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	8 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)	Arrow Lake Refresh (N3B)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Desktop Only	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2025 ?	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	8P + 32E	4P + 4E	4P + 8E
LLC	24 MB	24 MB ?	36 MB ?	?	8 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.)

 

[BorisTheBlade82]

Tomorrow is Christmas, so to all of you who celebrate Christmas, I wish you a merry Christmas.
To all of you who don't celebrate It, I wish you a nice weekend.

Thanks, to you and the bunch around here as well 😊

 

[LightningZ71]

Ouch we need indeed a GT3 for 128EUs. There is no GT1 by the looks of it though, so we can still hope GT3 will be the new standard for mobile rather than GT2 because GT2 seems to be the new default for desktop instead of GT1.


INTEL_PLATFORM_MTL_P:
if (intel_device_info_eu_total(devinfo) <= 64)
return intel_oa_register_queries_mtlgt2;
if (intel_device_info_eu_total(devinfo) <= 128)
return intel_oa_register_queries_mtlgt3;

If they can get the clocks up to decent levels, they should be able to achieve performance that's roughly on par with the 80EU configurations in Tiger Lake. That's not a particularly bad place to be as it is usable at 1080p/low in a lot of games, which is enough for a good sized chunk of the market. Compared to what was standard as recently as Comet Lake, that's a substantial improvement. Compared to Rocket Lake, Ice Lake and Alder Lake-S, it's still a step up.

 

[dangerman1337]

Looks like MTL Desktop may be cancelled after all:

https://twitter.com/x/status/1606342626654896129

Arrow Lake SMID increased to 1024 bits from the looks of it:

https://twitter.com/x/status/1606344296126246918

 

[Henry swagger]

I liked some of Geddagod analysis of leaks, their history and track record. Which is a good background info.

(Also, his other post where he speculated on timelines, and when certain decision would have to happen that would result in current state of product development / schedule) was a good analysis.

I think Henry swagger called Tom at MLID a "basement dwelling fraud" (not me). At least that's how I read it. His recording studio looks a little bit like a basement or attic.

The way I look at leaks and leakers (and value them) is when they provide some info, when there was none before.

Good example of that would be MLID video on Mi300 datacenter GPU.

For the rest, for information / leaks / analysis, from multiple sources - they provide range of expectations.

Then there is a whole category of rumors, rather than leaks, of, say Company X preparing Chip Y in certain time frame that will have certain characteristics - rumors that are not on any roadmaps yet. So, something to look forward to in the future, but very low likelihood that we will get from here to that product on a straight line.

So, I am not really "mad" when at any of the leakers, analysts who miss the target. Or would expect the past track record predict the future track record (as far as leaks), because you get what you get (from sources). As far as analysis, yeah, the track record matters more there.

Another point: One way to get the right answer is to present the wrong answer. And if they have good contacts in the industry, they may get someone to contact them to correct them.

I was calling mlid a fraud not you joe 😄

 

[Saylick]

Looks like MTL Desktop may be cancelled after all:

https://twitter.com/x/status/1606344296126246918

Arrow Lake SMID increased to 1024 bits from the looks of it:

https://twitter.com/x/status/1606344296126246918

You quoted the same Tweet twice, but I got you fam.

https://twitter.com/OneRaichu/status/1606342626654896129?s=20&t=Q4oJ4RUSczsWSAWHb2wduA

 

[trivik12]

Raichu did predict this a while back about this. Seems more like ice lake or hopefully tiger lake with at least gaming laptops.

 

[BorisTheBlade82]

Looks like MTL Desktop may be cancelled after all:

https://twitter.com/x/status/1606344296126246918

Arrow Lake SMID increased to 1024 bits from the looks of it:

https://twitter.com/x/status/1606344296126246918

8-wide and 700+ ROB indeed sounds impressive - finally on par or slightly better than M1.

 

[dangerman1337]

You quoted the same Tweet twice, but I got you fam.

https://twitter.com/OneRaichu/status/1606342626654896129?s=20&t=Q4oJ4RUSczsWSAWHb2wduA

Yeah edited it.

I mean if MTL-S can't be out by the end of 2023 then it has almost no reason to exist if ARL-S on TSMC N3 can be pushed out ASAP.

 

[poke01]

8-wide and 700+ ROB indeed sounds impressive - finally on par or slightly better than M1.

that's not impressive since Arrow lake is a 2024 product. So 4 years later Intel does what Apple did in 2020. Those chip designers that made the M1 and prior chips are industry leading engineers then. Now almost all those M1 people went to Qualcomm, it looks 2024 is going to be very interesting in the PC market.

