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.)

 

[Henry swagger]

Mods clean this thread.. so much amd talk 🤨

 

[Hitman928]

Regarding the display, OLED should use less power in theory. If it isn’t then companies are likely cheaping out. Good Adaptive Sync OLED displays should be capable of at least 1-120hz refresh.

No, OLEDs use more power. The only time you may save power is when the screen is largely very dark due to there being no backlight and so the vast majority of pixels are off whereas with other types of screen, the backlight will still be on (this changes with LED array panels with local dimming to whatever extent they can do local dimming). For normal usage though, OLEDs use significantly more power than other modern displays.

 

[moinmoin]

We need a similar solution for GPUs as well and a way to let Windows see “the happy path” to power savings.

What we really need is Windows to be optimized for such a setup. Many of the Windows processes could be run on these cores without compromising performance. Even if the CPU tile is active this could save some power.

In the last couple years Microsoft has repeatedly shown that it doesn't really care how well Windows' scheduler works. This works to the detriment of any true LP core that could work to save power as with a scheduler not cooperating you get the worst of all worlds, slow cores visibly affecting application performance as the scheduler ends up (ab)using the LP cores for normal processes anyway, and for better power saving the wrong non-optimized applications ends up being moved to the LP cores defeating the whole setup. So I don't have hope that this improves with Microsoft's help, and thus the approach is doomed.

 

[adroc_thurston]

In the last couple years Microsoft has repeatedly shown that it doesn't really care how well Windows' scheduler works.

But it works just fine.
Modern workloads are just perf-hungry and thus anything with a QoS priority gets moved onto the bigs.

 

[dullard]

But it works just fine.
Modern workloads are just perf-hungry and thus anything with a QoS priority gets moved onto the bigs.

I realize it was a joke, but at least use the proper terms.

Thread Affinity: software states which type of core to run on (big, little, etc). Windows doesn't get to choose if a thread affinity is stated by software. This is the correct long-term solution as the software makers should know best.

Thread Priority: software states how high up on the list the thread should be when there are too few resources to run all threads at the same time.

QoS priority is for networking.

 

[moinmoin]

But it works just fine.
Modern workloads are just perf-hungry and thus anything with a QoS priority gets moved onto the bigs.

Yeah, Microsoft doesn't even bother to take responsibility for the Windows ecosystem anymore. It's little more than a PR vehicle to push secondary services. For LP cores the simplest approach would be to lock them away from normal processes and only allow dedicated optimized ones that won't ever spill over to full E- or P-cores. But no, let's the usual mess keep festering, after all it works just fine.

And people keep wondering what competitive advantage Apple could possibly have.

This is the correct long-term solution as the software makers should know best.

I disagree. Especially the last couple of years with its many different odd setups, MTL LPe cores being only one of them, showed that no software can really be prepared in advance for such without knowing them. It's the job of the scheduler to make everything work well in the end even if software is badly prepared or even not at all.

 

[maddie]

Yeah, Microsoft doesn't even bother to take responsibility for the Windows ecosystem anymore. It's little more than a PR vehicle to push secondary services. For LP cores the simplest approach would be to lock them away from normal processes and only allow dedicated optimized ones that won't ever spill over to full E- or P-cores. But no, let's the usual mess keep festering, after all it works just fine.

And people keep wondering what competitive advantage Apple could possibly have.


I disagree. Especially the last couple of years with its many different odd setups, MTL LPe cores being only one of them, showed that no software can really be prepared in advance for such without knowing them. It's the job of the scheduler to make everything work well in the end even if software is badly prepared or even not at all.

Just a thought, could the AI engines help with this, learning over time?

 

[dullard]

I disagree. Especially the last couple of years with its many different odd setups, MTL LPe cores being only one of them, showed that no software can really be prepared in advance for such without knowing them. It's the job of the scheduler to make everything work well in the end even if software is badly prepared or even not at all.

You are correct in the fact that no software can predict future CPUs. However, you are missing the point.

Take a virus scan application for example. When someone opens up a virus scanner and says to scan a drive now, that should be on performance cores--whatever form the performance cores take now or in the future. No user wants to wait an eon after they put in a drive and ask for it to be scanned. They most likely want it scanned now and scanned fast.

