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

Page 244 - 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
651
536
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: 23,959
  • LNL.png
    LNL.png
    881.8 KB · Views: 25,428
Last edited:

DavidC1

Member
Dec 29, 2023
170
233
76
It don't matter whether Arrowlake has LPG or LPG+. The differences are negligible.

The bigger thing is only having 4 Xe cores, but it'll still be more than twice as fast as the 32 EUs Iris Xe in the Raptorlake generation.
 

Oneto

Member
Apr 2, 2017
65
72
91
It don't matter whether Arrowlake has LPG or LPG+. The differences are negligible.

The bigger thing is only having 4 Xe cores, but it'll still be more than twice as fast as the 32 EUs Iris Xe in the Raptorlake generation.
As someone who uses the iGPU when I'm between GPUs that is good news. Could be better, sure, but I don't use iGPU much anyway.
 
  • Like
Reactions: DAPUNISHER

DavidC1

Member
Dec 29, 2023
170
233
76
Please stop comparing apples to oranges.

MTL IGP with only 8 XEs is fed by 128-bit 5600-7500mbps (LP)DDR5(x). 20 XEs will need a lot more, unless clockspeed will take a drastic hit.
Not in 2027, which is 3 years away, which is enough time to bring huge changes in the computing space.

Nevermind the regular advances in DDR, the architectures will change to address lack of bandwidth such as Infinity Cache in RDNA.

This also assumes that Meteorlake's 8 Xe's are perfectly matched to bandwidth and any increase in compute results in imbalance. For all we know it might take 12 Xe to reach that limit.
If Intel wants to beat Apple's power efficiency, they need to let go of their addiction to HP node libraries.
That's just one aspect. They need to adopt a new mindset because comparing similar libraries(HP to HP, HD to HD), the Intel nodes are higher performance but less dense. Fundamentally it's been about manufacturing their CPUs so the nodes have been built from the ground up to be that way.

If/when Intel foundry gets big enough then changes will happen. Just like they did not focus on drivers until dGPU(unlike claims otherwise).
As someone who uses the iGPU when I'm between GPUs that is good news. Could be better, sure, but I don't use iGPU much anyway.
I think Arrowlake LPG+ changes are supposed to be 20% increase in performance brought on by the small architecture changes with larger L2 caches, and addition of XMX while still based on the Alchemist architecture.

So, it's not a tiny change, but nothing that'll change the big picture since the Xe core counts are staying the same, and at 4 Xe cores, it's still anemic.

It is in Lunarlake where the real good stuff happens with Battlemage-based Xe2 iGPU, and Celestial-based Xe3 in 2026 with Pantherlake.
 
Last edited:
  • Like
Reactions: Tlh97

poke01

Senior member
Mar 8, 2022
730
703
106

mikk

Diamond Member
May 15, 2012
4,140
2,148
136
Not in 2027, which is 3 years away, which is enough time to bring huge changes in the computing space.

Nevermind the regular advances in DDR, the architectures will change to address lack of bandwidth such as Infinity Cache in RDNA.

This also assumes that Meteorlake's 8 Xe's are perfectly matched to bandwidth and any increase in compute results in imbalance. For all we know it might take 12 Xe to reach that limit.


If we ignore other improvements like clock speed and uarch improvements getting 2.5x more units from 8/1024 up in 3 years would have been a big improvement. AMD used to have 8-11 CUs for many years. Then from 8CU to 12CU in 2 years and 12CU to 16CU in another 2 years, if this continues we get 20-24CUs in 2027 from their meaningful mobile lineup. Intel went from 512 to 1024 shader in 3 years.

Because Intel can gain more from clock speed, uarch and driver improvements in the next generations (coming from a lower base) they may not even need 20 Xe in Nova Lake. They already have 1024 units in Meteor Lake which is what AMD brings in Strix Point. With this in mind I doubt it's more than 16 Xe cores unless they come up with a special series like Strix Halo but how relevant can it be.

Raven Ridge 11CU/704 2017
Picasso 11CU/704 2019
Renoir 8CU/512 2020
Cezanne 8CU/512 2021
Rembrandt 8CU/512 2022
Phoenix 12CU/768 2023
Strix Point 16CU/1024 2025

Icelake 512 2019
Tigerlake 768 2020
Meteor Lake 1024 2024
 
  • Like
Reactions: igor_kavinski

Hulk

Diamond Member
Oct 9, 1999
4,223
2,012
136
If/when Intel foundry gets big enough then changes will happen. Just like they did not focus on drivers until dGPU(unlike claims otherwise).
I don't know how this factors into this discussion but I remember watching an interview with Raja where he said the issue with the drivers came because of the additional compute offered by the discrete hardware. If I remember correctly the point was that the iGPU was so weak from a hardware point of view that is would be "saturated" computationally with very little driver optimization whereas the discrete hardware needed a lot of software optimization for it to be fully utilized.

