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

Page 103 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Tigerick

Senior member
Apr 1, 2022
696
602
106
PPT1.jpg
PPT2.jpg
PPT3.jpg



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

ModelCode-NameDateTDPNodeTilesMain TileCPULP E-CoreLLCGPUXe-cores
Core Ultra 100UMeteor LakeQ4 202315 - 57 WIntel 4 + N5 + N64tCPU2P + 8E212 MBIntel Graphics4
?Lunar LakeQ4 202417 - 30 WN3B + N62CPU + GPU & IMC4P + 4E08 MBArc8
?Panther LakeQ1 2026 ??Intel 18A + N3E3CPU + MC4P + 8E4?Arc12



Comparison of die size of Each Tile of Meteor Lake, Arrow Lake, Lunar Lake and Panther Lake

Meteor LakeArrow Lake (20A)Arrow Lake (N3B)Lunar LakePanther Lake
PlatformMobile H/U OnlyDesktop OnlyDesktop & Mobile H&HXMobile U OnlyMobile H
Process NodeIntel 4Intel 20ATSMC N3BTSMC N3BIntel 18A
DateQ4 2023Q1 2025 ?Desktop-Q4-2024
H&HX-Q1-2025
Q4 2024Q1 2026 ?
Full Die6P + 8P6P + 8E ?8P + 16E4P + 4E4P + 8E
LLC24 MB24 MB ?36 MB ?12 MB?
tCPU66.48
tGPU44.45
SoC96.77
IOE44.45
Total252.15

LNL-MX.png

Intel Core Ultra 100 - Meteor Lake

INTEL-CORE-100-ULTRA-METEOR-LAKE-OFFCIAL-SLIDE-2.jpg

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.)



Clockspeed.png
 

Attachments

  • PantherLake.png
    PantherLake.png
    283.5 KB · Views: 24,006
  • LNL.png
    LNL.png
    881.8 KB · Views: 25,490
Last edited:

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
RCW PPC would be cool to know.
there's been a lot of discussion on redwood and crestmont cores. if all the bs about intel figuring out dlvr is any true and that's a stretch short of a miracle along with their advanced packaging and gaafet coming in line with future nodes then hoh hoh hoh santta's here.

meteor lake seems like a dud performance wise much like zen 4 is. it's here to set up a future for both x86 makers. my internal timeline of when gaafet bspd and dlvr become the standard across intel's range is jumbled up but I have no doubt gelsinger the gunslinger isn't getting intel's future in order, not all credit to him of course but tsmc better be looking in their rear view mirror to see gelsinger coming up on em snarling like a bull. amd can throw as many red tomaters as they want like the tomatina festival in bouni at intel with more cores to keep them at bay but eventually amd must also hunker down and deliver a respectful product. zen 5 will be a late 1h or early to mid 2h 24 product and arrow lake is a late q3 product. by then ddr5 prices will have normalised and board prices... the aibs can shove it where the sun don't shine but people won't be whining about prices then. both will be compelling products, who sels how much is up to anyone's guess.
 

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
This was expected. this works out better than what most of us were assuming was going to happen not long ago; which is to say the two generations side by side. based on the crap coming out about raptor's refresh it'll be a n a ok stop gapper until arrow and outpace zen 4 but no the 3d chips. meteor on laptop with its powerful igpu is way more important than it being on desktop. intel's best recourse is the mobile market where they have the upper handed in volume and being able to flood the market with their processor and getting favorable build qualities in laptop chasis from aibs where it may be a mystery meat concoction for amd branded requpped laptops.
 

Hougy

Member
Jan 13, 2021
80
62
61
They really tested MT instead of ST performance :\
It must be sad for Intel that they renamed all their nodes to try to more accurately show how comparable they are to the competitor's, but HWINFO, and probably cpu-z too, keeps showing Intel 4 as 7nm 😂. What do you think they will show for 20A and 18A? 5nm?
 
