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

Tigerick · Aug 22, 2022

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

511 · Apr 23, 2025

I for one am glad that itanium failed the way it did

Fjodor2001 · Apr 23, 2025

inquiss said:
Interesting data but I don't agree with the conclusion. Less memory bandwidth than a low clocked milan chip. If they give up the sorts of performance figures suggested, that's a couple of generations backwards from being memory starved.

That's a bit vague. Which specific data/info/conclusion in the post I linked to is it that you think is incorrect and why?

Abwx · Apr 23, 2025

Win2012R2 said:
Enables lots and lots of software to run NOW without changes, then add your own extentions to speed some stuff up if it's done better.

They moved because ops over 128 bits will beat 64 bits big time, plus it was terrible cheap choice to reuse registers for MMX, writing it was highly undesirable. Yes good thing Intel was dominant AND forward looking at the time, otherwise we'd still use MMX.

There were no 128b ops in SSE, you are confusing ops with registers width, ops were 32b precision, 64b precision came with SSE2 wich used 2 x 64b registers to store 2 operands.

jpiniero · Apr 23, 2025

511 said:
Cores are not for only Cinememe

We're talking about putting 32 or more E cores on the desktop socket. That seems pretty meme to me.

dullard · Apr 23, 2025

inquiss said:
Point 2...look at any new memory standard introduced....low volume initially with very high prices. Don't need a crystal balls here

I guess that I just don't think a few tens of dollars matters to someone buying a 52 core chip.

OneEng2 · Apr 23, 2025

dullard said:
I guess that I just don't think a few tens of dollars matters to someone buying a 52 core chip.

I just don't think there will be much volume sold at 52cores as it will likely be cost prohibitive for most.... and of limited utility to most people.

It seems like more of a Thread Ripper HPC product competitor. If that is the case, my question is how do you feed the beast?

inquiss · Apr 23, 2025

dullard said:
I guess that I just don't think a few tens of dollars matters to someone buying a 52 core chip.

How much memory are you buying here? This early in adoption prices will be closer to 5x to 10x ddr5 pricing. Just like when DDR launched

OneEng2 · Apr 23, 2025

inquiss said:
How much memory are you buying here? This early in adoption prices will be closer to 5x to 10x ddr5 pricing. Just like when DDR launched

Agree. Also, if history is any indicator, the cost of the routing and mb will increase some as well.

dullard · Apr 23, 2025

inquiss said:
How much memory are you buying here? This early in adoption prices will be closer to 5x to 10x ddr5 pricing. Just like when DDR launched

I'm back to my question, do you know DDR6 prices? Because your 5x to 10x pricing doesn't match history.

"When DDR4 launched it was approximately 30–40% more expensive than DDR3."

Kingston explains why DDR5 costs so much, and when it will be the dominant memory tech for PCs

Intel's Alder Lake is the first platform to support DDR5, but what can we expect from the future of RAM?

www.pcgamer.com

"DDR5 Will Probably Cost 50% to 60% More Than DDR4"

DDR5 Will Probably Cost 50% to 60% More Than DDR4

Steep premium for early adopters

www.tomshardware.com

And those price premiums didn't last long. "Overall, one DDR4-2133 chip became 18.4% cheaper in about two months and lost nearly 50% of its price in about eight months"

Price Check: Price Gap Between DDR3 and DDR4 Memory Almost Gone

www.anandtech.com

OneEng2 · Apr 23, 2025

dullard said:
I'm back to my question, do you know DDR6 prices? Because your 5x to 10x pricing doesn't match history.

"When DDR4 launched it was approximately 30–40% more expensive than DDR3."

Kingston explains why DDR5 costs so much, and when it will be the dominant memory tech for PCs

Intel's Alder Lake is the first platform to support DDR5, but what can we expect from the future of RAM?

www.pcgamer.com

"DDR5 Will Probably Cost 50% to 60% More Than DDR4"

DDR5 Will Probably Cost 50% to 60% More Than DDR4

Steep premium for early adopters

www.tomshardware.com

And those price premiums didn't last long. "Overall, one DDR4-2133 chip became 18.4% cheaper in about two months and lost nearly 50% of its price in about eight months"

Price Check: Price Gap Between DDR3 and DDR4 Memory Almost Gone

www.anandtech.com

And you imagine that Intel can afford 40-50% more cost for DDR6?

