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.)

 

[DAPUNISHER]

I know the rules changes take some getting used to. Reminder that vendor threads are not versus anymore. It has cut down on derailments and flame wars massively. You can make unfavorable compare and contrasts in the vendor threads it favors. But no longer in the one it does not. It invariably ends up being viewed as flame bait.

Mod DAPUNISER

 

[Abwx]

Entry with Highest ST score for Zen 5 Strix =

View attachment 103179

Entry with Highest ST score for LNL ES:

View attachment 103180

Do the math.

Perf/clock is within margin of error even on an ES sample in which the LNC core-only power is consuming perhaps 0.5x-0.6x power of a single Zen 5 core in Strix running ST Geekbench 6.

That s just the 8533 RAM that boost the score by a few % since GB is sensitive to RAM speed.

For the rest dunno what is the source of your power numbers, what is sure is that LNL use 30W to score 10200 pts in CB R23 while a 8945HS does 10600 pts at 20W, as for Strix there s no explicit and verified numbers but it should be logically better than Hawk Point.

All in all without SMT LNL has to give up some efficency in MT because at a given frequency throughput can be increased by 25% with power increasing by the same amount, while increasing frequency by 25% would require 70% more power.

 

[ErichZannMusician]

There have been no verifiable leaks of LNL CB23 MT scores.

Besides, that workload is irrelevant in a notebook form factor.

Memory speed difference might affect scores by 2-3 percent at most.

 

[Abwx]

There have been no verifiable leaks of LNL CB23 MT scores.

There was a leak published at 15 and 30W for a few tests.

Besides, that workload is irrelevant in a notebook form factor.

Memory speed difference might affect scores by 2-3 percent at most.

If memory affect the score by a few % that s a lot at this level, i talked of a 7% difference.

As for the load it s not relevant for real usage but it is when it comes to estimate the efficency.

 

[H433x0n]

5.7Ghz x 1.15x (IPC) / 6.2Ghz ~= 1.057 or roughly 6% higher ST performance.

This is probably the most accurate estimate in this thread.

There have been no verifiable leaks of LNL CB23 MT scores.

Besides, that workload is irrelevant in a notebook form factor.

Memory speed difference might affect scores by 2-3 percent at most.

There has but it’s irrelevant. You’re not buying a LNL laptop to render images on CPU.

15W = 8182
27W = 10212 (it’s way outside of its efficiency range at this point)

 

[dullard]

There have been no verifiable leaks of LNL CB23 MT scores.

Besides, that workload is irrelevant in a notebook form factor.

There has but it’s irrelevant. You’re not buying a LNL laptop to render images on CPU.

I fully agree with both of you that Cinebench (or similar workflows) on a thin and light laptop is not a very irrelevant benchmark. But, apparently that is all some people need to see to declare Lunar Lake a bad chip:

Page 378 - Discussion - Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads

Page 378 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[SiliconFly]

5.7Ghz x 1.15x (IPC) / 6.2Ghz ~= 1.057 or roughly 6% higher ST performance.

This is probably the most accurate estimate in this thread.

Oh pls. Comparing with KS? Lets try the more widespread K. And only 15%?

5.7Ghz x 1.20x (IPC) / 6.0Ghz ~= 1.14 or roughly 14% higher ST performance.

 

[Det0x]

Oh pls. Comparing with KS? Lets try the more widespread K.

What CPU (SKU) do you think Arrow Lake will be compared against in benchmarks ?
Or let me ask it in a other way, what is current "mainsteam" chart-topping cpu from Intel ?
*edit*
I checked around and it seems like most review sites dont actually normally list the 14900KS, so your right.
Just one example from one of the biggest sites:

CPU Benchmarks and Hierarchy 2024: CPU Rankings

We've run thousands of CPU benchmarks on all new and older Intel and AMD CPUs and ranked them.

www.tomshardware.com

Arrow lake should indeed be compared aginst the 14900K if/since the KS is not included i guess

 

[H433x0n]

Oh pls. Comparing with KS? Lets try the more widespread K. And only 15%?