  • Haha
Reactions: Tlh97 and moinmoin

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
The chip reports itself. Software parses the data. That's all. This is common knowledge 101. nm mean nothing and haven't meant nothing in years for any of the three major fabs. density varies greatly between each "node" from each manufacturer down to levels within a "node". you can ring out amazing performance on an old node with higher power draw to be determined and can also get crappy performance on a bleeding edge node. too many variables.
 

eek2121

Diamond Member
Aug 2, 2005
3,100
4,398
136
80%? you believe intel figured out how to run dlv whatever that stuff is called on desktop processors? you lot need to stop drinking the mlid and raichu energy drinks.
We have actual leaks, my dude. This particular chip (below) is "only" clocked 63% higher than the equivalent Raptor Lake chip while containing 2 more "e" cores (rumored to be on the SoC). The PL2 of 28W is unchanged. This isn't even the highest clocked SKU, and it appears to be an ES chip.

Now imagine this chip in a 2 lb ultrabook:
Intel-Meteor-lake-Computex-Screenshot-HWINFO-Laptop-1.jpg
I'm fairly pessimistic myself, but I can also do basic math. Unless there are IPC regressions, this looks to be a decent offering by Intel. We will see when the reviews and such drop, however.
 

Exist50

Platinum Member
Aug 18, 2016
2,452
3,102
136
there's been a lot of discussion on redwood and crestmont cores. if all the bs about intel figuring out dlvr is any true and that's a stretch short of a miracle along with their advanced packaging and gaafet coming in line with future nodes then hoh hoh hoh santta's here.
DLVR is very real, but it doesn't merit such hype. It's analogous to FIVR. Useful, but not groundbreaking.
Unless there are IPC regressions, this looks to be a decent offering by Intel.
I wouldn't be surprised if there's some memory latency sensitive test that shows a regression. Wouldn't really be the core's fault per se, but hard to isolate as well.
 

dullard

Elite Member
May 21, 2001
25,511
4,009
126
raising the base clocks by more than 80%.

80%? you believe intel figured out how to run dlv whatever that stuff is called

This particular chip (below) is "only" clocked 63% higher than the equivalent Raptor Lake chip

I think you two are talking past each other. @eek2121, your link at 63% does not YET prove your 80% claim. And 80% base increase does not necessarily translate into much benchmark performance increase. Especially since at least in that leak the turbo speed dropped 21%. Yes, this isn't a final version. But we don't YET have leaked data to show final performance.

@A///, Intel had DLVR in Raptor Lake but fused it off. Heck it is still in their Raptor Lake electrical specifications updated Feb 2023.
I'd love to know the story as to why it was fused off.

I personally am betting that Meteor Lake will have some nice features, which can in some cases give much improved performance. But, I think the real winner will be Arrow Lake.
 

H433x0n

Golden Member
Mar 15, 2023
1,209
1,572
96
Noobish question that I can hopefully get answered:

When Intel advertises a base clock - What exactly does it mean? I had assumed it meant this is the minimum sustainable clock speed at full utilization for the advertised PL2.. Is this a correct assumption?
 

Exist50

Platinum Member
Aug 18, 2016
2,452
3,102
136
Noobish question that I can hopefully get answered:

When Intel advertises a base clock - What exactly does it mean? I had assumed it meant this is the minimum sustainable clock speed at full utilization for the advertised PL2.. Is this a correct assumption?
The closest definition seems to be guaranteed clocks at PL1, though AVX/AMX workloads can have their own base clocks.
 

dullard

Elite Member
May 21, 2001
25,511
4,009
126
When Intel advertises a base clock - What exactly does it mean? I had assumed it meant this is the minimum sustainable clock speed at full utilization for the advertised PL2.. Is this a correct assumption?
@Exist50 is correct. I'll just add in Intel's definition that they list with every single processor on their ark website if you click the question mark for Base Power (Example: https://ark.intel.com/content/www/u...1370p-processor-24m-cache-up-to-5-20-ghz.html ):

"Processor Base Power
The time-averaged power dissipation that the processor is validated to not exceed during manufacturing while executing an Intel-specified high complexity workload at Base Frequency and at the junction temperature as specified in the Datasheet for the SKU segment and configuration"

This gives you Exist50's comment: If the processor is running at it's rated base speed, then the CPU will not exceed the processor's rated base power (PL1).