Perhaps in HPC, but certainly not in desktop.

dullard · Apr 23, 2025

OneEng2 said:
And you imagine that Intel can afford 40-50% more cost for DDR6?

Perhaps in HPC, but certainly not in desktop.

Intel isn't the one buying the memory. 32 GB of the fastest DDR5 officially supported by Alder Lake Desktop is $80. 50% more would be just an extra ~$40. Being concerned about $40 on what might be a $1k CPU is laughable. Of course that is just an assumption unless someone here can point to actual DDR6 pricing. Here is the first example that popped up on a search. https://www.newegg.com/patriot-memory-32gb-ddr5-6400/p/0RN-002U-004U1

If you need giant amounts of memory, the cost difference would likely be higher. But, again, I disagree that 52 cores with giant amounts of memory is main desktop territory. It will be HPC territory.

inquiss · Apr 23, 2025

dullard said:
I'm back to my question, do you know DDR6 prices? Because your 5x to 10x pricing doesn't match history.

"When DDR4 launched it was approximately 30–40% more expensive than DDR3."

Kingston explains why DDR5 costs so much, and when it will be the dominant memory tech for PCs

Intel's Alder Lake is the first platform to support DDR5, but what can we expect from the future of RAM?

www.pcgamer.com

"DDR5 Will Probably Cost 50% to 60% More Than DDR4"

DDR5 Will Probably Cost 50% to 60% More Than DDR4

Steep premium for early adopters

www.tomshardware.com

And those price premiums didn't last long. "Overall, one DDR4-2133 chip became 18.4% cheaper in about two months and lost nearly 50% of its price in about eight months"

Price Check: Price Gap Between DDR3 and DDR4 Memory Almost Gone

www.anandtech.com

Cool, an article from Jan 2022, DDR5 launched in October 2020. Price premium wasn't 30 to 40% at launch. Oh no.

dullard · Apr 23, 2025

inquiss said:
Cool, an article from Jan 2022, DDR5 launched in October 2020. Price premium wasn't 30 to 40% at launch. Oh no.

If you disagree with Kingston and PCGamer then take it up with them. Show me 5x to 10x for comparable memory.

LightningZ71 · Apr 23, 2025

Isn't 48+4 cores kind of pushing it on RAM capacity? If they stay dual channel and DDR5, that's a hard limit of 4 x 64GB or 256 GB ram, roughly 5-6GB per core/process. It was kind of tight with 32 threads on the dual CCD Ryzen processors for anything non-trivial in nature. It's one of the reasons that TR platform supported 4+ channels of RAM: capacity.

Fjodor2001 · Apr 23, 2025

LightningZ71 said:
Isn't 48+4 cores kind of pushing it on RAM capacity? If they stay dual channel and DDR5, that's a hard limit of 4 x 64GB or 256 GB ram, roughly 5-6GB per core/process. It was kind of tight with 32 threads on the dual CCD Ryzen processors for anything non-trivial in nature. It's one of the reasons that TR platform supported 4+ channels of RAM: capacity.

For what use cases would GB RAM / CPU core matter? I have a hard time seeing that max 256 GB RAM for 52 cores would be an issue in most cases.

Just as an example. Transcoding a 100 GB movie. You likely only have max 1 GB in RAM at a time anyway for the data being processed, probably much less since only a few frames are needed in RAM at the same time anyway.

OneEng2 · Apr 23, 2025

dullard said:
Being concerned about $40 on what might be a $1k CPU is laughable.

When you multiply that 40$ by 100 million units, suddenly it isn't a laughing matter at all.

dullard said:
If you need giant amounts of memory, the cost difference would likely be higher. But, again, I disagree that 52 cores with giant amounts of memory is main desktop territory. It will be HPC territory.