5.7Ghz x 1.20x (IPC) / 6.0Ghz ~= 1.14 or roughly 14% higher ST performance.

It'll probably mostly be compared with the 14900K, which changes it to ~9% higher 1T performance.

 

[SiliconFly]

It'll probably mostly be compared with the 14900K, which changes it to ~9% higher 1T performance.

He's used 15% IPC gains, whereas I'm expecting a little more than that. 19% is what seems to be a very popular sweet spot considering LNL 14% + ~5% additional for ARL-S due to better caches & memory.

5.7Ghz x 1.19x (IPC) / 6.0Ghz ~= 1.13 or roughly 13% higher ST performance. This is pretty much inline with the lower double digit performance increase that many leakers were claiming initially.

LNL is better at none, Zen 5 is 7% faster per clock in Geekbench and 5% faster per clock as well in Cinebench...

Don't worry. ARL-S is expected to closely match or slightly surpass Zen5 in ST. But it will definitely mop the floor with Zen5 in MT.

 

[Josh128]

He's used 15% IPC gains, whereas I'm expecting a little more than that. 19% is what seems to be a very popular sweet spot considering LNL 14% + ~5% additional for ARL-S due to better caches & memory.

5.7Ghz x 1.19x (IPC) / 6.0Ghz ~= 1.13 or roughly 13% higher ST performance. This is pretty much inline with the lower double digit performance increase that many leakers were claiming initially.


Don't worry. ARL-S is expected to closely match or slightly surpass Zen5 in ST. But it will definitely mop the floor with Zen5 in MT.

I fully expect the 3nm ARL-S 24 core to win the Cinebench MT wars vs the 4nm Zen 5 16 core. If it doesnt, it will be considered a huge faceplant by Intel. I'd say ~46K-48K R23 MT at stock whereas it appears Zen 5 gets ~42K stock and 44K-45K with PBO/CS tuning.

Even last years leaked internal Intel slide for ARL-S points to +15% MT, so applying that to 13900K's 40.5K R23 MT @ 253W yields ~46.5K or more.

I dont expect more than +15% ST IPC though. I think LNL needs all the IPC it can get to keep clocks down at these low power levels, thats why Lion Cove was designed they way it was, and increases to cache arent going to change it much. The core logic design of Lion Cove is not going to change for ARL vs LNL, same goes for Skymont.

 

[H433x0n]

He's used 15% IPC gains, whereas I'm expecting a little more than that. 19% is what seems to be a very popular sweet spot considering LNL 14% + ~5% additional for ARL-S due to better caches & memory.

Raptor Cove increased L2$ from 1.25mb to 2.0mb and that only brought gains of +2% int and +3% fp over Golden Cove. I wouldn’t expect anymore than that for the desktop variant of Lion Cove.

An IPC increase of 15% over RPL is actually on the high end.

 

[DavidC1]

Oh pls. Comparing with KS? Lets try the more widespread K. And only 15%?

5.7Ghz x 1.20x (IPC) / 6.0Ghz ~= 1.14 or roughly 14% higher ST performance.

Increasing L2 from 2.5MB to 3.0MB will in average not even increase performance by 1%, we'd be lucky to get 0.5%.

Mobile is slower than desktop but since it'll be compared to Raptorlake-S anyway, we'll see pretty much what Intel said, or 14-15% per clock.

@inf64 estimates are the most correct because of that.

Also, 5.7GHz/4.6GHz is EXACTLY what the deleted leak said many pages back.

 

[Henry swagger]

Geekbench 6 average ST score for 13/14900K = 3100 at 5.7 GHz

Lunar Lake ES Geekbench 6 ST score = 2700 at 4.9 GHz

Geekbench 6 ST PPC uplift on Lion Cove vs Raptor Cove based on ES silicon = 1% or within margin of error.