I remember this because I thought it was interesting and kind of the opposite of what I would have thought. I'm thinking that bottlenecks in the drivers were not "hit" with the iGPU but became an issue with the discrete hardware.
 

Doug S

Platinum Member
Feb 8, 2020
2,258
3,504
136
That's just one aspect. They need to adopt a new mindset because comparing similar libraries(HP to HP, HD to HD), the Intel nodes are higher performance but less dense. Fundamentally it's been about manufacturing their CPUs so the nodes have been built from the ground up to be that way.

If/when Intel foundry gets big enough then changes will happen. Just like they did not focus on drivers until dGPU(unlike claims otherwise).

I agree. I think one of the reasons Intel couldn't make it in mobile (the biggest reason being their idiotic insistence on x86 when that's not what the customers wanted) was because their processes were built from the ground up for high clock PC CPUs, and while they could tweak things for lower power enough to make laptop chips going down another big step in power to where smartphone SoCs live was another matter entirely.
 
  • Like
Reactions: lightmanek

coercitiv

Diamond Member
Jan 24, 2014
6,195
11,883
136
@Hulk Regarding your other post about Arrow Lake, AFAIK the config launching "soon" is 8+16, with 8+32 as a possible sequel.

MT throughput over RPL+ shoukd be offered by stronger cores in combination with higher clocks under full load (see MTL mobile gains over RPL mobile)

The other thread was closed, so we'll have to make it work here.
 

mikk

Diamond Member
May 15, 2012
4,140
2,148
136
Also this:
But how do they make up the MT performance? I think they replace the 8 logical P cores + 16 Gracemont E cores (24 total threads) with 24 Crestmont E cores.

ARL is getting Skymont E cores which makes a big difference. Crestmont on ARL would be lost without HT.
 

SiliconFly

Golden Member
Mar 10, 2023
1,043
536
96
I think Arrowlake LPG+ changes are supposed to be 20% increase in performance brought on by the small architecture changes with larger L2 caches, and addition of XMX while still based on the Alchemist architecture.

So, it's not a tiny change, but nothing that'll change the big picture since the Xe core counts are staying the same, and at 4 Xe cores, it's still anemic.
I seriously doubt LPG+ will bring in 20% performance uplift over LPG. Maybe single digit at best (definitely not more). Alchemist is a lost cause. They should bury it asap.

It is in Lunarlake where the real good stuff happens with Battlemage-based Xe2 iGPU, and Celestial-based Xe3 in 2026 with Pantherlake.
Lunar Lake & Panther Lake are exciting products I'm sure. But they're niche low-volume products that cater to a very small market. Main stream is always Arrow Lake (and sadly ARL refresh next year).
 

DavidC1

Member
Dec 29, 2023
170
233
76
The term fanboy or any similar insult is not permitted in the tech forums.
I don't know how this factors into this discussion but I remember watching an interview with Raja where he said the issue with the drivers came because of the additional compute offered by the discrete hardware. If I remember correctly the point was that the iGPU was so weak from a hardware point of view that is would be "saturated" computationally with very little driver optimization whereas the discrete hardware needed a lot of software optimization for it to be fully utilized.

I remember this because I thought it was interesting and kind of the opposite of what I would have thought. I'm thinking that bottlenecks in the drivers were not "hit" with the iGPU but became an issue with the discrete hardware.
If you look at just dGPUs, it's the mindset that determines everything. Rebar having big performance impact with their dGPU is another evidence of their iGPU mindset. They didn't need to worry because they didn't have dGPUs. Chipsandcheese micro-tests show that Alchemist has many imbalances which they point out partly are due to their inexperience in dGPU and partly due to their iGPU focus.

Scaling up anything isn't just "copy and paste" because unexpected problems occur. If it was that easy why wasn't A770 competitive at perf/watt and perf/mm2?

Same with Intel Foundry. If/when Intel gets broad customer base, then they'll focus on aspects other than what they were good at. Until then you can bet on them being behind. How can you know something you haven't done before?
Icelake 512 2019
Tigerlake 768 2020
Meteor Lake 1024 2024
If Intel gets to 20 Xe, then AMD will get to 26 CUs or similar. Competition sometimes seem like a waste because multiple companies sell products that end up being on the ballpark(or even similar) with each other with enough minor differences to keep redacted all hot and heavy.

Also, the one claiming 20 Xe is @Tigerick. I suggest you ask him.
 