I agree. These very high core count discussions seem to ignore cost and utility to the masses. I expect on the desktop for next gen, AMD will top out at 24c/48T and Intel will top out around 32c/32T. Both will be on DDR5 and will likely be bandwidth limited in apps that can actually load all cores (and threads).

In laptop, I am guessing half that but having a powerful graphics unit that will eat up the bandwidth for some implementations.

Fjodor2001 · Apr 23, 2025

OneEng2 said:
Intel will top out around 32c/32T

What P/E/LPE core count mix has been leaked so far at 32C/32T for Nova Lake?

Also, note that there has been a 52 core variant mentioned in news media lately.

OneEng2 said:
Both will be on DDR5 and will likely be bandwidth limited in apps that can actually load all cores (and threads).

What DDR5 bandwidth are you assuming, and why would the cores necessarily be bandwidth limited? I think that varies a lot per use case. In many use cases you don’t need that much bandwidth, as the cores are doing heavy CPU crunching on a limited data set.

inquiss · Apr 23, 2025

dullard said:
If you disagree with Kingston and PCGamer then take it up with them. Show me 5x to 10x for comparable memory.

Why would I take it up with them. I am suggesting to you that looking at memory prices in 2022 for a product that launched in 2020 isn't really a flex.

When DDR5 launched, 32 GB of 4800 was approx $1000. What's the cost of that today? $80? Did they drop the following year to around $400, yep. Is that still more than a few tens of dollars? Also yep. Especially since I imagine if you have 52c you're gonna go for the higher capacity, be mad not to, which is gonna cost you too

dullard · Apr 23, 2025

inquiss said:
Why would I take it up with them. I am suggesting to you that looking at memory prices in 2022 for a product that launched in 2020 isn't really a flex.

When DDR5 launched, 32 GB of 4800 was approx $1000. What's the cost of that today? $80? Did they drop the following year to around $400, yep. Is that still more than a few tens of dollars? Also yep. Especially since I imagine if you have 52c you're gonna go for the higher capacity, be mad not to, which is gonna cost you too

The 2022 article was discussing prices from 2020. If you had read it you would know. I'll quote it again "When DDR4 launched it was approximately 30–40% more expensive than DDR3"

You think 32 GB of DDR5-4800 was $1000 at launch? Or even $400? Hah. Here was the very first 32 GB DDR5-4800 kits at $310.99 (5 months before any CPU could use them).

The First DDR5-4800 32GB RAM Kit Hits the Market for $310

Grab some RAM for next-gen platforms that aren't out yet.

www.tomshardware.com

And here is a week or so before Alder Lake's launch with brand-name prices as low as $211.99 for 32 GB of DDR5-4800. At least in my world, $211.99 is not "approx $1000".

https://www.pcmag.com/news/ddr5-ram-pricing-ranges-from-116-to-369

Clearly you are just making up random "facts" and pretending that people can't check the real values.

inquiss · Apr 24, 2025

dullard said:
The 2022 article was discussing prices from 2020. If you had read it you would know. I'll quote it again "When DDR4 launched it was approximately 30–40% more expensive than DDR3"

You think 32 GB of DDR5-4800 was $1000 at launch? Or even $400? Hah. Here was the very first 32 GB DDR5-4800 kits at $310.99 (5 months before any CPU could use them).

The First DDR5-4800 32GB RAM Kit Hits the Market for $310

Grab some RAM for next-gen platforms that aren't out yet.

www.tomshardware.com

And here is a week or so before Alder Lake's launch with brand-name prices as low as $211.99 for 32 GB of DDR5-4800. At least in my world, $211.99 is not "approx $1000".
View attachment 122640

https://www.pcmag.com/news/ddr5-ram-pricing-ranges-from-116-to-369

Clearly you are just making up random "facts" and pretending that people can't check the real values.

