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	12 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 ?	12 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.)

 

[coercitiv]

The reason I did this was because of your post regarding my trying to outsmart the Intel engineers at various power levels for E and P frequencies. I was curious.

Back then I did it out of curiosity as well, if you remember I was surprised to see the desktop chips seem to behave better at low wattage than mobile Alder Lake chips. At the time I think I limited max P core clocks around 2.6-2.8Ghz for the 35W test, I remember it was above the expected frequency in MT loads and it somehow "fixed" the quirk, allowing me to run the MT test.

 

[moinmoin]

I highly doubt Zen 4c will have similar frequencies to normal Zen 4. We know they're not lowering IPC, and it's the same process, so really the only knob left to get a lot of density/power gains is frequency reduction. If Zen 4c could hit clocks similar to Zen 4, then frankly Genoa wouldn't have much reason to exist at all.

I've stated before that I consider Zen 4c as AMD extending the TAM of their previous mobile focused cores to servers. With those mobile cores the frequency was around the efficiency inflection point of the desktop/server core which usually is close to the base frequency of the latter. We'll see.

 

[mikk]

Roadmap update from MLID:

- Redwood Cove is targeting 15-25% more IPC over Raptor Cove
- Lion Cove is aiming to bring 22-34% more IPC over Redwood Cove
- Panther Cove is aiming to bring another massive IPC increase

 

[Tigerick]

OK, I am trying to compile next generation desktop CPU from both Intel and AMD which are supposed coming in 2024:

Intel 15th ARL-S	Cores	Total Cores + Threads	L3 Cache	iGPU	VS	AMD 8000 Zen 5	Total Cores + Threads	Shared L2 Cache	External L3 V-Cache
Core i5 (T)	6P + 8E	14 (20)	24 MB?	4 Xe Cores = 512ALU		8600X	12 (24)	12 MB	64 MB
Core i7 (T)	8P + 8E	16 (24)	30 MB?	4 Xe Cores = 512ALU		8700X	16 (32)	16 MB	64 MB
Core i9	8P + 16E	24 (32)	36 MB?	4 Xe Cores = 512ALU		8900X	24 (48)	24 MB	128 MB?
						8950X	32 (64)	32 MB	128 MB?

• The above comparison chart is preliminary, will be keep on updating it. Any insights please let me know.

 

[JayMX]

OK, I try to compile next desktop CPU from both Intel and AMD in Q4 2023/Q1 2024.The above comparison chart is preliminary, will be keep on updating it. Any insights please let me know.

Where are those "ZEN5/ Ryzen 8k" core numbers come from?
or is it just a speculation/ estimate of yours?

 

[lightisgood]

Roadmap update from MLID:

MLID said that "now it is expected to have at least a slight speed regression on the Intel 4 node compared to Raptor Lake".

I disagree with this.
We know that MTL had overcome TSV tax.

 

[Exist50]

Roadmap update from MLID:

- Redwood Cove is targeting 15-25% more IPC over Raptor Cove
- Lion Cove is aiming to bring 22-34% more IPC over Redwood Cove
- Panther Cove is aiming to bring another massive IPC increase

Lmao, I don't think it even makes sense to post this except for the comedic value.

 

[nicalandia]

delete

 

[Geddagod]

Raichu on twitter:

https://twitter.com/x/status/1590669654015741952

Lion Cove isn't part of royal core

 

[dangerman1337]

Raichu on twitter:

https://twitter.com/x/status/1590669654015741952

Lion Cove isn't part of royal core

Though I think Intel's Roadmap has shifted so much IDK what is "Royal Core" anymore. Like Arrow Lake a 2023 product now a (2H?) 2024 product.

Though I'll probably build a Z890 MTL-S i5 system hopefully Summer-ish next year and swap that i5 for a ARL-S i7 from the sounds of it (I mean Raichu didn't seem to shot down MTL-S going up to 6P?). I do wonder if LGA 1851 is going to be a long supported socket unlike previous sockets because it seems to be a very beefy socket for the long run. And I expect DDR5 for consumer to last a long time so making LGA 1851 long running since Motherboards these days are incredibly expensive (Z370 Taichi costed like £200 in the UK, now Taichi motherboards cost way over double that).

Motherboards damn expensive these days.

 

[inf64]

So MLID thinks MTL will target Golden Cove-like IPC improvement? If he believes his "sources", he's lost his mind I'm afraid. It reminds me of his 29% IPC jump for Zen 4 (Chips and cheese also posted that exact number for some reason).

 

[IntelUser2000]

Though I think Intel's Roadmap has shifted so much IDK what is "Royal Core" anymore. Like Arrow Lake a 2023 product now a (2H?) 2024 product.