Last edited by a moderator:
  • Like
Reactions: coercitiv

Hulk

Diamond Member
Oct 9, 1999
4,223
2,012
136
I tried to make the following post in a thread titled "Arrow Lake" because as we get close to a processor release we usually start a thread specifically for discussing the impact of that particular generation of CPU, rather than continuing with a more generalized discussion such as this one, which tends to spin off into different direction. The mods thought differently and locked that thread so I'm reposting here.



Since Arrow Lake is hopefully months away I think it deserves it's own thread. Especially now that Meteor Lake is available and we have some idea of how Intel 4 is performing.

To get this going I will submit my prediction for Arrow Lake.
I am thinking Arrow Lake will be a 8+24 part with max clock for the P's to be about 5.5GHz and about 4.3GHz for the E's.

Here's my reasoning.

Intel has painted themselves into the same corner they did with Haswell/Devil's Canyon when moving to Broadwell. We have discussed this. Devil's Canyon was on a very mature 22nm process while Broadwell was on a brand new 14nm process. Broadwell didn't have enough frequency and IPC to match much less beat Devil's Canyon so it was in reality a mobile only part. This is much like the situation with Raptor Lake and Meteor Lake today. Except this time it will be even harder for Intel to beat Raptor Cove because Raptor is monolithic vs tiles ARL and Raptor is currently reaching insane frequencies albeit at the expense of insane power.

What is Intel to do?

Based on current Meteor mobile frequency I think 5.5GHz for desktop is realistic for the 2nd go at Intel 4. I am also going to go out on a limb and predict Lion Cove will show about 15% better throughput (IPC) than Raptor Cove. But, this will be offset by the clock speed disadvantage and latency penalty of Arrow Lake. I will estimate this penalty to be 10%. About 6 or 7% do to clock speed reduction and the rest due to the disaggregated architecture.

This means that Arrow Lake would have a 5% performance improvement over Raptor Lake in ST. Many people will complain like mad about this not realize that Intel had to overcome a large clock deficit and tile latency to achieve it. Both the people who complain about the relative lack of ST performance increase and those who claim Intel did well for moving to a new node and tiled architecture will be correct in their appraisal;)

Okay so they manage to increase ST performance by removing HT structures from Lion Cove, freeing up area for the core to be wider, additional OoO logic, etc... They also offset the increased thermal density in moving to Intel 4 by removing HT from the big cores.

But how do they make up the MT performance? I think they replace the 8 logical P cores + 16 Gracemont E cores (24 total threads) with 24 Crestmont E cores. The E core clusters are quite area efficient and there should be room for them with the move to Intel 4. This not only improves on MT performance because E's are more performant than HT logical cores, but it also makes "life easier" for the Thread Director since now it only has to schedule P's and E's, no more logical threads in the mix. Also the penalty for putting a thread on a logical core when it should be on a P physical core is somewhat mitigated as that mistake would be now placing the thread on what is at least a physical E core. But I have not real evidence for this, it just seems reasonable. If Crestmont has better IPC than Gracemont then ARL will do even better in MT.

Finally ARL should have an efficiency advantage over Raptor Lake. Not only in terms of power but in terms of the very real problem that you can't really extract full performance from Raptor without a custom loop and/or delicate underclocking. If my hypothetical 8+24 5.5/4.3 ARL can run those frequencies on air or perhaps a 240 AIO then that is better than Raptor performance that can be attained by mere mortals without custom loops and God-like knowledge of undervolting and/or winning the silicon lottery and getting a highly binned part.

So in conclusion I see the 15900K being 8+24, no HT, running 5.5/4.3 with 5% better ST performance and 10% better MT performance, but this performance is actually attainable and won't require 300+ watts and commensurate cooling solutions.

Okay, have at me!
 

Hulk

Diamond Member
Oct 9, 1999
4,223
2,012
136
@Markfw,

I think ARL and Zen 5 are going to be very competitive. If I had to make a guess right now based on Meteor Lake performance and AMD's recent history of hitting Zen releases out of the ballpark I would bet that Zen 5 will be more performant and more efficient than Arrow Lake, but as always there will be arguments for and against both.

That's pretty much how I feel about Raptor Lake and Zen 4. They are very close in performance, but when you take power and heat into consideration it's hard to deny that Zen 4 is most likely the better choice for most people that just need to get work done. So why do I have a 14900K? I'll admit that I've always had Intel chips, I know them, know how to build systems, and I generally have fun tweaking the nuclear reactor that Raptor parts can be. But there are many times where I sit back in my chair and think, "Darn! Should I just get a 7950 and not worry about heat, and voltages, LLC, etc.. and just get my work done?"

