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.)

Saylick · Oct 23, 2024

Welp, the article got deleted. Guess it was legit.

511 · Oct 24, 2024

Lol i had cached copy i just closed but looks like Meh Lake I am waiting for TPU no review does in-depth testing like them

Goop_reformed · Oct 24, 2024

511 said:
Lol i had cached copy i just closed but looks like Meh Lake I am waiting for TPU no review does in-depth testing like them

Show them cache

controlflow · Oct 24, 2024

Saylick said:
Arrowlake is pretty disappointing. Not going to lie. Performance seems to have regressed in more situations than not compared to 14th gen. What’s really killing it is the high latency to DRAM. It’s over 100ms in AIDA. Even with 9300 MT/s memory, it’s still in the mid-90s range.

Edit: At least Intel got the power down, even though this is a CPU stress test.
View attachment 110151

FWIW, a user on Reddit with a 265K claimed that by default they got over 100ns latency and Windows was forcing VBS on by default. Disabling VBS supposedly reduced latency by 30% and got the latency to within ~10ns of their 14900KS with fast memory.

Saylick · Oct 24, 2024

controlflow said:
FWIW, a user on Reddit with a 265K claimed that by default they got over 100ns latency and Windows was forcing VBS on by default. Disabling VBS supposedly reduced latency by 30% and got the latency to within ~10ns of their 14900KS with fast memory.

That’s reassuring.

511 · Oct 24, 2024

IF VBS is screwing performance it can't be blamed on CPU either AMD or Intel it's MSFT messing things as always

Hitman928 · Oct 24, 2024

controlflow said:
FWIW, a user on Reddit with a 265K claimed that by default they got over 100ns latency and Windows was forcing VBS on by default. Disabling VBS supposedly reduced latency by 30% and got the latency to within ~10ns of their 14900KS with fast memory.

VBS on is the default for Win 11. 30% reduction from over 100 ns puts it in the 70 ns range. Still decently high but obviously much better.

511 · Oct 24, 2024

If they had VBS on it was going to be a bad showing It's just a way to force security without properly fixing the problem focusing on Kernel instead of shoving Ads/bloatware and Crapeare to users

DavidC1 · Oct 24, 2024

jdubs03 said:
I know it’s one review (edit: now deleted but likely to be back soon lul) and I said what I said a little bit ago.
But this isn’t a good look:
View attachment 110150

But how much power is it using in that R23 test? The AIDA64 results are showing a dramatic reduction. 273W on 14900K to 196W on 285K.

If it's too low like on CP2077 then something isn't working right.

Sounds a bit like Granite Rapids showing platform issues.

controlflow · Oct 24, 2024

I forgot to link to the Reddit user detailing the latency and VBS issue:

https://www.reddit.com/r/hardware/comments/1gabovq/comment/ltcsf08

Given that the current state of this platform is such that the reviewer has to be sure to have the right Windows build, right FW and set all the OS and system settings just right, pretty good odds that we see some big variation in results when the embargo lifts soon.

Kocicak · Oct 24, 2024

controlflow said:
I forgot to link to the Reddit user detailing the latency and VBS issue:

https://www.reddit.com/r/hardware/comments/1gabovq/comment/ltcsf08

Given that the current state of this platform is such that the reviewer has to be sure to have the right Windows build, right FW and set all the OS and system settings just right, pretty good odds that we see some big variation in results when the embargo lifts soon.

Well, it seems that OC3D tested with VBS disabled and his poor gaming results were not caused by this? The latency is not over 100 ns.

controlflow · Oct 24, 2024

Kocicak said:
Well, it seems that OC3D tested with VBS disabled and his poor gaming results were not caused by this? The latency is not over 100 ns.

View attachment 110164 View attachment 110165 View attachment 110166 View attachment 110167 View attachment 110168 View attachment 110169

Yeah not entirely clear what's happening there but there are lots of other weird and nonsensical things in the OC3D results beyond ARL numbers so I'm not sure what to make of it.

Would be curious to know what Windows build was used there for the ARL systems since the user on Overclockers UK forums showed significant differences between the current and preview Windows builds in ARL performance.

I think ARL isn't going to be anything to write home about for gaming even when configured perfectly though. I suspect it will do better than what OC3D shows here and probably trade blows with existing top end RPL refresh but doing so while using a sane amount of power.

DrMrLordX · Oct 24, 2024

Looks like we may need more than just third party reviews, but rereviews thanks to impending microcode fixes and problems with Windows slowing down Arrow Lake. What a mess.

