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	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 (20A)	Arrow Lake (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	4P + 4E	4P + 8E
LLC	24 MB	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.)

 

[Wolverine2349]

Trust me when I say I like Intel more than I should. But placing too much trust on Skymont doesn't sound very healthy. They awesome. But not that awesome. They have their limits. And these limits are the only reason Intel still includes Lion Cove core in the processor. Totally disabling P cores sounds a bit extreme imho.


I think he's trying to build his own custom version of M3/M4 using just the overclocked Skymonts which in my opinion might lead to disappointment. I'm not sure whether he's right or wrong or my understanding of his is off and don't want to pour cold water into his possibly interesting ideas. The question is, is it even feasible? idk.


Every person has to find their own way. And from what I can see, you're brave! But are you on the right track? Thats a question I can't answer.

You are very likely right. They may have limitations and heck you may not even being able to boot without one P core.

And they may in theory have Raptor Cove IPC but we shall see if it translates to real world usage.

There is 0 option shamefully from either Intel or AMD with more than 8 strong cores on a single die. Last thing was Comet Lake and that IPC is so outdated being Skylake like nevermind being stuck on PCIe Gen 3.

And really this would all be moot of only Intel would just make a 12 Lion Cove core die. But for some reaosn they will not. I doubt there is a financial problem stopping them. They are Intel afterall and have massive amount of money and resources regardless of their currrent setback.

But its preobablky not an ideal solution and Skymont probably is not really Raptor Cove IPC across all workloads. If its too good to be true it probably is which is why Lion Cove is included. Just cause core to core latency looks normal and good, does not mean the actual cores themselves clock normalized perform on par all workloads as Raptor Cove??

But maybe I will find out with benchmarks once its released to see what it really has in store.

 

[MoistOintment]

Man, I really wish LNL had more cores. And a larger power budget to feed those cores and get higher clockspeeds. And upgradable memory past 64GB. The laptops could be thicker to accommodate the extra cooling...

oh wait, whoops. I just described ARL-H.

If you don't understand why people want laptops that target peak efficiency at 17W instead of ~23W (a 35% increase), or why some people would rather the laptop budget go towards a better screen, speakers, Webcam, fit and finish, materials, keyboard, trackpad, etc. Then you clearly aren't a part of the target market for LNL. People need to stop wishing LNL would change to meet their needs instead of buying a plethora of other chips that already do.

LNL perfectly meets my needs. It's significantly faster than the i5-1035G7 that is more than sufficient for my work laptop. It's significantly better battery life. And it can actually run some games when I'm back at the hotel when I need to travel for work.

I need all day battery, snappy performance, sufficient gaming, and it needs to look nice in meetings.

 

[Wolverine2349]

Yeah indeed I'm just not sure it will have the outcome desired. highly unlikely to OC to 5Ghz without straying way outside efficiency curve at the very least.

I get the desire for high performance monolithic solutions with higher core counts, but they aint happening unfortunately. Can't push something that isn't into that realm though. the end outcome is performance at the end of the day, not latency figures measured in ns

Yeah I am unsure too as I said. I am just hoping based on IPC reports that say its like Raptor Cove IPC. Will that translate to real world Raptor Cove IPC and latency clock normalized.

If its too good to be true it probably is unfortunately. But I can always hope. IO just wish Intel or AMD would release 10-12 non-hybrid good CPU cores on a single die.

Its not coming from AMD based on their design limitations and refusal to have more than one stepping for DIY and OEMs and Threadripper (They just cram a bunch of 8 core CCDs in Threadripper and EPYC parts). Intel could do it with their design, but they so far have not.