Intel has a lot to overcome to improve on Raptor performance... moving to tiles and dealing with 6GHz clocks of Intel 7. AMD need only focus on improving a proven design so I think their chances of beating their previous generation are better than Intel's. The one area Intel could give them trouble is MT performance if they have the die budget to go +24 on the E's.

If the P's and E's show 5% IPC improvement, a hypothetical 8+24 ARL part, with no HT, running 5.5/4.3, would score over 46,000 in CB R23 MT with nearly 29000 of those points coming from the E's. Those pesky "Cinebench Accelerators" could be pesky for AMD. But then again Intel still has the issue with ST performance and would need 5.7GHz in with Lion Cove to match 6GHz Raptor Cove assuming 5% IPC increase for Lion Cove. Things are even more problematic for Intel if they release a 14900KS and its 6.2GHz CB ST scores. But they can conveniently "forget" about those scores come ARL launch time.

It just seems to me that based on lackluster MTL performance Intel has a big hill to climb with ARL while AMD can release Zen 5 with a 10% IPC bump and at least in my evaluation it would be a success. Plus AMD might have more headroom since Zen 4 clocks aren't in the insane region.

I'm excited to see where this goes over the next 8 or so months. I see three outcomes with #1 being the most likely.

1. Zen 5 and ARL are close in performance much like Zen 4 and RPL. The Earth rotates as usual.

2. ARL comes out big and is a solid win over Zen 5. I think this is a longshot.

3. ARL has big problems, low clocks and/or low/stagnant IPC and AMD pulls decisively into the lead. More likely to than #2 in my opinion. I'm not sure what would be worse for Intel, underperforming ARL or ARL delayed. Either could put Intel in the passenger seat for a long time. A delay would scare the you know what out of the board and investors... here we go 10nm fiasco again. The fact that Intel has just released the same top of the stack part (13900K, 13900KS, and 14900K) 3 times, and if rumors are correct, are about to for the 4th time (14900KS) doesn't make me swoon with confidence over their ability bring a plan to successful fruition.
 

SiliconFly

Golden Member
Mar 10, 2023
1,043
536
96
...and predict Lion Cove will show about 15% better throughput (IPC) than Raptor Cove. ...a 5% performance improvement over Raptor Lake in ST...
ST prediction is a bit narrow I think. Final ST may be just similar to RPC (with a touch of regression). Or it may even be up to 25% to 30% higher than RPC considering LNC is a grounds up new design and may not require higher clocks. Too soon to predict.

Finally ARL should have an efficiency advantage over Raptor Lake.

...I would bet that Zen 5 will be more performant and more efficient than Arrow Lake...

...but when you take power and heat into consideration it's hard to deny that Zen 4 is most likely the better choice... ...fun tweaking the nuclear reactor that Raptor parts...
People tend to forget what matters the most. Right from the moment Zen 1 was released, to up until now, Intel processors have always been on grossly inferior nodes compared to their Zen counterparts. And for the first time in many many years, Intel's desktop parts are actually being built on a better node than the competition. ARL desktop is getting around 2 to 2.5X node jump. This translates to massive efficiency gains assuming LNC is a half decent design.

There's a very very high chance ARL might beat Zen5 comfortably in efficiency. As for performance, thats still a question.
 
Last edited:

Hitman928

Diamond Member
Apr 15, 2012
5,262
7,888
136
ST prediction is a bit narrow I think. Final ST may be just similar to RPC (with a touch of regression). Or it may even be up to 25% to 30% higher than RPC considering LNC is a grounds up new design and may not require higher clocks. Too soon to predict.


People tend to forget what matters the most. Right from the moment Zen 1 was released, to up until now, Intel processors have always been on grossly inferior nodes compared to their Zen counterparts. And for the first time in many many years, Intel's desktop parts are actually being built on a better node than the competition. ARL desktop is getting around 2 to 2.5X node jump. This translates to massive efficiency gains assuming LNC is a half decent design.

There's a very very high chance ARL might beat Zen5 comfortably in efficiency. As for performance, thats still a question.

Zen 1 was on an inferior node to what Intel had at the time. Zen 4 and MTL are on roughly equivalent nodes. Arguably Zen 3 and Tiger Lake were on roughly equivalent nodes as well. I also don't expect there to be a large process difference between Zen 5 and ARL (density should be a large advantage but Intel typically sacrifices a lot of potential density gains for high frequency operation).
 
Last edited:

coercitiv

Diamond Member
Jan 24, 2014
6,195
11,883
136
The one area Intel could give them trouble is MT performance if they have the die budget to go +24 on the E's.
The leak we had from Intel's internal estimates (via Igor's Lab) was for a 8+16 die, I don't expect any 8+24 or 8+32 config to come this year.
 

