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.)

AcrosTinus · Nov 9, 2024

I might be reading too much into his answer but the way he answered the 3D-Cache questions implies that Intel might be trying to introduce something equivalent.

If Intel really can restore the core ultra 200S to raptor performance, they might regain my trust. I still won't be buying until APX, AVX10.2 and their big cache CPU is introduced.

The gap in gaming is like a death sentence and they need to fix it ASAP, even being within 5% to x3D is fine by me otherwise they are going to run into the Radeon trap, where the 3090 and 4090 with DLSS, FG, Cuda and much more have outrun the competition so far that most people default to Nvidia ignoring value.

gdansk · Nov 9, 2024

AcrosTinus said:
I might be reading too much into his answer but the way he answered the 3D-Cache questions implies that Intel might be trying to introduce something equivalent.

If Intel really can restore the core ultra 200S to raptor performance, they might regain my trust. I still won't be buying until APX, AVX10.2 and their big cache CPU is introduced.

The gap in gaming is like a death sentence and they need to fix it ASAP, even being within 5% to x3D is fine by me otherwise they are going to run into the Radeon trap, where the 3090 and 4090 with DLSS, FG, Cuda and much more have outrun the competition so far that most people default to Nvidia ignoring value.

It isn't a similar scenario. Note that the 3090 doesn't have FG but its Radeon competitor from the same era does. More over it didn't run away in general performance; only the 4090 did. People don't ignore value for features, they value features.

Arrow Lake still - generally - has more features than Granite Ridge and they add some value. But it's not enough to make up for the gaming performance gap to most people.

fastandfurious6 · Nov 9, 2024

title looks like a parody....

Intel Core Ultra 9 285H makes Geekbench debut with 16-core CPU and severe throttling

Intel's Core Ultra 9 285H Arrow Lake-H laptop processor has shown up on Geekbench. It has a 16-core CPU (6 P-cores + 10 E-cores) and boosts up to 5.4 GHz. It was tested alongside a Dell laptop with 64 GB RAM.

www.notebookcheck.net

sl0519 · Nov 9, 2024

AcrosTinus said:
I might be reading too much into his answer but the way he answered the 3D-Cache questions implies that Intel might be trying to introduce something equivalent.

If Intel really can restore the core ultra 200S to raptor performance, they might regain my trust. I still won't be buying until APX, AVX10.2 and their big cache CPU is introduced.

The gap in gaming is like a death sentence and they need to fix it ASAP, even being within 5% to x3D is fine by me otherwise they are going to run into the Radeon trap, where the 3090 and 4090 with DLSS, FG, Cuda and much more have outrun the competition so far that most people default to Nvidia ignoring value.

Isn't 3d v-cache invented by TSMC? Why can't they just use it?

gdansk · Nov 9, 2024

sl0519 said:
Isn't 3d v-cache invented by TSMC? Why can't they just use it?

No TSVs on core? SOIC-X not ready for N3 until next year? Who knows.

C1 · Nov 9, 2024

Intel pledges to ‘fix’ Arrow Lake after launch ‘didn’t go as planned’

511 · Nov 9, 2024

sl0519 said:
Isn't 3d v-cache invented by TSMC? Why can't they just use it?

TSV design consideration and many more things to consider

511 · Nov 9, 2024

fastandfurious6 said:
title looks like a parody....

Intel Core Ultra 9 285H makes Geekbench debut with 16-core CPU and severe throttling

Intel's Core Ultra 9 285H Arrow Lake-H laptop processor has shown up on Geekbench. It has a 16-core CPU (6 P-cores + 10 E-cores) and boosts up to 5.4 GHz. It was tested alongside a Dell laptop with 64 GB RAM.

www.notebookcheck.net

it is a windows issue lol i have tested it with my CPU as well windows smokes some crazy things

here are the results link

Alienware Alienware m16 R2 - Geekbench

Benchmark results for an Alienware Alienware m16 R2 with an Intel Core Ultra 9 185H processor.

browser.geekbench.com

Alienware Alienware m16 R2 - Geekbench

Benchmark results for an Alienware Alienware m16 R2 with an Intel Core Ultra 9 185H processor.

browser.geekbench.com

inf64 · Nov 9, 2024

C1 said:
Intel pledges to ‘fix’ Arrow Lake after launch ‘didn’t go as planned’

I'm not so optimistic they can fix anything performance-wise. Most reviewers got the expected performance (somewhat lower than 14900K which matches their slide deck). What they can fix is anti-cheat engines incompatibility and some crashes that were observed. Performance wise, it is what it is. There is no catching up to 35-40% deficit versus X3D parts.