There are three caveats in their statement. (1) CPUs might behave differently than their claims at extreme temperatures. (2) This is a specific workload but not the most power hungry workload possible (a power virus). A power virus might do something like switch every transistor from 0 to 1 and back every cycle which takes a lot of power to charge and discharge all of those transistors billions of times per second. (3) The words "time-averaged" gives them leeway for brief periods to go above or below (PL1)--they aren't trying to hit that target perfectly at every possible microsecond.

The unstated 4th caveat: If the processor is running at it's turbo speed then it most likely will run at a much higher power level (up to PL4 for a brief moment).
 
Last edited:

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
We have actual leaks, my dude. This particular chip (below) is "only" clocked 63% higher than the equivalent Raptor Lake chip while containing 2 more "e" cores (rumored to be on the SoC). The PL2 of 28W is unchanged. This isn't even the highest clocked SKU, and it appears to be an ES chip.

Now imagine this chip in a 2 lb ultrabook:
Intel-Meteor-lake-Computex-Screenshot-HWINFO-Laptop-1.jpg
I'm fairly pessimistic myself, but I can also do basic math. Unless there are IPC regressions, this looks to be a decent offering by Intel. We will see when the reviews and such drop, however.
I don't know which rumor release this pertains to but I like to take a very conservative approach on engineering sample results like this or the hardware but mostly in light because of the very recent leak where the clock freq is stuck at around 680 mhz. In the 12 hours since I've seen a lot of hot takes, mostly dumb, trying to extrapolate end performance. I'm not accusing you or anyone here of that because what I've seen has been lowballing intel big time.

Engineering samples can be real old or recent. The laptop in question with the clock issue could have been using an es gotten recently or it could be over a year to maybe 2 years old depending on how intel operates. I know gelsinger mentioned they powered on meteor lake for the first time a hell of a long time ago.

There was rumor a while back dlvr was solved by Intel and they got it working on mobile and desktop. it's why there's a lot of hot bull about the raptor refresh. I've seen high 5 all cores with mid 6 2 core boosts. while using only a little more than the package limit. not mentioned online in rumors I've been told my own dire rumors of the 14900K refresh having quite a bit of oc room from stock allowing it to soar while getting close but nowhere near the pl2 limit. This coupled with enduring 30 years of Intel and AMD's bs is why I'm so skeptical of anything I see until an official product launches and is reviewed by independent reviewers. This ain't my first rodeo and hopefully won't be my last.
 

eek2121

Diamond Member
Aug 2, 2005
3,100
4,398
136
I don't know which rumor release this pertains to but I like to take a very conservative approach on engineering sample results like this or the hardware but mostly in light because of the very recent leak where the clock freq is stuck at around 680 mhz. In the 12 hours since I've seen a lot of hot takes, mostly dumb, trying to extrapolate end performance. I'm not accusing you or anyone here of that because what I've seen has been lowballing intel big time.

Engineering samples can be real old or recent. The laptop in question with the clock issue could have been using an es gotten recently or it could be over a year to maybe 2 years old depending on how intel operates. I know gelsinger mentioned they powered on meteor lake for the first time a hell of a long time ago.