DDR5: Availability Improving, Prices Dropping

Time for an upgrade?

www.tomshardware.com

Hard to find prices at camelcamel myself as the memory available then is different to the available memory listed today. But this gives a flavour.

deasd · Apr 24, 2025

I tried to find the original source of 16P+32E rumor and unearth some interesting ancient pieces, yeah it was related to MLID:

Intel reportedly planning Arrow Lake Refresh featuring 8P+32E cores for 2025 debut - VideoCardz.com

Arrow Lake Refresh with 40 cores Intel’s latest client CPU roadmap has been presented by Moore’s Law is Dead. The video reiterates some rumors and solidifies the latest information about Intel’s plans for the next three years. However, the further we look into rumored and official roadmaps, the...

videocardz.com

I don't wanna talk about this guy but look at what he said in Oct 2023. Summary below:

ARL: 8P+16E+2LPE cores
LNL: 4P+4E+2LPE cores
ARL Refresh: 8P+32E+2LPE cores
PTL: 4P+8E+4LPE cores
NVL: 16P+32E+4LPE cores ===== source!!

After the Arrowlake announcement and core layout photo being published, it feels like that 8P+32E is technically impossible, unless Intel change the layout for a Refresh product but the information around nowaday don't hint anything about it. It might indicate either Intel changed its plan way earlier or anything more than 16E never exist before.

Now NVL 16P+32E is brought into sight and people suggests that it's (8P+16E)*2 CCDs, which sounds like a rumor mess to me.

Also those products which were fabbed in TSMC are wayyy different from the rumor, looks like the concern of cost drives final decision.

Win2012R2 · Apr 24, 2025

Abwx said:
There were no 128b ops in SSE, you are confusing ops with registers width, ops were 32b precision, 64b precision came with SSE2 wich used 2 x 64b registers to store 2 operands.

Naturally I meant ops that work on 128 bit registers.

Abwx · Apr 24, 2025

Win2012R2 said:
Naturally I meant ops that work on 128 bit registers.

That s not 128b registers as such but 4 registers of 32b each, SSE required that much registers because it wasnt as flexible as 3Dnow, i already posted this :

One advantage of 3DNow! is that it is possible to add or multiply the two numbers that are stored in the same register. With SSE, each number can only be combined with a number in the same position in another register. This capability, known as horizontal in Intel terminology, was the major addition to the SSE3 instruction set.

To wich we can add this :

because it is aliased to the x87 FPU, the 3DNow! and MMX register states can be saved and restored by the traditional x87 F(N)SAVE and F(N)RSTOR instructions. This arrangement allowed operating systems to support 3DNow! with no explicit modifications, whereas SSE registers required explicit operating system support to properly save and restore the new XMM registers(via the added FXSAVE and FXRSTOR instructions.)

End of debate.

Win2012R2 · Apr 24, 2025

Abwx said:
That s not 128b registers as such but 4 registers of 32b each, SSE required that much registers because it wasnt as flexible as 3Dnow, i already posted this :

It's still 128 bits operated on at the same time vs 64 bits, that was my point, and it's a register just like AVX-512 got 512 bits registers, even though they are also split in similar way as SSE.

Abwx · Apr 24, 2025

Win2012R2 said:
It's still 128 bits operated on at the same time vs 64 bits, that was my point, and it's a register just like AVX-512 got 512 bits registers, even though they are also split in similar way as SSE.

That s FI two 32b operands stored in two 32b registers, the result of the op (for instance 32b FMUL or FADD) is stored in a third 32b register, the fourth 32b register is used to store the result if it s needed for a following operation as the third register would be written off at the next step, that s why there s 4x 32b registers wich you confuse as being 128b.

Actual 128b registers exist only since AVX128.

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

Senior member

Attachments

Golden Member

Diamond Member

Lifer

Lifer

Elite Member

Senior member

Senior member

Senior member

Elite Member

Senior member

Elite Member

Senior member

Elite Member

Platinum Member

Diamond Member

Senior member

Diamond Member

Senior member

Elite Member

Senior member

Senior member

Senior member

Lifer

Senior member

Lifer