But, when the virus scanner just runs in the background all day long scanning misc. items, that should be on efficient cores and not on the performance cores. Yes, in the future newer efficient cores can and will come along that the software does not know about. The scheduler will have to determine if it goes on the E or LPe cores. But, the software has already done the main task of avoiding the P cores so the user's other software can run at full speed unaffected (excluding the few software programs specifically coded to use all cores of course). A later software revision can come along and specify LPe cores only (if the developer so chooses).

 

[DavidC1]

Idle minimum power (display set to minimum brightness, all wireless off, and power plan in Windows set to Energy Saving) = 7.1 W

7.1W idle power? Have you been living under a rock? If you were an average joe that only knows how to turn on/off computers, I'd understand but you are posting in AT forums.

https://www.notebookcheck.net/Acer-Swift-Edge-SFE16-review-A-lightweight-16-inch-laptop-with-AMD-power-and-an-OLED-display.726861.0.html

AC power: Idle minimum, 8.6W
Reader/Idle battery life: 653 mins, 54WHr. Or 4.96W

My Kaby-Y Yoga 11S uses less than 4W with the screen on while playing Youtube video. 8.6W isn't just bad, it's a disaster. 4.96W is a disaster. It should be under 3W screen-on idle. Dell XPS Icelake idled at 2.5W!

NBC's AC idle power results are near totally irrelevant for measuring battery power. In their excuse, you can only measure power properly if you open up the laptop and probe it, but that's difficult.

Instead for laptops you have a more relevant figure that is much easier and convenient to test, called battery life. Your wireless router at home is on 24/7 and people couldn't care less about that power, so 2-3W is irrelevant for AC, thus manufacturers don't care either.

However, under battery you do.

Unfortunately Meteorlake hasn't convinced me thoroughly that it caught up to AMD in perf/watt and battery life.

(Reminder: perf/watt and battery life is NOT the same. You could have wonderful perf/watt and terrible battery life and vice versa)

The seemingly endless miracle BIOS patches have ended up in handful of test systems and it just stopped. Oh well, wait for Lunarlake I guess.

I liked Intel too and is more confident about the future but let's get our facts straight first, eh?

@mikk You did not address that using N2 for Novalake is bad, especially if they have 18A for Pantherlake. There really is little reason.

 

[Khato]

You've convinced me, I agree now with your way of thinking about this. That's why, after searching near and far, I found the best possible CPU, the 7840u in a Lenovo laptop. It offers over 13 hours of browsing battery life on only a 57 Whr battery. Honorable mention would be the HP laptop with a 7840u. It actually is even more efficient but the battery life isn't as long due to a lower capacity battery so it gets knocked down a peg. You're going to need glasses because this one shines so bright! No need to check how they are configured, clearly when Lenovo or HP get a hold of a CPU they know how to do it right and Zen4 can really shine ahead of any competition!

https://www.notebookcheck.net/Lenovo-ThinkPad-T14s-G4-review-Business-laptop-is-better-with-AMD-Zen4.763581.0.html

And RPL-U can offer a minute shy of 14 hours of web browsing battery life on the same 57Whr battery - https://www.notebookcheck.net/Lenov...ing-expensive-business-flagship.728359.0.html

It's a great example of how meticulous implementation of the remainder of the laptop can yield greater gains in light workload battery life than superior CPU efficiency.

 

[DavidC1]

And RPL-U can offer a minute shy of 14 hours of web browsing battery life on the same 57Whr battery - https://www.notebookcheck.net/Lenov...ing-expensive-business-flagship.728359.0.html

Raptorlake-U can be efficient, so can Alderlake-U. It's just that there's a much greater variation where some fail totally. And it seems unlike previous Intel platforms, the battery life is a more direct tradeoff from performance, when it shouldn't matter.

They started regressing in Icelake, further in Tigerlake, but really came to fruition in Alder and Raptor. I had hoped that Meteorlake's LP-E cores would be better, but with Intel themselves admitting a mere 0.15W power savings, at best it's going back to some RPL-U levels.

