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 · Monday at 12:44 PM

igor_kavinski said:
Assuming they have already solved some or all of the tile related problems in Panther Lake, it seems logical that they can do something better with the ARL Refresh tiles, unless they decide it's too much work with not enough time.

Panther Lake is near monolithic only iGPU being the separate one.

igor_kavinski · Monday at 12:51 PM

511 said:
Panther Lake is near monolithic only iGPU being the separate one.

RAM is not on package so it will face the same latency issues as ARL unless they have solved that somehow.

If the above is to be believed, they lose this power advantage with Panther Lake and need to make up for it somehow. Either larger cache to reduce memory traffic or depend entirely on 18A's power efficiency.

511 · Monday at 12:57 PM

igor_kavinski said:
RAM is not on package so it will face the same latency issues as ARL unless they have solved that somehow.

View attachment 124770

If the above is to be believed, they lose this power advantage with Panther Lake and need to make up for it somehow. Either larger cache to reduce memory traffic or depend entirely on 18A's power efficiency.

IMC is on board the CPU die in Panther Lake also this is LNL specific optimization i doubt any non MoP solution will come close to this one.

igor_kavinski · Monday at 12:59 PM

511 said:
IMC is on board the CPU die in Panther Lake

Is this confirmed officially?

Fjodor2001 · Monday at 1:05 PM

coercitiv said:
Improving the NPU is still an exercise in futility at this point, it does not help short term and is likely to be too little for the future. Same applies to AMD.

Needed for the 40+ TOPS to get the Microsoft AI PC sticker. PC manufacturers will want that on their boxes when the Windows 12 upgrade cycle starts.

511 · Monday at 1:06 PM

igor_kavinski said:
Is this confirmed officially?

not officially but has been leaked somewhere i don't remember exactly where.

511 · Monday at 1:07 PM

Fjodor2001 said:
Needed for the 40+ TOPS to get the Microsoft AI PC sticker. PC manufacturers will want that on their boxes when the Windows 12 upgrade cycle starts.

Windows 12 is 70 TOPs . /jk

dullard · Monday at 1:09 PM

igor_kavinski said:
RAM is not on package so it will face the same latency issues as ARL unless they have solved that somehow.

From Chips and Cheese, they feel Intel has two separate latency issues which hold the performance back. (1) All the stops on the ring bus which include stops for cross die transfer. And (2) Intel's cache coherency protocols. With respect to Lunar Lake, there are only a few ring stops and the memory is on board and yet there is still high latency. That is their evidence that they think issue #2 is playing a significant role.

Cache coherency protocols ensure only one core can have a particular address cached in modified state. Thus the requesting core in this test will miss in all of its private cache levels. When the core with modified data gets a probe, it has to check all of its own private cache levels, both to read out the modified data and ensure the address is invalidated from its own caches. Lion Cove goes from two levels of core-private data caches to three, adding another step in the process.

A look back at Lunar Lake supports this theory. Lion Cove also sees high core-to-core latency there, even though Lunar Lake’s ring bus only serves four cores. Lunar Lake also displays high cross-cluster latencies, even though those transfers don’t cross die boundaries. How a memory subsystem handles cache coherency operations at the logical level can make a bigger difference than physical die boundaries. and Lunar Lake may be a good example.

Examining Intel's Arrow Lake, at the System Level

Arrow Lake is the codename for Intel's newest generation of high performance desktop CPUs.

chipsandcheese.com

In other words, having the RAM on package helps, but does not solve Intel's high latency.

Fjodor2001 · Monday at 1:10 PM

511 said:
Windows 12 is 70 TOPs . /jk

Where did you find that info? And in any case, I assume then Intel & AMD will be gunning for 70 TOPS (or whatever is needed) instead of 40 TOPS.

511 · Monday at 1:12 PM

Fjodor2001 said:
Where did you find that info? And in any case, I assume then Intel & AMD will be gunning for 70 TOPS (or whatever is needed) instead of 40 TOPS.

i said /jk at the end.