 

[Hulk]

Think about these IPC numbers for CB R23.
Metrics expressed in CB R23 points/GHz

265 - Gracemont
363 - Skymont (best info to date)
389 - Raptor Cove no HT
442 - Lion Cove (best info to date)
512 - Raptor Cove with HT running MT bench

Raptor Cove with HT is 16% better than Lion Cove.
Skymont is 37% better than Gracemont

If Raptor is running 5.7 all-core and 4.4 for Gracemont and if ARL is also running 4.4 Skymont, then the Lion Coves in ARL only need to run at 4.7GHz to equal Raptor Lake running full tilt nuclear reactor, 350W mode. Score here would be about 42,000, which is quite a feat for Raptor Lake without a custom loop and lots of volts. Point is 42,000 is a tough day-to-day bar to reach with Raptor Lake.

4.7/4.4 for ARL on the other hand, if performance estimates are true means we will most likely have a cool running CPU equaling the old top dog in nuclear meltdown mode.

Alternatively ARL should also be able achieve that score running 5/4.2, which should also be an easy air cooled target.

If you've ever tried to run a xx900K up to 42,000 CB R23 you know it's a high performance machine yet also a hot, temperamental one.

I'm kind of excited for ARL if these estimates hold up. I could see running something like 5.3/4.3 and having a really fast, efficient rig.

It's all supposition right now but we shall see...

 

[MoistOintment]

There is 0 option shamefully from either Intel or AMD with more than 8 strong cores on a single die...

And really this would all be moot of only Intel would just make a 12 Lion Cove core die. But for some reaosn they will not. I doubt there is a financial problem stopping them. They are Intel afterall and have massive amount of money and resources regardless of their currrent setback.

Intel has 2 desktop dies. It's imply not worth the cost to spin up a 3rd die with 12P cores just so it can lose in nT workloads to their 8+16 die, for the small subset of users who specifically want 12 P cores.

 

[Wolverine2349]

Intel has 2 desktop dies. It's imply not worth the cost to spin up a 3rd die with 12P cores just so it can lose in nT workloads to their 8+16 die, for the small subset of users who specifically want 12 P cores.

They have one desktop die the 8 + 16 one for Raptor Lake and upcoming Arrow Lake. Where is the 2nd die.

They had an 8+8 and 6 + 0 for Alder Lake. Have not seen any other deskotp dies rumored or around for others.

Where is the 2nd desktop die for REPL and ARL?

 

[sgs_x86]

They have one desktop die the 8 + 16 one for Raptor Lake and upcoming Arrow Lake. Where is the 2nd die.

They had an 8+8 and 6 + 0 for Alder Lake. Have not seen any other deskotp dies rumored or around for others.

Where is the 2nd desktop die for REPL and ARL?

RPL uses the lower end bins of the ADL 8+8 and 6+0 dies for its locked i5s and all i3s.
As for ARL, there is a 6+8 die made on 20A that serves the lower end skus. Launches Q1 2025.

 

[Wolverine2349]

RPL uses the lower end bins of the ADL 8+8 and 6+0 dies for its locked i5s and all i3s.
As for ARL, there is a 6+8 die made on 20A that serves the lower end skus. Launches Q1 2025.

20A has been scrapped and eliminated by Intel.

RPL and ADL, well the lower bins are really just ALD dies branded RPL so almost false marketing by Intel.

The only real Raptor Lake die is the 8 + 16 one.

Intel announces cancellation of 20A process node for Arrow Lake, goes with external nodes instead, likely TSMC [Updated]

Out with the Intel node, in with TSMC.

www.tomshardware.com

 

[511]

Finally Intel made an actual good decision

https://videocardz.com/newz/next-gen-z890-motherboards-to-ship-with-intel-default-profile-enabled-by-default

 

[9949asd]

Finally Intel made an actual good decision

https://videocardz.com/newz/next-gen-z890-motherboards-to-ship-with-intel-default-profile-enabled-by-default

That auto oc at stock bios was started at 4790k I think, finally back to the line. lol

 

[MoistOintment]

20A has been scrapped and eliminated by Intel.

RPL and ADL, well the lower bins are really just ALD dies branded RPL so almost false marketing by Intel.

The only real Raptor Lake die is the 8 + 16 one.

