores, it would make sense that they would either both have hyper-threading or neither have it.

Possible reasons to drop hyper-threading:

• Its gains are small (often 10% to 30% if there are gains).
• Those gains diminish in real life as more and more cores are added. The gains begin to be mostly seen in benchmarks.
• There are hyper-threading losses of ~10% in some software, especially industrially important software (like simulations)
• There are hyper-threading losses due to spectre and meltdown

So possibly, as the gains diminish and the losses are getting worse, Intel has decided to drop hyper-threading. Also, there are rumors of a replacement, but that is yet to be seen.

The real reason to drop HT is most likely to make scheduling easier. On Intel platforms currently you effectively have to schedule around 3 things with different performance characteristics - the physical cores, the additional threads and the E cores. LP-E cores are getting added in on top as well (eventually, even if they're not present on ARL-S) is going to be the 4th too. It seriously is probably just easier to properly assign workloads when you only have to worry about the P and E cores.

 

[SiliconFly]

What else is based on?

Leaks from many reliable sources in the past. Most of them confirmed that LNC won't have HT. LNC is not an extension of GLC/RPC/RWC. It's a new architecture and HT/SMT wasn't baked in on purpose. Probably for improved ST.

 

[SiliconFly]

Also, there are rumors of a replacement, but that is yet to be seen.

About that.Thats still bugging me. On one hand, I'm not able to accept the fact those unearthly claims are even real. But on the other hand, those patents and info leaks just add up. I'm stumped.

 

[Goop_reformed]

My dad's cousin who works at Nintendo who used to work at Intel told me ArrowLake = RaptorLake + 15% - 20%

 

[SiliconFly]

What's interesting in that leak is that the ARL CPU can boot with only E-cores. Haven't heard that being possible with ADL/RPL. The life of benchmarkers should get easier without having to use thread affinity tricks to isolate the performance of only E-cores. It also gives one the flexibility to live by only on E-cores and thereby less heat and noise and turning on the P-cores occasionally for heavier and infrequent workloads.

Thats would be awesome. Intel should start listening to people. Lisa does. Pat has a lot to learn from her.

 

[SiliconFly]

My dad's cousin who works at Nintendo who used to work at Intel told me ArrowLake = RaptorLake + 15% - 20%

Thats in line with most of the prediction. But it's better to wait for an ES. MTL fell short of some predictions. So it's better to be cautious.

 

[dullard]

IA cores just refer to CPU cores (versus GT cores for graphics). I don't think there's a distinction in this regard between P and E cores. I read the paper that the performance cores are disabled due to a hardware issue and then the next line says that there are 8 P-cores and that 8 of the CPU cores (without HT) are turned off. So between the two lines, you know that all of the P-cores were turned off and that they don't have hyperthreading.

But that leaves me with why are 24 cores allowed if the P-cores are turned off?

 

[dullard]

The real reason to drop HT is most likely to make scheduling easier. On Intel platforms currently you effectively have to schedule around 3 things with different performance characteristics - the physical cores, the additional threads and the E cores. LP-E cores are getting added in on top as well (eventually, even if they're not present on ARL-S) is going to be the 4th too. It seriously is probably just easier to properly assign workloads when you only have to worry about the P and E cores.

Correct. I just lumped that in with HT gets worse with more cores. But you are correct that having 3 (4 in the case of Meteor Lake's LP-E cores) different types of performance characteristics is a scheduling nightmare.