On a Sidenote Looks like my jokes/sarcasm are difficult to understand. 🥲

igor_kavinski · Monday at 1:16 PM

dullard said:
In other words, having the RAM on package helps, but does not solve Intel's high latency.

Wow. I missed that. So Lunar Lake is held back too!

511 · Monday at 1:16 PM

dullard said:
From Chips and Cheese, they feel Intel has two separate latency issues which hold the performance back. (1) All the stops on the ring bus which include stops stops for cross die transfer. And (2) Intel's cache coherency protocols. With respect to Lunar Lake, there are only a few ring stops and the memory is on board and yet there is still high latency. That is their evidence that they think issue #2 is playing a significant role.

Examining Intel's Arrow Lake, at the System Level

Arrow Lake is the codename for Intel's newest generation of high performance desktop CPUs.

chipsandcheese.com

In other words, having the RAM on package helps, but does not solve Intel's high latency.

Don't forget AMD Strix Point latency is in similar boat to Lunar Lake. These designs are optimized for power instead of performance.

I was never a fan of adding cache hieracrhy these adds more problem than they solve them i hope they fix it in Panther Cove.

igor_kavinski · Monday at 1:17 PM

511 said:
On a Sidenote Looks like my jokes/sarcasm are difficult to understand. 🥲

It is widely believed that you work for Intel so most of the stuff you say has some "official" ring to it

Fjodor2001 · Monday at 1:20 PM

511 said:
i said /jk at the end.

On a Sidenote Looks like my jokes/sarcasm are difficult to understand. 🥲

You should have posted this instead

511 · Monday at 1:33 PM

igor_kavinski said:
It is widely believed that you work for Intel so most of the stuff you say has some "official" ring to it

LMAO I don't work for Intel who is spreading this rumor if you are gonna do it do it properly. Make a thread at least with proper leaks 🤣.

igor_kavinski · Monday at 1:36 PM

511 said:
LMAO I don't work for Intel

Do you work for someone who works for Intel? Maybe stealing shredded documents out of their waste basket and spending all night gluing them together?

Thibsie · Monday at 3:56 PM

511 said:
https://twitter.com/x/status/1808621251965169774

That is hilarious 😂

DavidC1 · Monday at 11:14 PM

igor_kavinski said:
RAM is not on package so it will face the same latency issues as ARL unless they have solved that somehow.

Let me ask you this then. How was Raptorlake so fast in terms of latency? Think before you post man.

On package memory is absolutely insignificant in terms of latency. If it was the sole factor then the smartphone/tablet parts with memory right UNDERNEATH the chip would be absolutely the fastest, yet they are slow as snails.

Saying whether it's on package or not matters for latency means it's mostly limited by speed of electrons(which is similar to light), but RAM is hundreds of times too slow for that to matter.

Lunarlake's on package memory play is entirely about power, nothing else.

dullard said:
In other words, having the RAM on package helps, but does not solve Intel's high latency.

The real issue is that Intel is a shadow of what it used to be. C&C article is saying in a very roundabout manner that Intel has crappy engineering due to decades of mismanagement.

Before the brain drain, they could take a garbage design and make it ok.

511 · Monday at 11:23 PM

DavidC1 said:
The real issue is that Intel is a shadow of what it used to be, and the endless brain drain over past few decades is the reason for low performance on Intel.

Before the brain drain, they could take a garbage design and make it ok.

Totally many have retired or switched Jobs. Apple's design team is almost Ex-Intel. I have also seen somewhere Nvidia Litho team is ex Intel as well.

DavidC1 · Monday at 11:24 PM

511 said:
Totally many have retired or switched Jobs. Apple's design team is almost Ex-Intel.

Sometimes I get annoyed at Chips & Cheese articles. They KNOW perfectly why Arrowlake sucks.

Writing 10,000 word articles to try to "technically explain" it away doesn't change that.

It's like trying to explain why someone has shot themselves on their foot.