Intel announces cancellation of 20A process node for Arrow Lake, goes with external nodes instead, likely TSMC [Updated]

Out with the Intel node, in with TSMC.

www.tomshardware.com

ARL leaked SLUs show 4+4 Ultra 3 and 6+4 Ultra 5 non-K. That's implying a return to their 2 desktop die cadence.

The fact that they only had 1 RPL desktop die further reiterates my point.

 

[511]

ARL leaked SLUs show 4+4 Ultra 3 and 6+4 Ultra 5 non-K. That's implying a return to their 2 desktop die cadence.

The fact that they only had 1 RPL desktop die further reiterates my point.

RPL Was a product launched in hurry lol it was not as well planned as Intels other products

 

[whoshere]

9200MT DDR5 RAM!

 

[MarkPost]

This means 5600 is the officially supported memory speed?

 

[MoistOintment]

RPL Was a product launched in hurry lol it was not as well planned as Intels other products

I'm also aware of RPL's rushed development. It still stands that Intel isn't going to create 3 ARL desktop dies with one of them being a low volume niche that's worse in almost every metric except it may be more stable to run a hypervisor on.

 

[ikjadoon]

This means 5600 is the officially supported memory speed?

That is what I see, which would be no improvement for JEDEC max speeds in two years.

That'd be a shame. Perhaps the IMC binning did not go so favorably at the end here.

Or perhaps CAMM2 motherboards will get higher JEDEC speeds?

EDIT: and maybe 2-DIMM-slot motherboards, too? I notice ASRock correctly labeled its 4-DIMM slot Z690 motherboards as 4400 and not 4800.

 

[9949asd]

9200MT DDR5 RAM!

You need to buy a new cudimm xmp9200, hope the price will not be sky high.

 

[igor_kavinski]

CUDIMM Standard Set to Make Desktop Memory a Bit Smarter and a Lot More Robust

www.anandtech.com

Did something go very wrong somewhere? So Arrow Lake IMC can't support DDR5-6400 at 1.1V and it will instead be considered overclocking???

 

[Wolverine2349]

I'm also aware of RPL's rushed development. It still stands that Intel isn't going to create 3 ARL desktop dies with one of them being a low volume niche that's worse in almost every metric except it may be more stable to run a hypervisor on.

Would not be that low volume. Bit.Little hybrid has scheduling quirks and is worse for gaming than more P cores. Its got nothing to do with hypervisors. It would be better for gaming be smoother with no scheduling quirks and many other things as well including more stable and snappy.
The DIY market is not as niche as you think. And DIY market is heavy with gamers. If DIY market was so darn niche, why has Micro Center which caters specifically to DIY market expanding to Indianapolis, Charlotte, and Miami??

Come on Intel make a 12 + 0 Arrow Lake die You have a buyer in me, I and many others in this somewhat niche will pay top dollar as in double the price for such a CPU. Its worth the investment for that. Better IMC being more robust and stable probability as well.

 

[511]

Would not be that low volume. Bit.Little hybrid has scheduling quirks and is worse for gaming than more P cores. Its got nothing to do with hypervisors. It would be better for gaming be smoother with no scheduling quirks and many other things as well including more stable and snappy.
The DIY market is not as niche as you think. And DIY market is heavy with gamers. If DIY market was so darn niche, why has Micro Center which caters specifically to DIY market expanding to Indianapolis, Charlotte, and Miami??

Come on Intel make a 12 + 0 Arrow Lake die You have a buyer in me, I and many others in this somewhat niche will pay top dollar as in double the price for such a CPU. Its worth the investment for that. Better IMC being more robust and stable probability as well.

And enable AVX-512 on those P core dies 🤣

 

[Wolverine2349]

And enable AVX-512 on those P core dies 🤣

Yes thats another thing too worthwhile especially if you do PS5 emulation. Intel should do it. They are losing lots of buyers as such. They have a good opportunity to capitalize especially with AMD Zen 5 blunder and AMD's terrible latency in general.

