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)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	4P + 4E	4P + 8E
LLC	24 MB	24 MB ?	36 MB ?	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.)

 

[ErichZannMusician]

PL1+PL2= 15W

If that is the case, and if LNL preliminary estimated perf in CB 23 is +50% over MTL-U which scores ~6700 at the same power levels, then LNL is already ahead of Phoenix:



This is from Geekerwan.

Note that Intel's own nodes till Intel 7 don't scale well at low power, and LNL is on N3(B?) - which means that power scaling at the low end is going to be similar to competition.

So LNL should be doing 6700*1.5 ~= 10000 pts in CB R23 with MTP and PBP both capped at 15W (or 17W if you include the MoP).

 

[511]

If that is the case, and if LNL preliminary estimated perf in CB 23 is +50% over MTL-U which scores ~6700 at the same power levels, then LNL is already ahead of Phoenix:

View attachment 103379

This is from Geekerwan.

Note that Intel's own nodes till Intel 7 don't scale well at low power, and LNL is on N3(B?) - which means that power scaling at the low end is going to be similar to competition.

So LNL should be doing 6700*1.5 ~= 10000 pts in CB R23 with MTP and PBP both capped at 15W (or 17W if you include the MoP).

I have Ultra 9 185H i can try restricting it to 15W with 2P+8E+2LPE

 

[inf64]

I'm not sure if this was posted or how reliable it is (but it looks very much in line with the expectations):

https://twitter.com/x/status/1813228542315942088

 

[dullard]

I'm not sure if this was posted or how reliable it is (but it looks very much in line with the expectations):

https://twitter.com/x/status/1813228542315942088

If true, that would be at the top of the single core list, above the i-9 14900KS (143) and M3 Pro (142): https://www.cpu-monkey.com/en/cpu_benchmark-cinebench_2024_single_core

And basically in the middle of the i-9 14900K (2177) and i-9 14900KS (2252): https://www.cpu-monkey.com/en/cpu_benchmark-cinebench_2024_multi_core

According to this slightly different list, it would be the top of the single core and just below the i-9 14900K (2211): https://www.cgdirector.com/cinebench-2024-scores/

 

[inf64]

If true, that would be at the top of the single core list, above the i-9 14900KS (143) and M3 Pro (142): https://www.cpu-monkey.com/en/cpu_benchmark-cinebench_2024_single_core

And basically in the middle of the i-9 14900K (2177) and i-9 14900KS (2252): https://www.cpu-monkey.com/en/cpu_benchmark-cinebench_2024_multi_core

According to this slightly different list, it would be the top of the single core and just below the i-9 14900K (2211): https://www.cgdirector.com/cinebench-2024-scores/

We also have these two tweets:

One more model with 5.9Ghz ST boost

https://twitter.com/x/status/1814204150718050788

Zen 5 9950X ES scoring the same at ST at the same clock

https://twitter.com/x/status/1813595698459463687

 

[Geddagod]

If true, that would be at the top of the single core list, above the i-9 14900KS (143) and M3 Pro (142): https://www.cpu-monkey.com/en/cpu_benchmark-cinebench_2024_single_core

And basically in the middle of the i-9 14900K (2177) and i-9 14900KS (2252): https://www.cpu-monkey.com/en/cpu_benchmark-cinebench_2024_multi_core

According to this slightly different list, it would be the top of the single core and just below the i-9 14900K (2211): https://www.cgdirector.com/cinebench-2024-scores/

Would put it at an around ~16-17% IPC uplift compared to the 5.8GHz 13900k, not bad.

 

[Geddagod]

Surprised no one saw this tweet:

https://twitter.com/x/status/1814115377762496933

Or for those who dont wanna click on the link

 

[Philste]

Not bad, Intels ES usually never reach the clocks of final version, Intel has QS for that.

 

[Henry swagger]

Surprised no one saw this tweet:

https://twitter.com/x/status/1814115377762496933

Or for those who dont wanna click on the link
View attachment 103453

Must be testing a lunar lake laptop

 

[DavidC1]

The leaked ARL's MT performance is disappointing, as it should be higher than the ST gains. So still have to wait.

Surprised no one saw this tweet:

https://twitter.com/x/status/1814115377762496933

Or for those who dont wanna click on the link
View attachment 103453

It seems to be one rare generation where the BTB size didn't increase. That might explain less-than expected gains.

Skymont didn't get BTB increase either, but it benefits from the changes made to enhance the clustered decode scheme further.