There was rumor a while back dlvr was solved by Intel and they got it working on mobile and desktop. it's why there's a lot of hot bull about the raptor refresh. I've seen high 5 all cores with mid 6 2 core boosts. while using only a little more than the package limit. not mentioned online in rumors I've been told my own dire rumors of the 14900K refresh having quite a bit of oc room from stock allowing it to soar while getting close but nowhere near the pl2 limit. This coupled with enduring 30 years of Intel and AMD's bs is why I'm so skeptical of anything I see until an official product launches and is reviewed by independent reviewers. This ain't my first rodeo and hopefully won't be my last.
I am honestly unsure of DLVR. I know (but cannot prove) that Intel has hit 5.4 ghz on Intel 4. I suspect we won't see it in released products. We will see, however. I am almost positive MTL-P is hitting 3.1 ghz base clocks @ 28 watts unless someone did an amazing job at faking screenshots (which I doubt), Finally, I also strongly suspect there is at least one SKU with 3.4 ghz base based on linux DMESG.

Unsure how much of it comes down to process or power saving tech, but it IMO it really doesn't matter.
 
  • Like
Reactions: Tlh97 and A///

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
I am honestly unsure of DLVR. I know (but cannot prove) that Intel has hit 5.4 ghz on Intel 4. I suspect we won't see it in released products. We will see, however. I am almost positive MTL-P is hitting 3.1 ghz base clocks @ 28 watts unless someone did an amazing job at faking screenshots (which I doubt), Finally, I also strongly suspect there is at least one SKU with 3.4 ghz base based on linux DMESG.

Unsure how much of it comes down to process or power saving tech, but it IMO it really doesn't matter.
the article got deleted i guess they got a dmca notice from intel or msi. for dlvr the way I see it is it's baked into the silicon and can't be disabled without microcode changes at the bios level. it based on my personal dreams is it opens up the processor's ability to hit higher clocks while reducing the power needed through clever whatever you want to call it. in low power states or power plans it'll be efficient. I'm not sure where it'll stand with amd's power framework which amd talks about in their ryzen 5 years later video, and future power states, or whether dlvr is the first of many steps but both approaches are promising tech not just for mobile platforms but for non mobile platforms including desktop and dc. it'll take several generations for both amd to intel to fine tune and hone their development, but other tech dropping in consumer over the next 3-4 years is very exciting on multiple fronts.

for us mortals it's a quiet waiting game to hand over our hard earned cash.
 
Last edited:

A///

Diamond Member
Feb 24, 2017
4,351
3,158
136
hasen's website ran an article a few hours ago and then changed the date as if it were published 13 hours ago when it was up top in the fold. it's a rumor article from an mdipsh** video, nothing interesting but the rumored return of an 8+32 config pulled out of mdipsh**'s behind. There is also something called beast lake which sounds made up and rumors of zen 6 pushing more cores. I assume by zen 6 am6 will be the socket because Zen4 and its 3D are technically 2 generations of processors.

My humble take for Arrow Lake is an 8+16+2/4 combo. after slogging through some of @Geddagod's posts I realized the point of there being cores on the soc tile to avoid unncessary wake up on the compute tiles leading to lower power use unless more than 2 or 4 cores require unparking or wake status if you will. very interesting. gonna be fun to see how those will work mobile wise starting with meteor lake p in a few months hopefully.
 

TESKATLIPOKA

Platinum Member
May 1, 2020
2,523
3,038
136
Let's see.
1.) Meteor Lake will be faster than Raptor by an unknown amount.
2.) Arrow Lake will be 30-40% faster in ST than Raptor Lake and 40% faster in MT than Meteor Lake with the same core count.
3.) Panther lake will be 30-40% faster in ST and 15-20% faster in MT than Arrow Lake with the same core count.

Summary:
Panther lake will be according to MLID 69-96% faster in ST than Raptor Lake and 61-68% faster in MT than Meteor Lake and this with the same core count and within ~3 years.
Conclusion:
This is a total BS, especially that ST performance. Frequency won't increase significantly, so It means performance should come mostly from IPC.
>50% increase in IPC, what a nonsense in 3 years.

@dullard You are right about that ST performance, I fixed my post.
 