 

[Carfax83]

Arrow Lake SMID increased to 1024 bits from the looks of it:

https://twitter.com/x/status/1606344296126246918

My next upgrade! My 13900KF will serve me until then

 

[Exist50]

Looks like MTL Desktop may be cancelled after all:

https://twitter.com/x/status/1606342626654896129

Arrow Lake SMID increased to 1024 bits from the looks of it:

https://twitter.com/x/status/1606344296126246918

Huh, Raichu is usually pretty on point, but he's way off base here.

 

[Exist50]

Nope MLID is just a <redacted> leaker.
Check out my (very generous) estimation of MLID leak accuracy for Intel at 66%. Though more accurately it should be <60% after I make some corrections.
This is in comparison of other leakers, like Raichu, who have 90% accuracies from the who's who of leaks.
While obviously both of these estimations are unprofessional and bound to have mistakes, the general point is that MLID is NOT up to par with other leakers.
What's even worse is that he refuses to admit his mistakes, deletes videos where he got leaks wrong, and has a humongous ego.
Plus he rips off other leakers leaks, his most recent MTL/ARL leaks from his 2 most recent leak videos about those products are a great example of that:

• ARL P uses 20A while ARL S uses TSMC 3nm - ripped off Raichu
• MTL might have a limited desktop presence and coincide with a quick launch of ARL - ripped off Raichu
• MTL low end ARL high end - ripped off Kopite

Hell he was ripping off Raichu even before he started leaking these 2 next generations, with the information that ADL might be able to compete with the 5950x and not just the 5900x....
His leaks are bad, his attitude is bad, and he is constantly puts out super optimistic stuff for both Intel AND AMD and then claims "delays" or "failed execution" caused massive problems, or that he knew it all along:

• Zen 3 launch date (this might have an inkling of truth in it due to covid)
• Zen 4 IPC figures
• RDNA 3 clock speeds
• Idk what his excuse is going to be for Zen 5 but he also claimed that AMD is going to push that out at the end of 2023 when AMD confirmed it's a 2024 product (this one isn't that bad though tbh, time frame is pretty close)
• His recent MTL leak
• His entire ARC drama and the mythical A780

Also it's not just @Henry swagger who thinks MLID is pretty bad, a lot of people do too. There's a reason his leaks get laughed out of the room in many tech discussion forums outside youtube and twitter.
I'm also pretty sure Exist50 thinks MLID is a bad leaker too, but I doubt you're willing to confront him the same way you did Henry swagger considering how Exist50 demolished your Sapphire Rapids vs Genoa power draw estimations in the Intel Current and Future Lakes thread around a month ago. Funny you said he was trolling you too....
I just think it's ironic that you accuse other people (including me at one point, I think, on one of my first couple of posts ever in the Anandtech Forums) of trolling.

You could hit the character limit describing his mispredictions with Meteor Lake alone. To add on a few more, first he claimed it was using Ocean Cove (despite Ocean Cove being dead when he made that claim), then he insisted he actually knew it was Redwood Cove. Obviously he's flip flopped several times on MTL desktop, but he still seems to be insisting that Redwood Cove will be ~15-20% IPC, which is hilarious.

 

[Exist50]

At the end of July 2021, Intel announced Meteor Lake going up to the 125 watt power range. This is the main justification that Tom, and many others, use that Intel had always planned for a MTL-S lineup. And while that is true, we have to think about the time frame of these events.

Keep in mind, that slide was incorrect the day it was presented. Heard a lot of griping about that. Nor would it be the first time Intel marketing presented fundamentally wrong product details.

 

[Carfax83]

that's not impressive since Arrow lake is a 2024 product. So 4 years later Intel does what Apple did in 2020. Those chip designers that made the M1 and prior chips are industry leading engineers then. Now almost all those M1 people went to Qualcomm, it looks 2024 is going to be very interesting in the PC market.

Arrow Lake is likely going to be at least 5ghz though. It's one thing to get a super wide core to run at 3.2ghz or whatever the M1/M2 run at, and another to get it to run at 5ghz and above. And then you have the super wide vectors. It's going to be a beast!

 

[poke01]

Huh, Raichu is usually pretty on point, but he's way off base here.

How do you know he's off base?

 

[poke01]

Arrow Lake is likely going to be at least 5ghz though. It's one thing to get a super wide core to run at 3.2ghz or whatever the M1/M2 run at, and another to get it to run at 5ghz and above. And then you have the super wide vectors. It's going to be a beast!

You cannot have super wide cores running at high clock speeds. If you do you lose efficiency but its Intel and I have no faith in them when it comes to efficiency.