 

[9949asd]

The leaked ARL's MT performance is disappointing, as it should be higher than the ST gains. So still have to wait.

It seems to be one rare generation where the BTB size didn't increase. That might explain less-than expected gains.

Skymont didn't get BTB increase either, but it benefits from the changes made to enhance the clustered decode scheme further.

last time i saw，the es frequency is 4.7g. We don’t know what is the frequency that r24 run at.

 

[SmokSmog]

I'm not sure if this was posted or how reliable it is (but it looks very much in line with the expectations):

https://twitter.com/x/status/1813228542315942088

Bunch of malarkey

Same as this single core on Zen5.

 

[Josh128]

Agree, this result is a fake. We have an actual 9950X ES @5.65GHz scoring 143 in the Zen 5 thread. The same Haytam guy posted like 153 for 9950X ES a few days before that. Scores seem made up-- I dont believe Haytam is a reliable source. Plus, the guys handle has the word "meld" in it. Instant skepticism just from that.

 

[CouncilorIrissa]

Bunch of malarkey

Same as this single core on Zen5.

View attachment 103459

it's over intelbros...

Tech-Meld (Tech Meld)

I love ML, i have been learning it since summer of 2023, a lot of things have happened since, from basic knowledge to creation of self-made language models and diffusion models, everything is still...

huggingface.co

 

[DavidC1]

last time i saw，the es frequency is 4.7g. We don’t know what is the frequency that r24 run at.

The 4.7GHz ES you are talking about is DIFFERENT from this ES, as the ST performance is up to par.

But people are calling him out as fake, so there it is.

 

[9949asd]

The 4.7GHz ES you are talking about is DIFFERENT from this ES, as the ST performance is up to par.

But people are calling him out as fake, so there it is.

I think that guy was just make up the number, there are no such thing of 295k 275k 255k.

 

[Det0x]

last time i saw，the es frequency is 4.7g. We don’t know what is the frequency that r24 run at.

Did you miss this tweet ?

The performance numbers shared @ "4.7ghz" were never 4.7ghz in the first place

 

[mikk]

We also have these two tweets:

One more model with 5.9Ghz ST boost

https://twitter.com/x/status/1814204150718050788

Two Ultra 9-k doesn't make sense, most likely a fake account.

 

[poke01]

Look, no x86 CPU(Zen5 or ARL) is going above 150 ST (Cinebench 2024) at stock. Only one non-x86 CPU will, that would be Apple's M4 Max(4.5GHz) now that its got a ~17% ST boost over M3 Max.

If AMD or Intel boosted above 5.7GHz this gen, then yes 150 would have been possible.

 

[jdubs03]

Look, no x86 CPU(Zen5 or ARL) is going above 150 ST (Cinebench 2024) at stock. Only one non-x86 CPU will, that would be Apple's M4 Max(4.5GHz) now that its got a ~17% ST boost over M3 Max.

If AMD or Intel boosted above 5.7GHz this gen, then yes 150 would have been possible.

Just to follow up, it’s the M4 series in general that will be over 150 points. Based on geekbench (I know it’s not ideal), the M4 gets 19% better single-core scores compared to the M3 (it is cross OS, but the M2 differences aren’t much when using each OS). If that improvement remains the same, then the M4 should score around 170 points in Cinebench 2024.

 

[511]

Just to follow up, it’s the M4 series in general that will be over 150 points. Based on geekbench (I know it’s not ideal), the M4 gets 19% better single-core scores compared to the M3 (it is cross OS, but the M2 differences aren’t much when using each OS). If that improvement remains the same, then the M4 should score around 170 points in Cinebench 2024.

It will not cross 170 cause the geekbench inflation is due to the Matrix extension it will be around 165

 

[511]

Also i winded what happens if Cinebench ran on Linux than we would have both the macos and Linux with posix thread model minimising the difference

 

[jpiniero]

Two Ultra 9-k doesn't make sense, most likely a fake account.

Normally I would agree but the higher i9 could easily be an unobtanium model aimed at winning benchmarks and not a product intended to be sold.

 

[Hail The Brain Slug]

Normally I would agree but the higher i9 could easily be an unobtanium model aimed at winning benchmarks and not a product intended to be sold.

Seems analogous to the KS model.

 

[SiliconFly]

Normally I would agree but the higher i9 could easily be an unobtanium model aimed at winning benchmarks and not a product intended to be sold.

Considering the amount of meltdowns RPL and RPL-R K series is having, it is better that the higher i9 remains an unobtanium model.