Add 200 pts ST to the ES score for the final product = 2900 ST, and recalculate PPC uplift = 9% higher PPC for Lion Cove in Lunar Lake.

Normalize clock speeds, then at 5.7 GHz, assuming linear sclaing, Lion Cove in ARL-S would be 16 - 25% faster PPC than Raptor Cove in 13900K.

16% - being conservative, if there is little change between ES and final silicon.
25% - being cautiously optimistic.

Incidentally, this seems to be right in line with what MLID speculated.

The naysayers are in for a surprise.

Leaks will shut them up

 

[DavidC1]

Incidentally, this seems to be right in line with what MLID speculated.

The naysayers are in for a surprise.

MLID also claimed up to 34% gain for Lion Cove and that his "sources" thought Skymont is a "solid gain" over Crestmont".

Now that Intel revealed a whole bunch of information at Computex, his holy-and-can't-go-wrong sources revised it down to 15-25%, quite a big difference.

Also, I suggest you use the English language properly. Naysayer would be someone that says Intel can't do anything right. In this case you are using the word to say Intel, the manufacturer of the said CPU is wrong and it must be higher which is funny since most of the time people say the opposite and say the actual numbers MUST be lower than what Intel says.

 

[ErichZannMusician]

15W = 8182
27W = 10212 (it’s way outside of its efficiency range at this point)

Those aren't verifiable leaks.

Bionic Squash said that LNL scores 1.5x MTL-U in multi-core benchmarks, and specifically mentioned CB23.

You can check on Notebookcheck that the Core Ultra 7 165U already scores 10K points in CB23 mT.

So either Bionic Squash and Notebookcheck are wrong or your so-called "leak" is wrong.

 

[H433x0n]

Those aren't verifiable leaks.

Bionic Squash said that LNL scores 1.5x MTL-U in multi-core benchmarks, and specifically mentioned CB23.

You can check on Notebookcheck that the Core Ultra 7 165U already scores 10K points in CB23 mT.

So either Bionic Squash and Notebookcheck are wrong or your so-called "leak" is wrong.

I know both the guy on twitter (Jaykihn) and bionic_squash, their info isn’t contradictory. The performance info on Intel’s internal slides that they’ve given to partners claim that LNL is 1.5x MTL-U when both are limited to 15W iirc.

LNL isn’t a CPU that’s designed for nT workloads like rendering images. The design goal is to give MacBook Air type battery life and performance to windows laptops. LNL is in a totally different segment. Intel is making ARL-H & MTL refresh for those that want nT performance.

LNL is awesome, it’s the best mobile chip from Intel in a long time. It’s meant to fend off ARM in the windows ecosystem. LNL’s performance in cinememe is irrelevant.

 

[ErichZannMusician]

I know both the guy on twitter (Jaykihn) and bionic_squash, their info isn’t contradictory. The performance info on Intel’s internal slides that they’ve given to partners claim that LNL is 1.5x MTL-U when both are limited to 15W iirc.

LNL isn’t a CPU that’s designed for nT workloads like rendering images. The design goal is to give MacBook Air type battery life and performance to windows laptops. LNL is in a totally different segment. Intel is making ARL-H & MTL refresh for those that want nT performance.

LNL is awesome, it’s the best mobile chip from Intel in a long time. It’s meant to fend off ARM in the windows ecosystem. LNL’s performance in cinememe is irrelevant.

If the leaked CB23 scores are from Jaykinn, then it is likely that those are outdated information.

Because the 165U scores 10.5K in CB23 mT with a 21 W PL1.

Shave 1000-1500 pts off from that if you want to compare at 15 W, and even then LNL should score 14K in CB23 MT.

But yes, all this is irrelevant. The point is that Lion Cove, even with acceptable levels of IPC uplift, will have no problem keeping up with the competition at similar clock speeds.