Thunder 57 · Monday at 11:39 PM

DavidC1 said:
Sometimes I get annoyed at Chips & Cheese articles. They KNOW perfectly why Arrowlake sucks.

Writing 10,000 word articles to try to "technically explain" it away doesn't change that.

It's like trying to explain why someone has shot themselves on their foot.

I like their content usually. But I don't like them because they posed an open question; Should we move to Substack? The responses were a resounding no. They did it anyway. Why ask if you are going to do it regardless?

DavidC1 · Monday at 11:42 PM

Thunder 57 said:
I like their content usually. But I don't like them because they posed an open question; Should we move to Substack? The responses were a resounding no. They did it anyway. Why ask if you are going to do it regardless?

I like them too, but they have a lot of spurious articles. They are the best at what they are doing, which is the sad part. Because during the peak of Anandtech, the articles out on the internet were simply superb. Xbitlabs, Anandtech, Beyond3D, Realworldtech, they were miles better.

Guess it's way better than video articles which reduce content by 10x while increasing time to search for relevant data by 10x too. Pictures might be worth a thousand words, but thousand videos might not be worth one article.

They feel second rate. Also older articles didn't shy away from critiquing what they were reviewing. C&C doesn't know how or are scared to do it.

Thunder 57 · Monday at 11:52 PM

DavidC1 said:
I like them too, but they have a lot of spurious articles. They are the best at what they are doing, which is the sad part. Because during the peak of Anandtech, the articles out on the internet were simply superb. Xbitlabs, Anandtech, Beyond3D, Realworldtech, they were miles better.

Guess it's way better than video articles which reduce content by 10x while increasing time to search for relevant data by 10x too. Pictures might be worth a thousand words, but thousand videos might not be worth one article.

They feel second rate. Also older articles didn't shy away from critiquing what they were reviewing. C&C doesn't know how or are scared to do it.

Anandtech used to be the gold standard. Xbitlabs was nice they are back again but a joke. Never really used Beyond3D. RWT was great too.

DavidC1 · Monday at 11:56 PM

Thunder 57 said:
Anandtech used to be the gold standard. Xbitlabs was nice they are back again but a joke. Never really used Beyond3D. RWT was great too.

I'm talking about the old Xbitlabs of course. There's another site I can't remember which had absolute gems of articles. Technically they were superior to Anandtech, and I learned a lot of fundamentals there. Anandtech had more articles and had more relevant bits most of the time plus they were closer to industry.

The writers then put in a lot more work compared to what they got out of it. The writers nowadays put a fraction of work for lot more reward. Now it's all about chasing that juicy viewer count or $. Back then it felt as if the writers owed their fans a service. I wasn't a fan of Ian Cuttress either. I could immediately feel the quality went down.

Thunder 57 · Tuesday at 12:05 AM

DavidC1 said:
I'm talking about the old Xbitlabs of course. There's another site I can't remember which had absolute gems of articles. Technically they were superior to Anandtech, and I learned a lot of fundamentals there. Anandtech had more articles and had more relevant bits most of the time plus they were closer to industry.

The writers then put in a lot more work compared to what they got out of it. The writers nowadays put a fraction of work for lot more reward. Now it's all about chasing that juicy viewer count or $. Back then it felt as if the writers owed their fans a service. I wasn't a fan of Ian Cuttress either. I could immediately feel the quality went down.

I thought Ian was great at least when it came to CPU's. If anything I think Ryan was responsible for the downfall of Anandtech. He must've pissed someone off to stop getting GPU's to review. Then he went and blamed it on wildfires. And he turned down Ian's idea to have a youtube video portion for Anandtech in addition to the written articles.

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

Senior member

Attachments

Platinum Member

Lifer

Platinum Member

Lifer

Diamond Member

Platinum Member

Platinum Member

Elite Member

Diamond Member

Platinum Member

Lifer

Platinum Member

Lifer

Diamond Member

Platinum Member

Lifer

Golden Member

Golden Member

Platinum Member

Golden Member

Diamond Member

Golden Member

Diamond Member

Golden Member

Diamond Member