511 · Nov 9, 2024

inf64 said:
I'm not so optimistic they can fix anything performance-wise. Most reviewers got the expected performance (somewhat lower than 14900K which matches their slide deck). What they can fix is anti-cheat engines incompatibility and some crashes that were observed. Performance wise, it is what it is. There is no catching up to 35-40% deficit versus X3D parts.

yes no doubt about it hardly 4-5% more performance best case these would have been better off in HX designs

SiliconFly · Nov 9, 2024

inf64 said:
I'm not so optimistic they can fix anything performance-wise. Most reviewers got the expected performance (somewhat lower than 14900K which matches their slide deck). What they can fix is anti-cheat engines incompatibility and some crashes that were observed. Performance wise, it is what it is. There is no catching up to 35-40% deficit versus X3D parts.

I think their slide deck did show 285K slightly ahead of 14900K in many cases which we aren't seeing in reality. Thats what they're trying to fix.

inf64 · Nov 9, 2024

SiliconFly said:
I think their slide deck did show 285K slightly ahead of 14900K in many cases which we aren't seeing in reality. Thats what they're trying to fix.

Yeah but it is intel's slides, they would never show lower performance than the previous gen (which is expected given the compromises they made and lackluster IPC improvements). I really don't see this thing getting any better except for stability fixes.

sl0519 · Nov 9, 2024

511 said:
TSV design consideration and many more things to consider

So basically they didn't design their CPUs in a way that can fully utilize TSMC's 3d stacking technology, am I reading that right?

511 · Nov 9, 2024

sl0519 said:
So basically they didn't design their CPUs in a way that can fully utilize TSMC's 3d stacking technology, am I reading that right?

Yes you need TSVs to stack cache physically and also design the processor for that

OneEng2 · Nov 9, 2024

inf64 said:
I'm not so optimistic they can fix anything performance-wise. Most reviewers got the expected performance (somewhat lower than 14900K which matches their slide deck). What they can fix is anti-cheat engines incompatibility and some crashes that were observed. Performance wise, it is what it is. There is no catching up to 35-40% deficit versus X3D parts.

I agree. There is no way 285K can be "fixed" to compete with 9800X3D rather on the higher clocked versions soon to be released.

511 said:
Yes you need TSVs to stack cache physically and also design the processor for that

I do believe that there is no real reason that Intel can't use stacked cache (they are pretty good designers after all). There may well be little benefit to it without a major redesign though.

Just as Intel got little improvement for SMT due to their core design, it may be that the current architecture also doesn't benefit as much as Zen5 from stacked L3 ..... or because of the slow a** ring bus, it may get even more benefit. My point is that I suspect there is a good reason that Intel isn't doing it, and it isn't likely because they can't figure it out.

It is more likely that it doesn't make sense either from a performance or cost standpoint.

511 · Nov 9, 2024

OneEng2 said:
I agree. There is no way 285K can be "fixed" to compete with 9800X3D rather on the higher clocked versions soon to be released.

I agree on this one maybe 4-5% gains but that's about it should be able to get on par with 9950X though in games

SiliconFly · Nov 9, 2024

inf64 said:
Yeah but it is intel's slides, they would never show lower performance than the previous gen (which is expected given the compromises they made and lackluster IPC improvements). I really don't see this thing getting any better except for stability fixes.

What he basically says is, there is difference between their internal performance assessment and what they seeing out in the wild. And they're working on update(s) to fix just that. An improvement is coming. How much exactly is the question.

OneEng2 said:
I agree. There is no way 285K can be "fixed" to compete with 9800X3D rather on the higher clocked versions soon to be released.

285K's "fix" has nothing to do with 9800X3D. The "fix" is supposed to put 285K slightly ahead of 14900K just to avoid the currently observed regression in general workloads. Regression isn't good for sales & product image (like MTL).

OneEng2 said:
I do believe that there is no real reason that Intel can't use stacked cache (they are pretty good designers after all). There may well be little benefit to it without a major redesign though.

They can. They've chosen not to. Zen 5 has two stacked chiplets (cpu & llc) on top of the substrate. In Intel's case, it'll be three (imterposer, cpu & llc) which complicates things. The best solution for Intel is to put the llc in the interposer itself (something they've been working on for quite sometime, adm cache). It debuts next year with CWF & should enter mainstream later.

maddie · Nov 9, 2024

sl0519 said:
Isn't 3d v-cache invented by TSMC? Why can't they just use it?

SOIC stacked cache? I thought it was a collaboration between AMD and TSMC. Do we know if there was an exclusive use period in the agreement?

V-cache does not have to be SOIC, but you probably lose some benefits otherwise.

igor_kavinski · Nov 9, 2024

511 said:
Apparently Hallock was asked how did they improved productivity performance in Cinebench/Blackmagic etc he said by removing SMT LMAO should have said our E cores are better and P cores are S###

OMG no wonder AMD fired him!

poke01 · Nov 9, 2024

igor_kavinski said:
OMG no wonder AMD fired him!

Even with him gone AMD is no stranger to poor marketing. Remember the cringe RDNA3 keynote?

DavidC1 · Nov 9, 2024

511 said:
Apparently Hallock was asked how did they improved productivity performance in Cinebench/Blackmagic etc he said by removing SMT LMAO should have said our E cores are better and P cores are S###

In certain applications and benchmarks, this might be literally true.

