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.)

 

[ondma]

MTL may sell. But it's still a mediocre product.

Well, I'm now thoroughly convinced anything with RWC is mediocre (including Granite Rapids).

LNC+SKT is their real future.

Skymont seems like a great product, the kind of improvement Intel needed in their P cores. But how much improved is Lion Cove, actually, over RWC for desktop use? I dont really care that much about the E cores for desktop. Just give me a big jump in P core performance and efficiency.

 

[dullard]

You are either mis-interpreting or deliberately obfuscating the point of my post. It is most certainly about both performance and release dates.

I have been speaking about your specific post here:

Problem is that ARL is late to the party. What reason will their to buy it if AMD has a comparable or better performing chip out 6 months earlier?.

ARL is not 6 months after Zen 5. I am simply trying to correct that fallacy. If I misinterpreted or obfuscated "6 months earlier", please help correct me.

If you wish to have a discussion about performance, go ahead. Last we left it, I used your claim of Zen 5 having comparable performance better than ARL and went with that. Do you have anything concrete to add to your original performance estimate? We will find out in a few months which of those claims is true. But that has nothing to do with your timeline post that we have been discussing for days. A post which was based on ARL being 6 months after Zen 5.

If you want to discuss performance and do not have anything performance-wise to add (especially unlikely given that you have no idea what you are talking about with respect to hyperthreading), how about we go back to your first post on this. What happens to Intel if:
1) Arrow Lake is bad compared to Zen 5
2) Arrow Lake is comparable
3) Arrow Lake beats Zen 5

 

[SiliconFly]

I have been speaking about your specific post here:

ARL is not 6 months after Zen 5. I am simply trying to correct that fallacy. If I misinterpreted or obfuscated "6 months earlier", please help correct me.

If you wish to have a discussion about performance, go ahead. Last we left it, I used your claim of Zen 5 having comparable performance better than ARL and went with that. Do you have anything concrete to add to your original performance estimate? We will find out in a few months which of those claims is true. But that has nothing to do with your timeline post that we have been discussing for days. A post which was based on ARL being 6 months after Zen 5.

If you want to discuss performance and do not have anything performance-wise to add (especially unlikely given that you have no idea what you are talking about with respect to hyperthreading), how about we go back to your first post on this. What happens to Intel if:
1) Arrow Lake is bad compared to Zen 5
2) Arrow Lake is comparable
3) Arrow Lake beats Zen 5

My two cents. ARL and Zen5 are separated by around 2.5 months at best. Not 6.

ARL top 8P+16E is going to take the MT crown over Zen5 top X3D part due to too many performant E cores. ST is yet to be seen. But I think it's gonna be very close.

(i.e, Skymonts in ARL are gonna be faster than the Skymonts in LNL).

Skymont seems like a great product, the kind of improvement Intel needed in their P cores. But how much improved is Lion Cove, actually, over RWC for desktop use? I don't really care that much about the E cores for desktop. Just give me a big jump in P core performance and efficiency.

Like I mentioned once before, it appears the P core team has reached an evolutionary dead end. Now that they have ported the design to industry standard tools and made it more modular and node agnostic, they need to slide and dice it and come up with something awesome. Otherwise, the entire team needs to be dissolved. They're very close to becoming irrelevant if they don't have anything good to show real soon.

Intel P core team sucks! (as of now)

 

[TwistedAndy]

Like I mentioned once before, it appears the P core team has reached an evolutionary dead end. Now that they have ported the design to industry standard tools and made it more modular and node agnostic, they need to slide and dice it and come up with something awesome. Otherwise, the entire team needs to be dissolved. They're very close to becoming irrelevant if they don't have anything good to show real soon.

Yes, but things are more complicated here.

It's a classic situation for big projects (P-core/Cove architecture). Usually, they have some kind of a mix of ancient legacy stuff and some new parts. Once the project grows, it becomes harder and harder to make some changes and implement new features. At a certain point, you have to do some refactoring, remove some old legacy structures, make it more flexible, modular, decoupled, etc., just to be able to move forward.