Remember, it was in Comet/Whiskeylake days when there was a roughly 50% gap in battery life/whr compared to ARM. Considering Intel exerts enormous engineering resources into the PC ecosystem, problems with Meteorlake can and will be blamed at Intel because they are largely responsible. Now, I reckon some failures are because MTL is such a delayed product and I have a rule of thumb that delay = worse product.

With Intel less talk the better. They were very quiet previous to Core 2. The Atom team does not blabber either. Tiger was talked about endlessly with three separate events talking about the same thing. It still got clobbered by Ryzen so still a disappointment. Meteorlake had big slide days too.

 

[Khato]

Raptorlake-U can be efficient, so can Alderlake-U. It's just that there's a much greater variation where some fail totally. And it seems unlike previous Intel platforms, the battery life is a more direct tradeoff from performance, when it shouldn't matter.

Indeed, it's easy for OEMs to offer up laptops with horrible battery life. But that aspect is not the fault of either AMD or Intel and affects both equally.

It is fair to note that the Intel U series trades multi-thread performance for better battery life, a nice little advantage for the target market.

 

[eek2121]

No, OLEDs use more power. The only time you may save power is when the screen is largely very dark due to there being no backlight and so the vast majority of pixels are off whereas with other types of screen, the backlight will still be on (this changes with LED array panels with local dimming to whatever extent they can do local dimming). For normal usage though, OLEDs use significantly more power than other modern displays.

That is simply not true except on the brightest of backgrounds. LED monitors use a backlight, and regardless of brightness, always have that backlight at a fixed setting. OLED pixels are self emitting. Unless you are using white/bright colors all day, OLED will use less power. To put another way, show a dark grey or black blank image full screen on your monitor. Unless brightness or refresh rate don't match, OLED will consume less power.

For movies? OLED wins in both power consumption and image quality.

Don't believe me? look it up with a reputable source.

 

[adroc_thurston]

That is simply not true except on the brightest of backgrounds.

It is true, self-emissive panels are PITA to drive.

 

[Khato]

It's pretty easy to look up figures for OLED lumens/W and compare to LED - it's at least a 2.5x advantage for LED. The adoption of quantum dot color generation rather than filters was a major boon for LED based display efficiency.

 

[hemedans]

That is simply not true except on the brightest of backgrounds. LED monitors use a backlight, and regardless of brightness, always have that backlight at a fixed setting. OLED pixels are self emitting. Unless you are using white/bright colors all day, OLED will use less power. To put another way, show a dark grey or black blank image full screen on your monitor. Unless brightness or refresh rate don't match, OLED will consume less power.

For movies? OLED wins in both power consumption and image quality.

Don't believe me? look it up with a reputable source.

It depend with size too, OLED TV use more power than LED TV.

From rtngs

 

[Hitman928]

It depend with size too, OLED TV use more power than LED TV.

From rtngs

Even with smaller screens it holds true though the absolute difference isn't so drastic.

 

[Asterox]

Ok, maybe HT and AVX512 is too expensive.

Intel Arrow Lake-S CPU spotted with 24 threads, no Hyper Threading and AVX512 support - VideoCardz.com

Intel Arrow Lake-S with no hyper threading More evidence surfaces claiming that Intel’s next-generation client CPU series will not support hyper threading. Hyper Threading, is a simultaneous multithreading technology that allows physical cores in Intel CPUs to perform two tasks at the same time...

videocardz.com

 

[SiliconFly]

Intel does it again. 15th gen Raptor Lake Refresh Refresh is now confirmed.

https://www.tweaktown.com/news/9591...cessors-leaving-arrow-for-high-end/index.html

This makes it 10+++++

 

[Oneto]

Intel does it again. 15th gen Raptor Lake Refresh Refresh is now confirmed.

https://www.tweaktown.com/news/9591...cessors-leaving-arrow-for-high-end/index.html

This makes it 10+++++

Having read the article, no, no it's not confirmed. They did say that there would be a "budget" LGA1700 chip though

 

[DavidC1]

That is simply not true except on the brightest of backgrounds. LED monitors use a backlight, and regardless of brightness, always have that backlight at a fixed setting. OLED pixels are self emitting.

