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)	Arrow Lake Refresh (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Desktop Only	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2025 ?	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	8P + 32E	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.)

 

[SiliconFly]

This is exactly what I mentioned a while back. This test shows MTL running with both the GPU & CPU tile off most of the time. And that makes one hell of a difference...

Beats AMD 7840 by more than 2X in power efficiency.

 

[Hitman928]

View attachment 90408

This is exactly what I mentioned a while back. This test shows MTL running with both the GPU & CPU tile off most of the time. And that makes one hell of a difference...

Beats AMD 7840 by more than 2X in power efficiency.

When doing nothing with the computer. . .

I mean, it's kind of cool to see it working, but no one will see a benefit like this under actual use case scenarios.

 

[dullard]

When doing nothing with the computer. . .

I mean, it's kind of cool to see it working, but no one will see a benefit like this under actual use case scenarios.

No one has a computer idle or in light load ever? You really think that?

I only know the experience of people that I see using computers. But they tend to be idle or low usage almost all the time. Typing a document, emailing, web-browsing, reading, playing music, cooking, conference calls, watching videos, etc. That is what I personally see most users doing most of the time.

 

[SiliconFly]

When doing nothing with the computer. . .

I mean, it's kind of cool to see it working, but no one will see a benefit like this under actual use case scenarios.

True. But the real world scenario is somewhere in between. Actually, people tend to idle a lot.

No one has a computer idle ever? You really think that?

I only know the experience of people that I see using computers. But they tend to be idle or low usage almost all the time.

I guess in a typical 8 hour work day, the systems tend to idle a lot more than we can imagine. Especially considering workloads like browsers & office/productivity apps in which their main and worker threads usually tend to spend most of the time in idle states.

 

[Khato]

No one has a computer idle ever? You really think that?

I only know the experience of people that I see using computers. But they tend to be idle or low usage almost all the time.

Indeed. Have to keep in mind that from the perspective of a computer, even 1 second of nothing happening is quite a bit of 'idle' time. I know my usual web browsing has plenty of longer pauses than that while reading through a page. And if normal basic interaction (scrolling, typing, etc) all stays on the low power e-cores as it should, then the periods between needing to make use of the primary compute cores can be quite a bit longer.

 

[TESKATLIPOKA]

View attachment 90408

This is exactly what I mentioned a while back. This test shows MTL running with both the GPU & CPU tile off most of the time. And that makes one hell of a difference...

Beats AMD 7840 by more than 2X in power efficiency.

It doesn't have >2x better efficiency than Phoenix per se. Just those tested PHX laptops are not that great in this regard. This is the problem with laptops in general, that they have big differences in power consumption under light loads despite comparable specs.

Here is ThinkPad T14s G4 with R7 7840U and 57Wh battery.Link

If I normalized It to 75Wh I would get ~17.5 hours.
MTL would still win, but not with that big of a margin.

 

[Saylick]

View attachment 90408

This is exactly what I mentioned a while back. This test shows MTL running with both the GPU & CPU tile off most of the time. And that makes one hell of a difference...

Beats AMD 7840 by more than 2X in power efficiency.

No one has a computer idle or in light load ever? You really think that?

I only know the experience of people that I see using computers. But they tend to be idle or low usage almost all the time. Typing a document, emailing, web-browsing, reading, playing music, cooking, conference calls, watching videos, etc. That is what I personally see most users doing most of the time.

That comparison was just refreshing Chrome every 15 seconds, which I'm not sure is equivalent to typing a document, emailing, conference calls, or watching videos. I suspect typing a Word document will fire up the compute tile.

Using a smartphone analogy, I personally see that comparison as something closer to standby battery life, which while useful is not something most users care about since screen-on-time is more important, and screen-on-time is more analogous to tasks that would fire up the compute tile (if this makes sense).

 

[SiliconFly]

Indeed. Have to keep in mind that from the perspective of a computer, even 1 second of nothing happening is quite a bit of 'idle' time. I know my usual web browsing has plenty of longer pauses than that while reading through a page. And if normal basic interaction (scrolling, typing, etc) all stays on the low power e-cores as it should, then the periods between needing to make use of the primary compute cores can be quite a bit longer.

I think even while typing or even scrolling in a browser, the LP E cores should be able to handle it comfortably as the draw call isn't handled by the CPU tile.

That comparison was just refreshing Chrome every 15 seconds, which I'm not sure is equivalent to typing a document, emailing, conference calls, or watching videos. I suspect typing a Word document will fire up the compute tile.

Using a smartphone analogy, I personally see that comparison as something closer to standby battery life, which while useful is not something most users care about since screen-on-time is more important, and screen-on-time is more analogous to tasks that would fire up the compute tile (if this makes sense).