I think that's the reason why Intel decided to clean up the architecture in Lion Cove and switch from "Sea of Fubs" to "Sea of Cells". Usually, the project's success depends on how successful you are at dealing with the technical debt and refactoring.

At the same time, Skymont is a way more perspective architecture just because it's smaller and much more innovative.

 

[SiliconFly]

...dealing with the technical debt and refactoring.

My point too.

Once they came out with RWC, the whole world fell in love with the P core team for their awesomeness. And now that they're about to deliver a massive IPC uplift with LNC, there's this overwhelming feeling of admiration for the P core team! Wonder what incredible things they're gonna accomplish next.

 

[jur]

Skymont design definitely seems very modern and sophisticated, but the question is: "can it achieve same IPC, vector throughput, frequency... as P-CORE without having P-core size (and power)?". It seems that Intel's P-core dedicate a lot of die space to lower latency instructions and vector units. Also, large die size of P-core makes it less power dense, which enables it to achieve higher frequency and power.

 

[SiliconFly]

Skymont design definitely seems very modern and sophisticated, but the question is: "can it achieve same IPC, vector throughput, frequency... as P-CORE without having P-core size (and power)?". It seems that Intel's P-core dedicate a lot of die space to lower latency instructions and vector units. Also, large die size of P-core makes it less power dense, which enables it to achieve higher frequency and power.

Well, it's been almost 3 years since Golden Cove and the P core team haven't come out with anything worthwhile till date. At this point, Pat can't even blame his predecessors for this mess. This one definitely belongs to him.

 

[AMDK11]

In my humble opinion, LionCove is a new approach to core design and construction methodology.

Since the beginning, Intel has used a unified schedule for FPUs and ALUs with common execution ports, and now with LionCove, Intel is moving away from this pattern. Overall, LionCove provides plenty of resources and logic (Single wide 8 way decoder, including an 8x larger prediction block!) to achieve the highest possible single thread IPC.

Skymont's philosophy is still to take up as little space as possible and make many compromises to keep the logic less complex.

According to Intel, splitting the schedule separately into ALU and FPU + new P-Core design techniques and methods will allow for larger changes to be made in a shorter time. I wouldn't worry about the future of P-Core, but of course, as always, time will tell.

 

[TwistedAndy]

Skymont design definitely seems very modern and sophisticated, but the question is: "can it achieve same IPC, vector throughput, frequency... as P-CORE without having P-core size (and power)?". It seems that Intel's P-core dedicate a lot of die space to lower latency instructions and vector units. Also, large die size of P-core makes it less power dense, which enables it to achieve higher frequency and power.

Yes, it's pretty realistic to expect that Skymont will be close to Raptor Cove in terms of IPC. Technically, Skymont offers a higher throughput (8-wide) compared to Raptor Cove (6-wide), but it has simpler predictors, smaller buffers, etc.

As for Lion Cove, I think that the improvement of 15% in IPC is pretty realistic as well.

Intel will probably keep the HT for ARL-S and HX to show better numbers in Cinebench. It makes sense.

Also, I expect to see lower PL1 and PL2. 253W+ PL2 is too much.

 

[AMDK11]

LionCove at ArrowLake will be different. Intel itself admitted that logic was removed from the LunarLake variant to save space and power consumption. One such thing is that the L1 and L2 communication bandwidth in ArrowLake will be 128B instead of 64B like in LunarLake.
Intel says the 64B's L1 and L2 communication links provide a smaller footprint, which is key for LunarLake. But ArrowLake doesn't need such compromises, where performance comes first.

I suspect there may be more, such as ROBs or execution units. But this will only be known for sure during the ArrowLake presentation.

Certainly much faster communication between tiles, etc.

 

[SiliconFly]

LionCove at ArrowLake will be different. Intel itself admitted that logic was removed from the LunarLake variant to save space and power consumption. One such thing is that the L1 and L2 communication bandwidth in ArrowLake will be 128B instead of 64B like in LunarLake.