 

[scannall]

Leaks will shut them up

Leaks mean literally nothing. They are for entertainment purposes only. From any company. See, them. Enjoy them, but until real third party reviews come out there is no value.

 

[DavidC1]

The E core in Arrowlake clocks 200MHz faster than a super binned 14900KS, 100MHz higher than the rumored 4.6GHz. The increase couldn't have come free, and may be responsible for extra die area increases on the E core. They may have seen it as necessary, but I'm not too happy about it. They need to cut the clocks down, and even on E cores cut it to 12 stage or lower.

Apple's design decisions make a lot of sense. Another thing they can do is adopt their massive L1 cache scheme, which will lower power consumption and improve performance. Will be only possible at low enough clocks though.

The FPU blocks have been literally doubled, and it comes off of having FMA support on Gracemont, so it probably added 30% increase in area alone.

 

[Abwx]

Because the 165U scores 10.5K in CB23 mT with a 21 W PL1.

In the two Notebookcheck scores that you linked they measure around 35-40W in Cinebench for 10.2k-10.7k, for the higher score the laptop consume 60W, wich is certainly not 20W for the CPU.

 

[Det0x]

Can this be real ? 🤣

https://twitter.com/x/status/1812913340865474946

Intel Sets Higher Temperature "TJMax" & Power "Electrical DC" Specifications For Arrow Lake CPUs​

Intel's Arrow Lake & Panther Lake CPUs will feature a higher temperature limit of 105C compared to 100C on Raptor Lake CPUs.

Intel Arrow Lake CPUs Get Increased Temperature "TJMax" & Power Ratings Versus Raptor Lake, Up To 105C

Intel's Arrow Lake & Panther Lake CPUs will feature a higher temperature limit of 105C compared to 100C on Raptor Lake CPUs.

wccftech.com

Board makers such as MSI recently introduced a new BIOS feature called "CPU Fast Throttle Threshold" which allowed users to manually set TJMax limits beyond the official rating, up to 115C which will help you degrade your CPU silicon even faster and let you experience the joys of an unstable PC.

 

[poke01]

Can this be real ? 🤣

https://twitter.com/x/status/1812913340865474946

Intel Sets Higher Temperature "TJMax" & Power "Electrical DC" Specifications For Arrow Lake CPUs​

Intel's Arrow Lake & Panther Lake CPUs will feature a higher temperature limit of 105C compared to 100C on Raptor Lake CPUs.

Intel Arrow Lake CPUs Get Increased Temperature "TJMax" & Power Ratings Versus Raptor Lake, Up To 105C

Intel's Arrow Lake & Panther Lake CPUs will feature a higher temperature limit of 105C compared to 100C on Raptor Lake CPUs.

wccftech.com

That’s be expected Apples TJMax is 110C on their N3B node SoC, it’s got nothing to do with efficiency but then Apple also doesn’t push 200watts.

 

[Det0x]

Can this be real ? 🤣

https://twitter.com/x/status/1812913340865474946

Intel Sets Higher Temperature "TJMax" & Power "Electrical DC" Specifications For Arrow Lake CPUs​

Intel's Arrow Lake & Panther Lake CPUs will feature a higher temperature limit of 105C compared to 100C on Raptor Lake CPUs.

Intel Arrow Lake CPUs Get Increased Temperature "TJMax" & Power Ratings Versus Raptor Lake, Up To 105C

Intel's Arrow Lake & Panther Lake CPUs will feature a higher temperature limit of 105C compared to 100C on Raptor Lake CPUs.

wccftech.com

Seems like power limits also are going up.
Must say this don't fill me with confidence..

 

[SmokSmog]

ARL-S tops up at 8P+16E/24T, right? Does that mean the Ryzen 9 9950X with 16C/32T will dominate the nT benchmarks, or will the uplift from Lion Cove and Skymont make up for the loss of thread counts (HT threads)?

ARL 5.4ghzP 4.6ghzE should do 47K+ R23 MT going by current Intel claims about Lion cove and Skymont IPC.