 

[Carfax83]

You cannot have super wide cores running at high clock speeds. If you do you lose efficiency but its Intel and I have no faith in them when it comes to efficiency.

Raptor Lake is pretty damn wide and it will be hitting 6ghz next month. As far as efficiency, from my experience Raptor Lake is fairly power efficient when manually tuned, which to me suggests the main problem is that the factory settings having unlimited power draw and ridiculously high voltages.

 

[poke01]

Raptor Lake is pretty damn wide and it will be hitting 6ghz next month. As far as efficiency, from my experience Raptor Lake is fairly power efficient when manually tuned, which to me suggests the main problem is that the factory settings having unlimited power draw and ridiculously high voltages.

thats good to hear

 

[Glo.]

Looks like MTL Desktop may be cancelled after all:

https://twitter.com/x/status/1606342626654896129

Arrow Lake SMID increased to 1024 bits from the looks of it:

https://twitter.com/x/status/1606344296126246918

1024 Bit SIMD and 8-Wide will result in 3000 pts in Geekbench 5 ST, with today's even moderate boost clock speeds.

 

[Hulk]

Raptor Lake is pretty damn wide and it will be hitting 6ghz next month. As far as efficiency, from my experience Raptor Lake is fairly power efficient when manually tuned, which to me suggests the main problem is that the factory settings having unlimited power draw and ridiculously high voltages.

While I can appreciate the achievement of getting a production CPU to 6GHz if this is just a 2 core turbo 6GHz then it is useless outside of advertising. In reality there are no applications that benefit from 1 or 2 cores boosting ONLY when all other cores are not active. Without actually shutting down the other cores in the BIOS it's almost impossible to hit 5.8GHz on the 13900K I have found.

Now with that being said I would assume that if these super binned parts can hit 6GHz then perhaps they can do lower frequencies at lower voltage and power than the 13900K. We shall see.

 

[igor_kavinski]

1024 Bit SIMD and 8-Wide will result in 3000 pts in Geekbench 5 ST, with today's even moderate boost clock speeds.

Oh so now that AMD has AVX512, Intel is moving on and their literature is going to point out all the limitations of AVX512 and how their 1024-bit vector instructions will solve everything and make the world a better place. They will conveniently leave out the fact that the CPU will get even hotter and downclock even more aggressively.

 

[turtile]

that's not impressive since Arrow lake is a 2024 product. So 4 years later Intel does what Apple did in 2020. Those chip designers that made the M1 and prior chips are industry leading engineers then. Now almost all those M1 people went to Qualcomm, it looks 2024 is going to be very interesting in the PC market.

I don't think making a core wide really makes it impressive. Apple can afford to use a lot of die area because they automatically take a profit from the device sale. If they can hit almost twice the clock speed of M1, that will be an impressive design.

 

[Carfax83]

While I can appreciate the achievement of getting a production CPU to 6GHz if this is just a 2 core turbo 6GHz then it is useless outside of advertising. In reality there are no applications that benefit from 1 or 2 cores boosting ONLY when all other cores are not active. Without actually shutting down the other cores in the BIOS it's almost impossible to hit 5.8GHz on the 13900K I have found.

It's probably going to be TVB and single core only if I had to guess. Hitting 6ghz on more than one core is a lot to ask for. All core boost might go up by 100mhz or more, from 5.5ghz to 5.6/5.7ghz.

Now with that being said I would assume that if these super binned parts can hit 6GHz then perhaps they can do lower frequencies at lower voltage and power than the 13900K. We shall see.

Perhaps. But out of the box power consumption and thermals is likely to be high no matter what without manual tuning. The auto voltages on these motherboards pump too much voltage into the CPU.

 

[igor_kavinski]

Supposing AVX1024 rumor is true, does this mean ARL will be P-cores only or they gonna beef up the E-cores with AVX1024 too? If the E-cores don't show up in ARL, it will confirm that hybrid cores was just a stopgap solution to have a fighting chance against AMD.

 

[moinmoin]

Supposing AVX1024 rumor is true, does this mean ARL will be P-cores only or they gonna beef up the E-cores with AVX1024 too? If the E-cores don't show up in ARL, it will confirm that hybrid cores was just a stopgap solution to have a fighting chance against AMD.

I think it's obvious Intel wanted there to be a solution making P core with and E cores without AVX-512 at the same time possible, didn't manage to make it work in time and delayed the real solution. ARL may well be their second go at it.

A good solution (and positive surprise) would be Intel reworking AVX to finally work more like SVE with the rumored introduction of AVX-1024, allowing E cores to handle all instruction even without native 1024 bit support.