Self emitting or not it does nothing for reducing power. Many average people expected power reduction based on not requiring backlight.

Reality, OLED screens on laptops are generally blamed for poor battery life, because that's how much more power it uses. What looks like on the high level and what it actually is can be very different.

 

[FlameTail]

It is possible for OLED laptops to improve battery life greatly from where they are now.

OLED tech is improving at a rapid pace, much faster than LCDs perhaps. Samsung Display for instance, releases a new OLED material set almost every year. The Galaxy S24 series just debuted their latest M13 AMOLED technology. Each new technology brings much improved power efficiency. Unlike smartphones, it seems laptops are much slower in adopting these new technologies.

Google Pixel 8 Pro display deep dive: Brain finally meets brawn

What happens when Google uses premium components?

www.xda-developers.com

Aside from OLED materials sets, another thing that can help improve power efficiency in OLEDs is the use of LTPO/oxide TFT + VRR. This has already been implemented in smartphones with great success. LTPO/oxide TFTs are inherently more power efficient than LTPS in normal OLEDs, and it also enables dynamic VRR. This means the display refresh rate can go dynamically adjust to suit the content on the screen, going as low as 1 Hz when displaying static content.

Aside from hardware, software features can also help improve power efficiency. Look at Asus' Lumina OLED. They have implemented some special software techniques, which reduces OLED power consumption.

 

[ikjadoon]

That is simply not true except on the brightest of backgrounds. LED monitors use a backlight, and regardless of brightness, always have that backlight at a fixed setting. OLED pixels are self emitting. Unless you are using white/bright colors all day, OLED will use less power. To put another way, show a dark grey or black blank image full screen on your monitor. Unless brightness or refresh rate don't match, OLED will consume less power.

For movies? OLED wins in both power consumption and image quality.

Don't believe me? look it up with a reputable source.

Unfortunately for laptop OLEDs, much of the web uses a light background (i.e., high APL). In fact, even Notebookcheck's own light website consumed virtually the same power as a 100% white background.

Because of this, OLED is more likely to carry a battery life penalty for browsing based battery life tests. Notebookcheck ran the numbers in a 2019-era OLED:




They have a somewhat-comparable comparison (4K60 OLED vs 1080p240 IPS): the OLED lost the battery life comparison by ~10%.

//

In another comparison, at minimum brightnesses Notebookcheck's own site (via screenshot) consumed more power than an all white background on a 2019-era OLED and only slightly less at maximum brightness.



The bane of any OLED laptop owner is a white website with higher brightnesses.

//

More to the topic, this review summary claimed "great battery life", but it's hardly one hour longer than the average. This is the HP Spectre x360 14 2024.

//

This is the Zenbook 14 OLED w/ the U7 155H.



Some more 1T power numbers (I assume CPU package power). I thought the i7-13700H at 18W was very competitive.

//

Battery life here is much improved with the PCMark10 test. This is great to see, though some is due to the test including offline media playback and MTL just runs laps around RPL there.



Unfortunately, the AI benchmarking was a mess and I don't blame KitGuru: turned out ASUS used the wrong API framework (good luck to consumers catching that) until a Windows update somehow reverted it, Windows tried to downgrade the GPU drivers, and the UL test seems tough to connect to real usage.

 

[SiliconFly]

View attachment 93255

Looks like MTL battery life has had an phenomenal improvement! What has changed in the last 2 months? New bios patches/pcode updates/windows drivers/os patches/etc? Or is it something else entirely like a new stepping or larger battery, etc?

 

[dullard]

Looks like MTL battery life has had an phenomenal improvement! What has changed in the last 2 months? New bios patches/pcode updates/windows drivers/os patches/etc? Or is it something else entirely like a new stepping or larger battery, etc?

ikjadoon specifically stated that it was because of offline media playback being included in the test. I assume that means that the laptop doesn't have to actively download the media while playing it. Just a guess here, but does that mean that only the LPE cores and media engine is running and the rest of the CPU cores can shut down in this case?