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.)

Abwx · Oct 10, 2024

dullard said:
Intel has discrete graphics cards that it can work with. Practically no one uses them, but technically it is possible. And there are some people (like me) that actually use the iGPU on desktop chips.

For most gamers though, with discrete non-Intel graphics cards, it is just a Windows scheduler tweak optimized for specific game needs.

Ok, thank you.
Also you said that at 300W gaming perf should be 20% better than at 250W, that s just impossible, perfs doesnt scale linearly with power, moreover in games, at best in CB it will perform 5-6% better from 250 to 300W, you misinterpreted the slide you are refering to.

Hulk · Oct 10, 2024

Wolverine2349 said:
If that's true that is very disappointing and sad.

I really wanted an on par CPU gaming wise with 7800X3D or at least Raptor Lake that had more than 8 cores and a good in built hardware scheduler and no dual CCD infinity fabric crossing crap.

Maybe raising power limits similar to Raptor Lake levels will make it better.

The selling point seems better power efficiency.

I wonder if it can perform belter than Raptor Lake and on par with 7800X3D if power limits are removed, but this time with strong platinum stability and no degradation problems. Well power limits removed to an extent. Obviously do not want 400W or even over 300W like Raptor Lake, but the 250W or maybe up to 300W CPU only.

How bottlenecked are games by CPU's these days? Are we talking about frame rates in the 50's or 60fps or in the hundreds of fps?

DavidC1 · Oct 10, 2024

GTracing said:
Yup, and the compiler is different, and we don't know what clock speed Intel ran the "iso-clock" benchmarks at, and the benchmark is "simulated" (not run on the actual hardware?), and they claim a 10% margin for error,

10% margin means 10% of the 32% and 72%, so 28-35%, and 65-79%, not 22-42% and 62-82% as some think. Also it's plus or minus. That's still an awesome gain.

GTracing said:
But I do think that it's within the realm of possibility that skymont gets within spitting distance of the p core in IPC.

Yea with the revised down 9% for Lion Cove it's hard to determine whether Lion Cove is faster than Skymont ISO clock in Integer at all.

TESKATLIPOKA · Oct 10, 2024

dullard said:
Intel's slides show that gaming performance is linear with power for the 285K (same performance per watt, at least when going from 125 W to 250 W). So, if that slide is correct, then 250 W power is double the gaming performance as 125 W. If that holds at higher power levels, then 300 W would have about a 20% gaming boost over 250 W.
View attachment 109120

If you though a bit about It then you would have realized that graph is a total nonsense or doesn't mean only the CPU but the whole system!

TPU compared 14900K at different TDPs and the average increase was only 78% and best case of 128% was in CP2077. Wow right?
Not really, because that was only HD resolution and 35W vs 253W TDP(7.2x), real power consumption was 16W vs 144W(9x)!
Performance difference between 125W vs 253W was not even 3% while consuming 38W more during gaming. Link

@9949asd Of course, because games don't use that many cores or at full power. That's why perf. difference between 125-250W should be small.

Hulk · Oct 10, 2024

Hitman928 said:
Here's a better slide for efficiency comparison (kind of as there are no hard numbers on the graph and it only shows Cinebench). Anyone have a similar chart for the 9950x and 14900k to compare relative curves?

View attachment 109123

Ah yes, here we go with the charts that if I handed in when I was an engineering student I would have ended up with a failing grade.

No description on the y-axis.
No idea of the applications tested.

TESKATLIPOKA · Oct 10, 2024

Hulk said:
Ah yes, here we go with the charts that if I handed in when I was an engineering student I would have ended up with a failing grade.

No description on the y-axis.
No idea of the applications tested.

At least we know Intel is employing engineering dropouts.

9949asd · Oct 10, 2024

TESKATLIPOKA said:
If you though a bit about It then you would have realized that graph is a total nonsense or doesn't mean only the CPU but the whole system!

TPU compared 14900K at different TDPs and the average increase was only 78% and best case of 128% was in CP2077. Wow right?
Not really, because that was 35W vs 253W! Difference between 125W vs 253W was not even 3%. Link

Yea, the 253w is power limit. The gaming power only around 80w-120w

9949asd · Oct 10, 2024

Well, look at the price of z890+285k+cudimm and the gaming performance 😂
I think better just save the money for the upcoming rtx 5090🤣

igor_kavinski · Oct 10, 2024

Hulk said:
How bottlenecked are games by CPU's these days? Are we talking about frame rates in the 50's or 60fps or in the hundreds of fps?

I have a feeling he thinks that more P-cores are the perfect weapon against the Ryzen X3D chips...