I suspect there may be more, such as ROBs or execution units. But this will only be known for sure during the ArrowLake presentation.

Certainly much faster communication between tiles, etc.

Sincerely hoping your prediction comes true! If ARL's LNC under performs compared to competition, it's gonna be a huge disappointment as the same LNC is being reused in next gen (2025Q3) desktop parts as well I think.

 

[ondma]

I have been speaking about your specific post here:

ARL is not 6 months after Zen 5. I am simply trying to correct that fallacy. If I misinterpreted or obfuscated "6 months earlier", please help correct me.

If you wish to have a discussion about performance, go ahead. Last we left it, I used your claim of Zen 5 having comparable performance better than ARL and went with that. Do you have anything concrete to add to your original performance estimate? We will find out in a few months which of those claims is true. But that has nothing to do with your timeline post that we have been discussing for days. A post which was based on ARL being 6 months after Zen 5.

If you want to discuss performance and do not have anything performance-wise to add (especially unlikely given that you have no idea what you are talking about with respect to hyperthreading), how about we go back to your first post on this. What happens to Intel if:
1) Arrow Lake is bad compared to Zen 5
2) Arrow Lake is comparable
3) Arrow Lake beats Zen 5

Well, again, you still refuse to recognize the intent of my post, even after I restated it. Lets quit arguing about how much later than Zen 5 ARL will be. I will grant you 3 months later, although I still think considering Intel's execution record that is an optimistic estimate for good availability. Still, I never said how late it will be is the only important factor. If it is later, but performance is a home run, that is acceptable. I will give you my take on the 3 scenarios you posted, and be done with this argument.

1) ARL worse than Zen 5: Intel is in deep trouble in the desktop. I dont see any way you can deny this. Intel will be permanently behind unless they hit a home run or AMD falters.

2) ARL is comparable to Zen 5: Intel will be chronically doomed to be equal at best, and behind with every new AMD release, unless Intel hits a home run at some point. And, no, I dont consider spamming improved E cores for MT benchmarks to be taking the lead in performance. ST and gaming still matter. On that front, BTW, Intel needs something to counter AMD's vCache.

3.) ARL beats Zen 5 (I would add by a significant margin like 10% or so) The best possible case for intel, although still not as desirable as if they came out at the same time or before Zen 5. They will lose a considerable amount of sales to Zen 5 before ARL comes out, but at least they could make up for some of them if ARL is clearly ahead in performance.

And BTW, I consider myself an Intel supporter, and have not bought AMD for the desktop since the Athlon 2600. I would love for Intel to give me a reason to stick with them for my next desktop upgrade. I thought ARL might be it, but, wow, unless every leak is wrong I dont see it.

 

[AMDK11]

Sincerely hoping your prediction comes true! If ARL's LNC under performs compared to competition, it's gonna be a huge disappointment as the same LNC is being reused in next gen (2025Q3) desktop parts as well I think.

LionCove has not said its last word yet. We will see how dividing schedules into separate ones for ALU and separate ones for FPU will work in many scenarios and loads. What will this look like in terms of adding a level of cache.

There is one more thing, namely in the LionCove presentation the predictor is as much up 8x larger, which may suggest that the variant for LunarLake is smaller than the one in ArrowLake.

I'm not saying it will be 100%, but it is possible that the IPC in ArrowLake will be higher than everyone expects.

 

[DavidC1]

(i.e, Skymonts in ARL are gonna be faster than the Skymonts in LNL).

The differences are 5% per clock due to fabric and rest in clocks.

I suspect there may be more, such as ROBs or execution units. But this will only be known for sure during the ArrowLake presentation.

Certainly much faster communication between tiles, etc.

There's no reason to expect Arrowlake's LNC to be noticeably faster. Arrowlake is also a delayed product and needs to overcome the deficiency that Meteorlake has too.

It'll be competitive merely due to the fact that the competition also only offers 16% gain, and the 5.7GHz clock is similar to the deleted peak clock for Arrowlake too.