Last edited:

dullard

Elite Member
May 21, 2001
25,511
4,009
126
2.) Arrow Lake will be 30-40% faster in ST and 40% faster in MT than Meteor Lake with the same core count.
Note: the portion in red is NOT what MLID said. So the rest of your math is not what he claims either.

He specifically claimed (without any evidence) of 30% to 40% ST faster than Raptor Lake not Meteor Lake (and not even Raptor Lake Refresh). That would be 30% to 40% ST with 2 node shrinks and a new core (3 node shrinks if you count skipping Intel 3). That would require Arrow Lake turbo to be roughly, mid 6 GHz to mid 7 GHz depending on IPC gains.

I'm not saying MLID is correct, but you took his rumors way past even what MLID claims.
 
Last edited:

ondma

Diamond Member
Mar 18, 2018
3,005
1,528
136
Note: the portion in red is NOT what MLID said. So the rest of your math is not what he claims either.

He specifically claimed (without any evidence) of 30% to 40% ST faster than Raptor Lake not Meteor Lake (and not even Raptor Lake Refresh). That would be 30% to 40% ST with 2 node shrinks and a new core (3 node shrinks if you count skipping Intel 3). That would require Arrow Lake turbo to be roughly, mid 6 GHz to mid 7 GHz depending on IPC gains.

I'm not saying MLID is correct, but you took his rumors way past even what MLID claims.
I could believe that ArL could be 30% (40% seems a bit much) faster than Raptor Lake. Like you said, that is two generations, and also Alder/Raptor/Raptor Refresh has been stalled for a while with small IPC gains. I am not that familiar with Panther, but another huge gain on top of Arrow Lake seems very unrealistic. Maybe it is the long rumored Royal Core????
 

Exist50

Platinum Member
Aug 18, 2016
2,452
3,102
136
This is a total BS, especially that ST performance. Frequency won't increase significantly, so It means performance should come mostly from IPC.
>50% increase in IPC, what a nonsense in 3 years.
Maybe Royal could do that, but Cougar Cove? Lol. This should be clear by now, but a single team is simply incapable of delivering significant yearly IPC gains.
MLID is a very good source, most of his leaks have come true or very close to. The times he has been incorrect has been due biased sources
He's been full of it. Pretty much every number he ever throws out is straight BS. Occasionally he gets some detail right, but it's pointless when the only people who can find the needle in the haystack already know.
I am not that familiar with Panther, but another huge gain on top of Arrow Lake seems very unrealistic. Maybe it is the long rumored Royal Core????
Royal is Royal. Lion Cove is not Royal. Cougar Cove isn't Royal. Panther Cove isn't Royal. Royal is Royal. Beast Lake may be the SoC, but the core name is not ambiguous.
 
  • Like
Reactions: Tlh97

SiliconFly

Golden Member
Mar 10, 2023
1,541
897
96

3:2 aspect ratio suggests to equip itself with MTL-P.
MTL-P gets 60% frequency improvement at iso-power (c.f. i7-1370P/6P8E/1.9GHz/28W).
When compared to Intel 7, Intel 4 gets 20% frequency improvement at iso-power. Or we can expect Intel 4 to have around 35% to 40% less power usage compared to Intel 7 at iso-freq (either one but not both at the same time).

60% frequency improvement for MTL-P at iso-power over previous gen RPL-P equivalent is just not possible i think.
 
Last edited:

eek2121

Diamond Member
Aug 2, 2005
3,100
4,398
136
When compared to Intel 7, Intel 4 gets 20% frequency improvement at iso-power. Or we can expect Intel 4 to have around 35% to 40% less power usage compared to Intel 7 (either one but not both at the same time).

60% frequency improvement for MTL-P at iso-power over previous gen RPL-P equivalent is just not possible i think.
Likely a combination or architectural improvements, DLVR, and other power management improvements.

This graph from AT is actually pretty informative:

1685737952442.png