DavidC1 · Oct 10, 2024

Lol, so they kinda sandbagged the Skymont performance by saying 32% for the E cores and 9% for the P cores, when for Skymont it was just for Int, and for the P cores it was combined Int/FP.

I don't know why the gains got lowered from the initial 14%. Considering Lion Cove adds additional FP units too, it's very disappointing. That means Integer gains are possibly lower than 9%.

The Youtube Arrowlake video has the Intel guy basically ignoring Lion Cove and just praising Skymont. This is basically telling us what their future arch is.

dullard · Oct 10, 2024

Hulk said:
Ah yes, here we go with the charts that if I handed in when I was an engineering student I would have ended up with a failing grade.

No description on the y-axis.
No idea of the applications tested.

It is a poorly made chart, but not to the extent that you are mentioning.

(1) There is a y-axis description. Not a good description, but there is at least a description.

(2) There is a description of the application listed in that very same PDF file:

Hitman928 · Oct 10, 2024

DavidC1 said:
Lol, so they kinda sandbagged the Skymont performance by saying 32% for the E cores and 9% for the P cores, when for Skymont it was just for Int, and for the P cores it was combined Int/FP.

I don't know why the gains got lowered from the initial 14%. Considering Lion Cove adds additional FP units too, it's very disappointing. That means Integer gains are possibly lower than 9%.

The Youtube Arrowlake video has the Intel guy basically ignoring Lion Cove and just praising Skymont. This is basically telling us what their future arch is.

The initial 14% was for LNL over MTL P-cores. Multiple people were pointing out that this probably meant there would be a smaller increase for ARL vs RPL, but they largely got ignored/shut down.

gdansk · Oct 10, 2024

Hitman928 said:
The initial 14% was for LNL over MTL P-cores. Multiple people were pointing out that this probably meant there would be a smaller increase for ARL vs RPL, but they largely got ignored/shut down.

Not necessarily ignored. For example, I saw these comments but thought the other changes to Lion Cove from LNL to ARL would end up with some increase. And that'd it'd probably be pretty close to 14% overall. But it doesn't seem to be the case.

alcoholbob · Oct 10, 2024

Hulk said:
How bottlenecked are games by CPU's these days? Are we talking about frame rates in the 50's or 60fps or in the hundreds of fps?

I think generally the framerates are "okay" for CPU bottlenecked games like BG3 for instance, unless you are obsessively trying to get games to max out at your monitors high hz refresh rate.

The bigger problems are generally 1% lows, but then it's not just the CPU but also a combination of other bottlenecks, like storage speed (when streaming textures).

DavidC1 · Oct 10, 2024

Hitman928 said:
The initial 14% was for LNL over MTL P-cores. Multiple people were pointing out that this probably meant there would be a smaller increase for ARL vs RPL, but they largely got ignored/shut down.

Yes, but the E core portion is even better than expected now. From 38%/68% over LPE cores in Meteorlake to 32%/72% over cores in Raptorlake Refresh.

This is despite the rumor that Arrowlake might have 1.1GHz reduction in ring frequency.

MoistOintment · Oct 10, 2024

Hitman928 said:
They leave them at JEDEC? Not even enable EXPO/XMP? It's literally a single click on the first page of most BIOSes. Personally, I don't know anyone who built their own PC that doesn't at least enable them. Almost every pre-built gaming PC will come with them enabled as well.

There are millions of PC gamers and hundreds of us on forums. It's pretty typical for someone to watch some simple guide on how to connect the parts together, install Windows, and go from there.

I'm sure most, if not all here, are gonna enable EXPO/XMP, but it's definitely typical for most users (and prebuilts) to run JEDEC.

I think Intel and AMD should always present their performance numbers at JEDEC. If the manufacturer feels confident enough to present their performance data at above JEDEC speeds, then why not make those speeds the officially supported speed? But I would prefer independent reviewers to enable reasonable EXPO/XMP.

Hitman928 · Oct 10, 2024

MoistOintment said:
There are millions of PC gamers and hundreds of us on forums. It's pretty typical for someone to watch some simple guide on how to connect the parts together, install Windows, and go from there.

I'm sure most, if not all here, are gonna enable EXPO/XMP, but it's definitely typical for most users (and prebuilts) to run JEDEC.

I think Intel and AMD should always present their performance numbers at JEDEC. If the manufacturer feels confident enough to present their performance data at above JEDEC speeds, then why not make those speeds the officially supported speed? But I would prefer independent reviewers to enable reasonable EXPO/XMP.