SiliconFly

Golden Member
Mar 10, 2023
1,043
536
96
Zen 1 was on an inferior node to what Intel had at the time. Zen 4 and MTL are on roughly equivalent nodes. Arguably Zen 3 and Tiger Lake were on roughly equivalent nodes as well. I also don't expect there to be a large process difference between Zen 5 and ARL (density should be a large advantage but Intel typically sacrifices a lot of potential density gains for high frequency operation).
Desktop comparison:

2019: Zen2 N7 - Comet Lake 14nm
2020: Zen3 N7P - Rocket Lake 14nm
2022: Zen4 N4 - Raptor Lake 7nm

Note:
Zen1 was never really competition. Only the later parts.
In desktop, Intel was stuck in 14 for many years (during Zen2 & Zen3).
And in mobile, Tiger Lake's 10SF was no comparison to more advanced N7P.
Ignoring MTL as it appears to be a dud (as of now).
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,551
14,505
136
Desktop comparison:

2019: Zen2 N7 - Comet Lake 14nm
2020: Zen3 N7P - Rocket Lake 14nm
2022: Zen4 N4 - Raptor Lake 7nm

Note:
Zen1 was never really competition. Only the later parts.
In desktop, Intel was stuck in 14 for many years (during Zen2 & Zen3).
And in mobile, Tiger Lake's 10SF was no comparison to more advanced N7P.
Ignoring MTL as it appears to be a dud (as of now).
Zen 5 expectations are high, but I am cautious until it arrives. ARL to me is a real unknown, as not too many credible leaks. I will just wait until they both arrive and are benchmarked before I even begin to take a stand with either one.

But a nice surprise from Intel would be welcome.

Edit: And I agree the MTL appears to be a dud right now.
 

Hitman928

Diamond Member
Apr 15, 2012
5,262
7,888
136
Desktop comparison:

2019: Zen2 N7 - Comet Lake 14nm
2020: Zen3 N7P - Rocket Lake 14nm
2022: Zen4 N4 - Raptor Lake 7nm

Note:
Zen1 was never really competition. Only the later parts.
In desktop, Intel was stuck in 14 for many years (during Zen2 & Zen3).
And in mobile, Tiger Lake's 10SF was no comparison to more advanced N7P.
Ignoring MTL as it appears to be a dud (as of now).

So From Zen1 on, Intel always had a "grossly inferior node" as long as we ignore 2 of the 4 generations, including Zen 1? On desktop, obviously the situation was worse, but that's not the only market Intel was lagging in and is actually the least important one (outside of embedded). On the mobile side, Intel had the better or roughly equivalent process for at least 3 of the 4 generations, if not 4 of the 4 (though released later than the AMD counterparts). Only Zen 2 on mobile you could argue had a very clear process lead throughout its entire lifetime. Tiger Lake's 10SF was much, much closer to Zen 2's N7 process than anyone else's 14nm/12nm process. It was more similar to Zen+'s process on GF 12 nm compared to Intel's 14 nm but in reverse where Zen 2's process was the better one. ADL's Intel 7 versus Zen 3's TSMC 7(P) was again, fairly close. I would definitely give the nod to TSMC's process (especially in mobile) but the difference wasn't near as much as the difference in the final products.

Desktop Zen 4 is also on N5, not N4. Zen 5 desktop will be on N4 (probably N4P). Intel will actually have the better process on desktop with ARL for the first time in a while, but I don't think it will help them as much as you think. The process difference won't be that large and I expect the Zen 5 architecture to just be better, particularly within sane power limits. Could be wrong, we'll see, not too much longer to find out.
 
Last edited:

dullard

Elite Member
May 21, 2001
25,059
3,410
126
Nova Lake (2026 H2) maybe on 14A and/or 18A and/or 18A-P and/or TSMC N2. Lots of possibilities. Too soon to tell.
Too soon to tell of course. But now we can even speculate about 10A in 2028. https://www.tomshardware.com/pc-com...-for-fully-ai-automated-factories-with-cobots

The chart there is very informative: https://cdn.mos.cms.futurecdn.net/iUo6natDCxZNAprYxFxg5M-1200-80.jpg
Intel 4/3 won't really reach high production levels until 2026 while Intel 20A/18A will be at higher levels of production than Intel 4/3 from here on out. That shows a lot of the reasoning behind Intel's large array of products in the next 2 years, each one a niche product, while Raptor Lake keeps getting produced again and again.

Also it shows that even Intel is lumping them together as half nodes (Intel 4/3, Intel 20A/18A).