Sunaiac · Oct 24, 2024

Youtubers already rubbing their paws, gonna be able to milk this even more than Zen 5

Goop_reformed · Oct 24, 2024

DrMrLordX said:
Looks like we may need more than just third party reviews, but rereviews thanks to impending microcode fixes and problems with Windows slowing down Arrow Lake. What a mess.

What problems with windows?

DavidC1 · Oct 24, 2024

DrMrLordX said:
Looks like we may need more than just third party reviews, but rereviews thanks to impending microcode fixes and problems with Windows slowing down Arrow Lake. What a mess.

Sigh, so it'll be just like Granite Rapids. Serious software issues on launch, hopeful for a software fix.

SiliconFly · Oct 24, 2024

igor_kavinski said:
They had no plan B this time. Went all in on tiles and paid the price with their inexperience. They must've known this at least a year back. Still did not take any corrective measures to try to get the performance up to snuff. They must've thought that the Intel brand will just sell itself. Maybe it will. To an extent. For gamers and most rational enthusiasts, they almost do not exist anymore, unless they release Bartlett Lake next year to save face.

I don't think it's entirely accurate. They clearly know what it entails. Tiles are necessary cos they can't do large monolithic on advanced nodes anymore due to prohibitive costs. And compared to AMD's older substrate based chiplets design, Intel's tiles are still far ahead. In fact, Zen 6 itself is expected to use tiles (with interposer) instead of chiplets.

SiliconFly · Oct 24, 2024

jdubs03 said:
Tbh after the whole Zen 5 thing, I’ve just decided I’ll come to a conclusion after all of the reviews are in. Hopefully no required Windows updates/new firmware versions are needed to have an established understanding.

At this point, the way things are going, we're going to need a new µcode update, a bios update and a OS update to get the cpu to even run decently.

511 · Oct 24, 2024

Looks like Microsoft is screwing around with x86
First Zen 5 issues
LNL Issues
ARL Issues
We sure Microsoft is not pushing ARM ?

Saylick · Oct 24, 2024

SiliconFly said:
I don't think it's entirely accurate. They clearly know what it entails. Tiles are necessary cos they can't do large monolithic on advanced nodes anymore due to prohibitive costs. And compared to AMD's older substrate based chiplets design, Intel's tiles are still far ahead. In fact, Zen 6 itself is expected to use tiles (with interposer) instead of chiplets.

Hmmm. Last time I checked, tiles are chiplets. Also, I thought Zen 6 was using silicon bridges, not an interposer?

SiliconFly · Oct 24, 2024

Hitman928 said:
VBS on is the default for Win 11. 30% reduction from over 100 ns puts it in the 70 ns range. Still decently high but obviously much better.

I ain't sure we can't totally blame M$ for VBS issues. VBS has been there for many years and works real well on many generations of CPUs from both camps. If ARL has an issue with VBS in Win11, then it points more towards Intel than M$ (or maybe it's more like an interop issue).

coercitiv · Oct 24, 2024

Goop_reformed said:
What problems with windows?

We have reports that disabling VBS improves memory latency, people expect we'll see big swings in performance from one review to another, based on wheter security features are (properly) disabled or left enabled to emulate the casual user experience.

Kocicak · Oct 24, 2024

coercitiv said:
... wheter security features are (properly) disabled or left enabled to emulate the casual user experience.

You mean improperly then???

SiliconFly · Oct 24, 2024

Saylick said:
Hmmm. Last time I checked, tiles are chiplets. Also, I thought Zen 6 was using silicon bridges, not an interposer?

Not entirely. Tiles are not like AMD chiplets. Conversely, AMD chiplets can't be called Tiles either.

Tiles are an advanced from of chiplet design compared to older chiplet design AMD uses. AMD chiplets are hardwired on substrate. Intel Tiles sit on interposer that sits on top of substrate.

AMD is expected to switch to Intel's advanced Tile like chiplet configuration with Zen 6.

coercitiv · Oct 24, 2024

Kocicak said:
You mean improperly then???

I meant properly. We've had cases when reviewers thought some security features were disabled but were in fact still active. What we think about security and the recommended settings is another issue.

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

Senior member

Attachments

Diamond Member

Golden Member

Senior member

Member

Diamond Member

Golden Member

Diamond Member

Golden Member

Golden Member

Member

Golden Member

Member

Lifer

Member

Senior member

Golden Member

Golden Member

Golden Member

Golden Member

Diamond Member

Golden Member

Diamond Member

Golden Member

Golden Member

Diamond Member