I agree with manufacturers presenting numbers with officially supported speeds, I'm just saying that the vast majority of gamers won't be running that speed and so the real meaningful results are with easy to achieve non-JEDEC memory speeds.

KompuKare · Oct 10, 2024

DavidC1 said:
Lol, so they kinda sandbagged the Skymont performance by saying 32% for the E cores and 9% for the P cores, when for Skymont it was just for Int, and for the P cores it was combined Int/FP.

I don't know why the gains got lowered from the initial 14%. Considering Lion Cove adds additional FP units too, it's very disappointing. That means Integer gains are possibly lower than 9%.

The Youtube Arrowlake video has the Intel guy basically ignoring Lion Cove and just praising Skymont. This is basically telling us what their future arch is.

Might the missing P cores gains have been a relatively late things due changes made after the RPL degradation failures became public?

Mainly the rumoured reduced ring bus frequency but maybe other things had to be dialed down too?

This not being Intel's on process might make them more cautious too as any accelerated aging tests might be harder.

coercitiv · Oct 10, 2024

I would argue the gamers who are willing to check on early manufacturer presentations (directly or through the press) are much more likely to use XMP memory. The people with JEDEC memory in their rigs may not even know what Core Ultra is until they decide to buy a new system. Whatever Intel and AMD communicate during launch would simply not get to them. Whether the first segment is bigger than the other is not even that important.

DavidC1 · Oct 10, 2024

KompuKare said:
Might the missing P cores gains have been a relatively late things due changes made after the RPL degradation failures became public?

Mainly the rumoured reduced ring bus frequency but maybe other things had to be dialed down too?

This not being Intel's on process might make them more cautious too as any accelerated aging tests might be harder.

That's the problem. Whatever that'll make the P core score better should also improve E cores too.

cannedlake240 · Oct 10, 2024

DavidC1 said:
Lol, so they kinda sandbagged the Skymont performance by saying 32% for the E cores and 9% for the P cores, when for Skymont it was just for Int, and for the P cores it was combined Int/FP.

I don't know why the gains got lowered from the initial 14%. Considering Lion Cove adds additional FP units too, it's very disappointing. That means Integer gains are possibly lower than 9%.

The Youtube Arrowlake video has the Intel guy basically ignoring Lion Cove and just praising Skymont. This is basically telling us what their future arch is.

Don't underestimate the P core, it could still outlast all other Intel uarch if it is seen as the lower risk, safe option. Switching over the CPU lineups to a new core is probably expensive as well, which Intel won't be happy about

coercitiv · Oct 10, 2024

cannedlake240 said:
Don't underestimate the P core, it could still outlast all other Intel uarch

The P core is not a uarch, it's a concept above implementation, describes resource allocation and performance targets. The uarch can be anything as long as it has the muscle.

gdansk · Oct 10, 2024

coercitiv said:
The P core is not a uarch, it's a concept above implementation, describes resource allocation and performance targets. The uarch can be anything as long as it has the muscle.

Exactly. And genealogically Skymont may have two sons for all we know...

DavidC1 · Oct 10, 2024

cannedlake240 said:
Don't underestimate the P core, it could still outlast all other Intel uarch if it is seen as the lower risk, safe option. Switching over the CPU lineups to a new core is probably expensive as well, which Intel won't be happy about

If switching allows Intel to have a much superior product, then they'll get more money over the long run, which is the point of investing.

32% Int/72% FP means there's a possibility there's less than 5% advantage for Lion Cove over Skymont, and that's with a full 3x core size difference. There's 23% top clock difference too, but not looking too impressive now.

Maybe that 5% difference can be made up simply by having the L3 cache as fast as Raptorlake? Who knows? Whatever made Lion Cove go from 14% to 9% would reasonably result in 32/72% going to 37%/75%.

If Arctic Wolf achieves another 30% general purpose gain, then we'll have a situation where Intel can claim it beats the last gen P core performance in outright performance, not just per clock.

cannedlake240 · Oct 10, 2024

coercitiv said:
The P core is not a uarch, it's a concept above implementation, describes resource allocation and performance targets. The uarch can be anything as long as it has the muscle.

"The P core" as in Core or P6 lineage, so Golden cove, LionCove...

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

Senior member

Attachments

Lifer

Diamond Member

Golden Member

Platinum Member

Diamond Member

Platinum Member

Member

Member

Lifer

Golden Member

Elite Member

Diamond Member

Diamond Member

Diamond Member

Golden Member

Member

Diamond Member

Golden Member

Diamond Member

Golden Member

Senior member

Diamond Member

Diamond Member

Golden Member

Senior member