Typing a word document and emailing shouldn't technically wake up the CPU tile. It's basically a bunch of draw calls. Conference calls might, but for how long I'm not sure. Watching videos is power consuming, even with the CPU tile off I think.

 

[Abwx]

No one has a computer idle or in light load ever? You really think that?

I only know the experience of people that I see using computers. But they tend to be idle or low usage almost all the time. Typing a document, emailing, web-browsing, reading, playing music, cooking, conference calls, watching videos, etc. That is what I personally see most users doing most of the time.

Then it s not connected to the internet, because the slightest activity will wake up the main cores, here they present a corner case where a same page is refreshed in Chrome, dunno if that could be called a normal usage, the ones you are talking about require the other cores to wake up...

 

[Hitman928]

No one has a computer idle or in light load ever? You really think that?

I only know the experience of people that I see using computers. But they tend to be idle or low usage almost all the time. Typing a document, emailing, web-browsing, reading, playing music, cooking, conference calls, watching videos, etc. That is what I personally see most users doing most of the time.

The other web browsing test we have which isn't just, ctrl+r the same cached page every 15 seconds, did not seem to show a big improvement from using the LPE cores. Any time you do something that actually interacts with the computer, the process should get kicked to the P-cores because you want high responsiveness. Hopefully someone can thoroughly test this, but I just highly doubt the LPE cores will be used much outside of long idle situations. The only battery life test we have seen that shows MTL with a large lead over the competition is that situation. Even if you are just scrolling through a website and then going to a new one, opening word in between, etc., I don't see the LPE cores coming into play.

Edit: Even Intel does not claim large improvements during web browsing. Clearly this is not where the LPE cores are intended to come into play. Intel is claiming they only come into play when you are not interacting with the computer (i.e., idle and video playback situations). Unfortunately for Intel, the video playback tests we have from 3rd parties also do not show the kind of improvements they claim.

 

[Saylick]

The other web browsing test we have which isn't just, ctrl+r the same cached page every 15 seconds, did not seem to show a big improvement from using the LPE cores. Any time you do something that actually interacts with the computer, the process should get kicked to the P-cores because you want high responsiveness. Hopefully someone can thoroughly test this, but I just highly doubt the LPE cores will be used much outside of long idle situations. The only battery life test we have seen that shows MTL with a large lead over the competition is that situation. Even if you are just scrolling through a website and then going to a new one, opening word in between, etc., I don't see the LPE cores coming into play.

Edit: Even Intel does not claim large improvements during web browsing. Clearly this is not where the LPE cores are intended to come into play. Intel is claiming they only come into play when you are not interacting with the computer (i.e., idle and video playback situations). Unfortunately for Intel, the video playback tests we have from 3rd parties also do not show the kind of improvements they claim.

Yeah, this is the crux of MTL's efficiency gains: does your intended workload fire up the compute tile? If so, then don't expect a big jump in battery life over RPL. If not, then you get a lot more battery life.

I personally think that having better battery life when you're doing something is far more helpful than having better battery life when you're not doing much.

 

[dullard]

I hate that AMD is being lazy coz Intel is so slow.

Now there is a troll statement if I ever saw one.

 

[jpiniero]

Note that chart is using the 165H and not the 155H.

 

[Hitman928]

Note that chart is using the 165H and not the 155H.

Shouldn't matter when the argument is having the LPE cores with the compute tile powered down is the benefit.

 

[SiliconFly]

It doesn't have >2x better efficiency than Phoenix per se. Just those tested PHX laptops are not that great in this regard. This is the problem with laptops in general, that they have big differences in power consumption under light loads despite comparable specs.

Here is ThinkPad T14s G4 with R7 7840U and 57Wh battery.Link
View attachment 90410
If I normalized It to 75Wh I would get ~17.5 hours.
MTL would still win, but not with that big of a margin.

Yep. It's just a one off result and more than 2X maybe a bit off. But I just wanted to hilight the scenario where the LP E cores actually shine. It's just that many don't talk about the importance of the LP E cores.

In a typical home and/or office scenario where people are just using simple browsers and/or office apps, I strongly believe, it's the LP E cores that'll be active more than 95% of the time during the day. This may hold true for people in sales and marketing, but not the design or development teams. May hold true for doctors and lawyers, but won't for gamers or enthusiasts. Just saying.

 

[dullard]

Edit: Even Intel does not claim large improvements during web browsing

Your image shows 7% better than the competition. Maybe that isn't a large improvement to you, but it is a heck of a lot better than where Intel was.

For heavy CPU use cases, you'll have to wait for Arrow Lake next year for Intel to be more competitive.

 

[controlflow]

View attachment 90398

Looking good for casual use.

What is the source of this?

 

[SiliconFly]

Maybe that isn't a large improvement to you, but it is a heck of a lot better than where Intel was.

So true.