What about next year? Do we really expect the competition to be Zen 6? If it's ARL-R vs Zen 5-R then it'll be again similar.

Zen 1: Mar 2, 2017
Zen 2: Nov 14, 2019(20 months)
Zen 3: Nov 17, 2020(12 months)
Zen 4: Sep 27, 2022(22 months)
Zen 5: July 31, 2024?(22 months)

That leaves Zen 6 to be Fall 2026, just in time for Pantherlake to counter it, is it not?

 

[DavidC1]

In my humble opinion, LionCove is a new approach to core design and construction methodology.

Since the beginning, Intel has used a unified schedule for FPUs and ALUs with common execution ports, and now with LionCove, Intel is moving away from this pattern. Overall, LionCove provides plenty of resources and logic (Single wide 8 way decoder, including an 8x larger prediction block!) to achieve the highest possible single thread IPC.

Lion Cove takes nearly 3x the size and uses 20-80% more power ISO-process.

If you take the square root law, 2x the size = 41%, 2.5x = 58%

The problem is the team and the ideology behind it. It's still all big, bigger, and biggest.

During their best days they were still quite slow and innovations few and far between. Meanwhile, the E core team brings new ideas every big updates.

Yes, it's pretty realistic to expect that Skymont will be close to Raptor Cove in terms of IPC. Technically, Skymont offers a higher throughput (8-wide) compared to Raptor Cove (6-wide), but it has simpler predictors, smaller buffers, etc.

Which is a much saner way of doing things. CPUs have been power limited for eons. Therefore, every addition has to be done with efficiency in mind. Skymont does, the P core team has been in expansion mode since post-Sandy Bridge. C&C notes this regarding decoders in their Skymont article about Apple chips.

They expanded retire to 16-wide because they could reduce expansions elsewhere and ends up saving overall.

 

[H433x0n]

Well, again, you still refuse to recognize the intent of my post, even after I restated it. Lets quit arguing about how much later than Zen 5 ARL will be. I will grant you 3 months later, although I still think considering Intel's execution record that is an optimistic estimate for good availability. Still, I never said how late it will be is the only important factor. If it is later, but performance is a home run, that is acceptable. I will give you my take on the 3 scenarios you posted, and be done with this argument.

1) ARL worse than Zen 5: Intel is in deep trouble in the desktop. I dont see any way you can deny this. Intel will be permanently behind unless they hit a home run or AMD falters.

2) ARL is comparable to Zen 5: Intel will be chronically doomed to be equal at best, and behind with every new AMD release, unless Intel hits a home run at some point. And, no, I dont consider spamming improved E cores for MT benchmarks to be taking the lead in performance. ST and gaming still matter. On that front, BTW, Intel needs something to counter AMD's vCache.

3.) ARL beats Zen 5 (I would add by a significant margin like 10% or so) The best possible case for intel, although still not as desirable as if they came out at the same time or before Zen 5. They will lose a considerable amount of sales to Zen 5 before ARL comes out, but at least they could make up for some of them if ARL is clearly ahead in performance.

And BTW, I consider myself an Intel supporter, and have not bought AMD for the desktop since the Athlon 2600. I would love for Intel to give me a reason to stick with them for my next desktop upgrade. I thought ARL might be it, but, wow, unless every leak is wrong I dont see it.

What does beating Zen 5 mean? Why is it incumbent on Intel whom already has majority market share to beat Zen 5 by a big margin and not vice versa? If it’s competitive on all metrics that’s arguably a loss for the company with less market share.

 

[AMDK11]

There's no reason to expect Arrowlake's LNC to be noticeably faster. Arrowlake is also a delayed product and needs to overcome the deficiency that Meteorlake has too.

Here's why. Until now, all processors, whether mobile, Core or Xeon, used exactly the same P-Core. Intel itself, together with LionCove, admitted that the design method allows for various design variants. For example, smaller FPU, no HT, which also suggests the use of a smaller ROB, UOP Cache, smaller execution units or a smaller cache (area) with 64B communication.