Let's say an Application A behaves as the following:
-1-12 threads it scales nearly linearly
-13-20 threads it starts tapering off
-21-24 threads the gains really diminish
-Supports a maximum of 24 threads

While in an ideal scenario, a 8P core with HT + 16E cores should not be slower, because it'll prioritize E cores over HT, not all scenarios behave perfectly and when you throw vastly varying usage into the mix it's even more true.

Oftentimes, there's a non-negligible performance loss when HT is added in an application where the thread count is maxed out without HT.

This will be even more true in Arrowlake, where Skymont is far more capable relatively than it did in previous generations, because you now have diminishing returns pegged to HT versus Skymont and adding conflicts in scenarios where supported thread count is lower than maximum available.

511 · Nov 9, 2024

DavidC1 said:
In certain applications and benchmarks, this might be literally true.

Let's say an Application A behaves as the following:
-1-12 threads it scales nearly linearly
-13-20 threads it starts tapering off
-21-24 threads the gains really diminish
-Supports a maximum of 24 threads

While in an ideal scenario, a 8P core with HT + 16E cores should not be slower, because it'll prioritize E cores over HT, not all scenarios behave perfectly and when you throw vastly varying usage into the mix it's even more true.

Oftentimes, there's a non-negligible performance loss when HT is added in an application where the thread count is maxed out without HT.

This will be even more true in Arrowlake, where Skymont is far more capable relatively than it did in previous generations, because you now have diminishing returns pegged to HT versus Skymont and adding conflicts in scenarios where supported thread count is lower than maximum available.

While this may be true but the benchmarks in question were cinebench and Balckmagic i can consider this for blackmagic but Cinbench is purely carried by Sktmont can you say this in context of cinebench🤣
Here is the timestamp

DavidC1 · Nov 9, 2024

SiliconFly said:
Thats fantastic! I was a bit worried that Arctic Wolf might swell up a bit to nearly 1:2 ratio.

Skymont grew a lot because of the doubled FP section, which is likely responsible for 20-30% growth in die area alone. Without that, it would have been well under 1mm2, and be close to 1:4 ratio again.

Intel's numbers of 32% Int and 72% FP means the gain due to doubled FP is approximately 1.72/1.32 or 30%. 20-30% die area growth for 30% FP improvement.

The 30% integer gain came at the cost of roughly 30% core area increase when you take out the doubled FP, assuming ~1.6x N3B density gain over Intel 4, which is an excellent figure. It might be even more optimistic than this, depending on the actual difference between N3B and Intel 4.

This is what the E core team does, likely they have a goal they have and works to meet that goal.

Since the 8% performance projection means the P core will grow too, the only way I see going to 1:2 ratio is they put even more work at aiming at replacing the P cores, not just being a companion chip. So a 40-50% gain.

alcoholbob · Nov 9, 2024

AcrosTinus said:
I might be reading too much into his answer but the way he answered the 3D-Cache questions implies that Intel might be trying to introduce something equivalent.

If Intel really can restore the core ultra 200S to raptor performance, they might regain my trust. I still won't be buying until APX, AVX10.2 and their big cache CPU is introduced.

The gap in gaming is like a death sentence and they need to fix it ASAP, even being within 5% to x3D is fine by me otherwise they are going to run into the Radeon trap, where the 3090 and 4090 with DLSS, FG, Cuda and much more have outrun the competition so far that most people default to Nvidia ignoring value.

There's already rumors of a 144MB L3 cache part for Nova Lake being considered, but this is something they are looking for a late 2026 or early 2027 release. Certainly not going to be out anytime soon and X3D parts will likely have no competition for the next 2 or 3 years.

Ranulf · Nov 9, 2024

Another link on the hot hardware podcast info:

Intel Identifies Issues with the Launch of Its Core Ultra 200S Desktop Arrow Lake Desktop CPUs and Promises Fixes Are Forthcoming

There's good news on the horizon for the Core Ultra 200 series Arrow Lake processors as Intel identifies performance-hindering issues.

www.thefpsreview.com

Transcript:
“I think what people have been interested to hear is what happened. I can’t go into all the details yet, but we identified a series of multifactor issues at the OS level, at the BIOS level, and I will say that the performance we saw in reviews (through no fault of reviewers) is not what we expected and not what we intended. The launch just didn’t go as planned. That has been a humbling less for all of us, inspiring a fairly large response internally to get to the bottom of what happened and to fix it.”

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

Senior member

Attachments

Member

Diamond Member

Senior member

Member

Diamond Member

Platinum Member

Golden Member

Golden Member

Diamond Member

Golden Member

Golden Member

Diamond Member

Member

Golden Member

Senior member

Golden Member

Golden Member

Diamond Member

Lifer

Platinum Member

Golden Member

Golden Member

Golden Member

Diamond Member

Platinum Member