My 13th gen laptop is enough to heat my entire home when it's under full load.

And revisiting my previous observation, now that too many different benchmark results are there. Performance benchmarks are pretty much straight forward. And they should be able to cover most of the real world scenarios (for MTL and other CPUs as well.)

But power efficiency benchmarks are a different beast. All these time they were pretty much focusing on stress tests and they won't represent MTL in real world scenarios anymore due to MTL new architecture. We're are gonna need either newer tests or updated benchmark suites.

 

[SiliconFly]

Intel's mature high clocking nodes really have a tendency to screw them over, don't they? It's what, third time when their new lineups lose in ST compared to predecessors (Haswell -> Broadwell, RPL -> MTL, Whiskey Lake -> Ice Lake)?

MTL's Intel 4 is not exactly a top of the line node if you ask me. It brings in a minor bump in efficiency but just not enough. And it uses HP cells. So, MTL's CPU tile neither has world class performance nor the efficiency we need. The only saving grace for MTL is it's LP E cores, without which it's efficiency will sink like a rock.

ARL should rectify this issue with 20A by the end of Q3 2024 (hopefully if Intel executes as expected).

 

[Hitman928]

Your image shows 7% better than the competition. Maybe that isn't a large improvement to you, but it is a heck of a lot better than where Intel was.

For heavy CPU use cases, you'll have to wait for Arrow Lake next year for Intel to be more competitive.

That is Intel's own numbers which I don't put much trust in (see their claims during the live yesterday vs. tests today). The point of sharing the graph from Intel, however, is that even Intel isn't claiming large gains from the LPE cores during interactive use of the computer. If the LPE cores were able to take over for web browsing, you should see a much larger advantage during this test, something more like the busy idle or video playback test, but you don't, you see the same relative performance as a Teams meeting which is for sure going to be running on the compute tile.

 

[controlflow]

When doing nothing with the computer. . .

I mean, it's kind of cool to see it working, but no one will see a benefit like this under actual use case scenarios.

I agree with your point but it looks reasonable under load too. No it isn't able to compete with Apple but it at least dramatically improves on mobile RPL. I think MTL is alright but I wanted to see better PPW given the new node. The new iGPU is a very welcome improvement but I'm not too impressed with the performance of the CPU cores. Intel needs to replace RWC with something much better.

 

[Geddagod]

MTL's Intel 4 is not exactly a top of the line node if you ask me. It brings in a minor bump in efficiency but just not enough. And it uses HP cells. So, MTL's CPU tile neither has world class performance nor the efficiency we need. The only saving grace for MTL is it's LP E cores, without which it's efficiency will sink like a rock.

ARL should rectify this issue with 20A by the end of Q3 2024 (hopefully if Intel executes as expected).

Don't think this is on the node team

 

[H433x0n]

No one has a computer idle or in light load ever? You really think that?

No, they do image rendering and cinebench specifically below 35W.

MTL's Intel 4 is not exactly a top of the line node if you ask me. It brings in a minor bump in efficiency but just not enough.

I don't think it's the node. I think it's a combination of the Redwood Cove core being disappointing and the most recent iteration of Intel 7 being much better than it's given credit for. For the car guys, the Raptor & Redwood Cove cores remind me of a big turbo evo, where it just takes forever to 'spool' up.

 

[Hitman928]

I agree with your point but it looks reasonable under load too. No it isn't able to compete with Apple but it at least dramatically improves on mobile RPL. I think MTL is alright but I wanted to see better PPW given the new node. The new iGPU is a very welcome improvement but I'm not too impressed with the performance of the CPU cores. Intel needs to replace RWC with something much better.

View attachment 90414
View attachment 90415
View attachment 90416

Yes, battery life is improved, though to what degree is hard to determine. Other tests show the improvement to be minimal at best when comparing the same laptop with the MTL update. Battery life tests are really a system test and vary wildly between models and settings (even if normalizing for the battery) which is why getting power settings, monitor resolution, brightness, etc. are very important for understanding how they actually compare. Notebookcheck usually is pretty good about outlining what settings they are using and trying to normalize as much as possible between models, so I am looking forward to their results but it sounds like they might not get the chance to run those tests until January.

 

[Abwx]

I agree with your point but it looks reasonable under load too. No it isn't able to compete with Apple but it at least dramatically improves on mobile RPL. I think MTL is alright but I wanted to see better PPW given the new node. The new iGPU is a very welcome improvement but I'm not too impressed with the performance of the CPU cores. Intel needs to replace RWC with something much better.


View attachment 90416

So basically in the heavy loading "test" the 7840U laptop consume 50W, wich is quite a lot for a U APU wich is suposed to go up to 30W, and the MTL laptop 35W, if that s not made up numbers dunno what it is...

And of course no idea of the quantity of work processed...