I don't know what the difference in average IPC will be, but I hope it will be more than just 2-3%.

Now LionCove doesn't just refer to a specific core. LionCove is now a general project that can create different core variants with optimization for different needs.

Just an example: For ArrowLake the FPU may have 3-4x ADD while the LunarLake variant has 2x ADD. Or ROB 630 instead of 576.

 

[DavidC1]

I don't know what the difference in average IPC will be, but I hope it will be more than just 2-3%.

Look at what was needed to get 14% in Lion Cove and 16% in Zen 5. Now tell me 2-3% is a small amount.

Also take into account many insider sources talk about serious issues in the P core team.

Gracemont put the E core team on the map. Skymont puts doubt about the big core team. Arctic Wolf is going to make it obsolete.

 

[AMDK11]

Look at what was needed to get 14% in Lion Cove and 16% in Zen 5. Now tell me 2-3% is a small amount.

Also take into account many insider sources talk about serious issues in the P core team.

Gracemont put the E core team on the map. Skymont puts doubt about the big core team. Arctic Wolf is going to make it obsolete.

Take into account that the Zen5 IPC curve includes AVX512 while the LionCove IPC curve does not.

IPC curve (average +19%) GoldenCove also includes AVX512 (located physically in the core) later disabled but not removed.

now remove AVX512 from the IPC curve in SunnyCove, GoldenCove, Zen4 and Zen5. The average will be even lower.

LionCove
Decode 6->8-Wide +33%
ALU 5->6 +20%
AGU 5->6 +20%

Yes, the P-Core team had problems. It was about outdated architectural design tools and methods. But I would be reserved about rumors claiming this is still the case. I believe that there is considerable progress and that a new strategy has been chosen.

Remember that the successor to LionCove is also in development because you say as if there would be no next generation P-Core.

 

[jur]

That leaves Zen 6 to be Fall 2026, just in time for Pantherlake to counter it, is it not?

Pantherlake is supposed to launch in 2025 on 18A...

 

[ondma]

What does beating Zen 5 mean? Why is it incumbent on Intel whom already has majority market share to beat Zen 5 by a big margin and not vice versa? If it’s competitive on all metrics that’s arguably a loss for the company with less market share.

Granted Intel will continue to sell products. However, just look where they are now from a market cap, profitability, and market share position vs 10 or 15 years ago when they had clear product leadership and were executing consistently in a timely manner. You can get by on your marketing and reputation for a period of time, but ultimately, having at best equal, and arguably inferior products will catch up to you. As for the rest of the question, I have explained this already in my posts with dullard. I am not going to get into it again.

 

[DavidC1]

Pantherlake is supposed to launch in 2025 on 18A...

Not desktop apparently. Either way they'll probably be close again.

MTL may sell. But it's still a mediocre product.

Well, I'm now thoroughly convinced anything with RWC is mediocre (including Granite Rapids).

LNC+SKT is their real future.

If the part about 8-wide decode in Granite Rapids version is true, then it won't be so bad.

Pat Gelsinger: "ten-plus percent changes in the core..."

 

[AMDK11]

Moreover, the IPC curve of SunnyCove, GoldenCove, Zen4 and Zen5 most likely also includes SMT/HT, which also affects the average IPC increase result. The LionCove variant in LunarLake physically lacks HT logic.

 

[H433x0n]

Granted Intel will continue to sell products. However, just look where they are now from a market cap, profitability, and market share position vs 10 or 15 years ago when they had clear product leadership and were executing consistently in a timely manner. You can get by on your marketing and reputation for a period of time, but ultimately, having at best equal, and arguably inferior products will catch up to you. As for the rest of the question, I have explained this already in my posts with dullard. I am not going to get into it again.

If you compare the quality of information that both companies provided at Computex it would suggest there isn’t much to worry about.

 

[gdansk]

If you compare the quality of information that both companies provided at Computex it would suggest there isn’t much to worry about.

If we go by quality of information provided Intel shouldn't worry about Apple at all. But Lunar Lake shows they do.