That's cause MLID is pretty much gaslighting everyone and he's gained so much fame that every "press"(regurgitators is more like it) is referring to him. Look how he constantly changes his predictions.

Raichu and @Exist50 is right. Royal Core happens way later, like post-Panther Lake. Everything else is just a buildup of what they have right now. Also he said Arrowlake-S is coming almost at the same time as Meteorlake-S.

@Tigerick Core i9 and likely even i7 is Arrowlake with 8+16 not 6+16 Meteorlake.

he's lost his mind I'm afraid.

I think he's perfectly aware of what he's doing.

 

[Exist50]

Though I think Intel's Roadmap has shifted so much IDK what is "Royal Core" anymore. Like Arrow Lake a 2023 product now a (2H?) 2024 product.

Royal is Royal. It's really that simple. No double meaning.

That's cause MLID is pretty much gaslighting everyone and he's gained so much fame that every "press"(regurgitators is more like it) is referring to him. Look how he constantly changes his predictions.

It's baffling, really. I guess you need nothing more than persistence, ego, and a complete willingness to lie to become a professional "leaker".

 

[poke01]

Royal is Royal. It's really that simple. No double meaning.

How would we know what Royal is?

Like Royal could be Nova Lake? Is Royal going THAT powerful that as soon it leaks/releases that we know that that CPU arch is Royal based?

 

[Exist50]

How would we know what Royal is?

Fair point. But I hope that even if you won't take my word for it, Raichu's is worth more than MLID's, low bar though that may be.

Like Royal could be Nova Lake? Is Royal going THAT powerful that as soon it leaks/releases that we know that that CPU arch is Royal based?

I do expect Royal will be that distinctive, yes. But even if not, it's the name of the core. So they'll probably just say it...

 

[A///]

Side effects of drug use and the development of a self important complex.

 

[eek2121]

From what I gather, MTL-S sounds like this:

P-Cores - IPC: single digit gains - Clocks: minor gains, especially in multicore workloads.

E-Cores - IPC: Double digit gains - Clocks: Decent gains.

Both: lower power consumption.

So I imagine total single thread improvements will be somewhere around 7-12%, multicore will likely be 30-40%. Just a guesstimate based on what I am reading now. Power limits down to 233W maybe?

 

[Kaluan]

From what I gather, MTL-S sounds like this:

P-Cores - IPC: single digit gains - Clocks: minor gains, especially in multicore workloads.

E-Cores - IPC: Double digit gains - Clocks: Decent gains.

Both: lower power consumption.

So I imagine total single thread improvements will be somewhere around 7-12%, multicore will likely be 30-40%. Just a guesstimate based on what I am reading now. Power limits down to 233W maybe?

I think (trustworthy) rumours surrounding MTL's TDP and PL2 will be the biggest indicator if Intel has something decent lined up, we'll see it in the (admittedly fake, since RTL needs to run unlocked/over 253W PL2 to truly compete with Raphael) TDP/PL2 numbers.
Lower than 241-253W desktop/115-170W mobile turbos? Will probably be good. Same? 50/50. More? It will be an unmitigated disaster, like Cannon Lake.

 

[IntelUser2000]

I love how specific the numbers are, even though performance benchmarking cannot be.

18% Sunny Cove and 19% Golden Cove. 51% for Zen 1. Again 19% for Zen 3.

Really? You could literally change one benchmark and swing it easily by 1-2%. 18 or 19% is called 15-20%. That's how we used to do it back in the days. That's the fundamentals you learn in grade-school math!

Same with power numbers.

Lower than 241-253W desktop/115-170W mobile turbos? Will probably be good. Same? 50/50. More? It will be an unmitigated disaster, like Cannon Lake.

Mobile is fine because they cannot have PL1=PL2 or the thermal system will be overloaded, unlike desktop where bricks are used for heatsinks. That in itself is the check for out of bounds TDP limits.

We may be headed to a future in desktops where the case itself is the heatsink. Actually there's a passive heatsink where it is indeed that big. Then you can crank CPU/GPU power to the limit where 1500W wall socket limit will be reached.

 

[TESKATLIPOKA]

I expect Meteor Lake to do great even against Phoenix.
Here is a table about MT performance at different power limits. Link: ComputerBase

As you can see, If you limit the power to 45W then 13900K is the most efficient one and 6+8 13600K is just a bit worse than 7700x.
A monolithic 8C Phoenix should perform better at 45W in this table, not sure by how much.
Increasing the power for 7700x by 20W(+44%) increased performance by just 20%(82% -> 98%), but the chiplet overhead is hardly 20W, so Phoenix at 45W shouldn't be 20% faster unless the supposed N4 process and additional changes makes up the difference. Let's say Phoenix will be 15% faster at 45W than 7700x limited to 45W.