People are turned off by RPL stability issues and many hate scheduling quirks of Big.Little. With Arrow Lake, you can attract many re buyers those who hate Big.Little with a 12 + 0 die. Otherwise many will just go X3D route and be stuck with only 8 cores where as with 12 + 0 die you would put a dent in X3D sales big time for gamers how also want thread heavy games and no scheduling stupidity of BIg.LIttle gimmicks.

 

[cannedlake240]

hate Big.Little

Big little is probably going to after NVL or the one after that. Rumor is that Intel will either cancel the Atom line and leaving only the P core or will create a unified P core architecture from Core and Atom cores. Apparently Royal core was also aimed to eventually replace both P, E cores, before it was canned

 

[MS_AT]

Yes thats another thing too worthwhile especially if you do PS5 emulation. Intel should do it. They are losing lots of buyers as such. They have a good opportunity to capitalize especially with AMD Zen 5 blunder and AMD's terrible latency in general.

People are turned off by RPL stability issues and many hate scheduling quirks of Big.Little. With Arrow Lake, you can attract many re buyers those who hate Big.Little with a 12 + 0 die. Otherwise many will just go X3D route and be stuck with only 8 cores where as with 12 + 0 die you would put a dent in X3D sales big time for gamers how also want thread heavy games and no scheduling stupidity of BIg.LIttle gimmicks.

Until current console generation dies, then 12 + 0 die is a gimmick for gaming, really. You would better spend the additional die space on bigger caches so they could catch up to x3d. By the time more than 8 cores will be a must in gaming your 12 + 0 ArrowLake will be obsolete. Unless of course you need more cores for background stuff but then Skymont will serve you well, won't it?

 

[Wolverine2349]

Until current console generation dies, then 12 + 0 die is a gimmick for gaming, really. You would better spend the additional die space on bigger caches so they could catch up to x3d. By the time more than 8 cores will be a must in gaming your 12 + 0 ArrowLake will be obsolete. Unless of course you need more cores for background stuff but then Skymont will serve you well, won't it?

Not sure about that. More than 8 cores is not a must for gaming. But neither is more than 6. Some games start to stutter with 6 but 6 will run any game. More than 8 cores can help with some games today. There is a difference between beneficial and a must.

I suppose part of my hope is maybe Big.Little will actually not have scheudling issues this time since Skymont is supposed to be so good this time unlike Gracemont. The gap between IPC of Skymont and Lion Cove is so much closer than Gracemont and Golden/Raptor Cove that maybe all scheduling quirks of Big.Little will be fixed? AM I right on that?

Wasn't part of reason Big.Little caused scheduling issues with some games is because some games got caught on a Gracemont core and Gracemont had such much lower IPC and latency compared to Holden/Raptor Cove and game performance tanked as such. But with Skymont they should handle games as well as a 4.7GHz Golden Cove so if one gets on a core no big deal? Or not really and Big.Little has other scheduling problems besides just that. Or will the better thread director fix that as well.

 

[cebri1]

Not sure about that. More than 8 cores is not a must for gaming. But neither is more than 6. Some games start to stutter with 6 but 6 will run any game. More than 8 cores can help with some games today. There is a difference between beneficial and a must.

I suppose part of my hope is maybe Big.Little will actually not have scheudling issues this time since Skymont is supposed to be so good this time unlike Gracemont. The gap between IPC of Skymont and Lion Cove is so much closer than Gracemont and Golden/Raptor Cove that maybe all scheduling quirks of Big.Little will be fixed? AM I right on that?

Wasn't part of reason Big.Little caused scheduling issues with some games is because some games got caught on a Gracemont core and Gracemont had such much lower IPC and latency compared to Holden/Raptor Cove and game performance tanked as such. But with Skymont they should handle games as well as a 4.7GHz Golden Cove so if one gets on a core no big deal? Or not really and Big.Little has other scheduling problems besides just that. Or will the better thread director fix that as well.

I’m just realizing my 10700K is slower than those tiny e-cores