6+8 Meteor Lake should have the same configuration as 13600K and If you check the above table the 13700K with +2 P cores performs only 1% better at 45W, so having only 6+8 is not that surprising for a mobile chip.

Meteor Lake should use Intel 4 and that supposedly reduces power consumption by 40% at ISO frequency or being 20% faster at ISO power.
I made a table based on the above table where I apply Intel 4 gains to Raptor Lake.

13700k at 65W is 100%	7600x	7700x	7900x	7950x	Phoenix	13600K	13700K	13900K
Performance at 45W	69%	82%	83%	84%	94%	79%	80%	92%
Intel 4 at ISO power	69%	82%	83%	84%	94%	95%	96%	110%
Intel 4 at ISO frequency	69%	82%	83%	84%	94%	79% at 27W	80% at 27W	92% at 27W

Meteor Lake should use Redwood Cove for P-cores and Crestmont for E-cores, meaning higher IPC, so the performance at 45W would be even better than what I calculated above.

P.S. Only today I found out that Meteor Lake supposedly has additional 2 LPE cores inside the SoC, meaning It has in reality 16 cores in total(6P+8E+2LPE). Videocardz

 

[Hulk]

Meteor Lake should use Intel 4 and that supposedly reduces power consumption by 40% at ISO frequency or being 20% faster at ISO power.

What is the source for this claim again? Seems like high altitude pastry.

Otherwise known as "pie in the sky."

 

[TESKATLIPOKA]

What is the source for this claim again? Seems like high altitude pastry.

Otherwise known as "pie in the sky."

Meteor lake using Intel 4 process. Link or Link
Intel 4 process info where they mention -40% and +20%. Link

I also found this. Link

Actually, It looks like It was at ~2.1GHz. At 3GHz you can still shave off ~1/3 of power or gain ~15% of performance.

 

[igor_kavinski]

We may be headed to a future in desktops where the case itself is the heatsink. Actually there's a passive heatsink where it is indeed that big. Then you can crank CPU/GPU power to the limit where 1500W wall socket limit will be reached.

I wish they would do that and redesign the enthusiast PC design from the ground up. Hot air circulating in a confined space is just stupid.

 

[DrMrLordX]

I wish they would do that and redesign the enthusiast PC design from the ground up. Hot air circulating in a confined space is just stupid.

There are a few case designs that attempt just that:

 

[Glo.]

I expect Meteor Lake to do great even against Phoenix.
Here is a table about MT performance at different power limits. Link: ComputerBase
View attachment 70965
As you can see, If you limit the power to 45W then 13900K is the most efficient one and 6+8 13600K is just a bit worse than 7700x.
A monolithic 8C Phoenix should perform better at 45W in this table, not sure by how much.
Increasing the power for 7700x by 20W(+44%) increased performance by just 20%(82% -> 98%), but the chiplet overhead is hardly 20W, so Phoenix at 45W shouldn't be 20% faster unless the supposed N4 process and additional changes makes up the difference. Let's say Phoenix will be 15% faster at 45W than 7700x limited to 45W.

6+8 Meteor Lake should have the same configuration as 13600K and If you check the above table the 13700K with +2 P cores performs only 1% better at 45W, so having only 6+8 is not that surprising for a mobile chip.

Meteor Lake should use Intel 4 and that supposedly reduces power consumption by 40% at ISO frequency or being 20% faster at ISO power.
I made a table based on the above table where I apply Intel 4 gains to Raptor Lake.

13700k at 65W is 100%	7600x	7700x	7900x	7950x	Phoenix	13600K	13700K	13900K
Performance at 45W	69%	82%	83%	84%	94%	79%	80%	92%
Intel 4 at ISO power	69%	82%	83%	84%	94%	95%	96%	110%
Intel 4 at ISO frequency	69%	82%	83%	84%	94%	79% at 27W	80% at 27W	92% at 27W

Meteor Lake should use Redwood Cove for P-cores and Crestmont for E-cores, meaning higher IPC, so the performance at 45W would be even better than what I calculated above.

P.S. Only today I found out that Meteor Lake supposedly has additional 2 LPE cores inside the SoC, meaning It has in reality 16 cores in total(6P+8E+2LPE). Videocardz

For APUs its not the CPU side that is the key, but its always the iGPU.

Till this day, we've always had an APU/SOC that always had CPU more powerful than the iGPU, but we're entering an era, where its the iGPU that is more powerful than the CPU.

And the performance of the iGPU will always define who won this round of competition. Not the CPU.