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.)

 

[DavidC1]

There is no Panther lake desktop, and it arrives later than Clearwater. Desktop is ARL-R with a larger NPU. Clearwater is the lead product for 18A, just as iphone chips are for leading edge TSMC nodes. If your lead product is outsourced all of a sudden, it means the node is completely unusable, and Intel failed to yield Ribbonfet and Powervia

Clearwater Forest is lead Xeon on 18A, not lead 18A. It was Sierra Forest that was lead Intel 3.

It seems like on Computex they were implying on PTL first, but since client products launch end of year anyway, not sure what to make of this. They probably come roughly the same time.

Yeah it’s a full Intel chip. I think he confused FS and CWF

Intel's next-gen Falcon Shores AI chip: ordered on TSMC 3nm node, CoWoS advanced packaging

Intel reportedly placed orders for its next-generation AI chip codenamed 'Falcon Shores' with TSMC, using its CoWoS advanced packaging, 3nm process node.

www.tweaktown.com

Falcon shores being on N3 is a big fail too:

Original Falcon Shores

 

[Wolverine2349]

In a world where 9800X3D is not limited to 5Ghz boost clock. 5.4Ghz should be easy to achieve and there you go, 13% faster than 7800X3D.

How will that happen when the 3D V cache stack is so voltage and heat sensitive. No way it will be able to clock much higher than 7800X3D.

AMD promised the same thing with 7800X3D over 5800X3D and it was multiplier locked when they stated way before hand it would be unlocked. And it is as clock limited compared to vanilla Zen 4 as 5800X3D is limited compared to vanilla Zen 3.

And since vanilla Zen 5 is barley any faster than vanilla Zen 4 and way behind 7800X3D , hard to imagine 9800X3D will be much if any faster than 7800X3D with a clock bump that will not exist as per reasons and evidence above.

 

[Hulk]

Here are the stats of my 14900K doing what is for me a heavy lift. Rendering a video project from Vegas Video 21 using Happy Otter Scripts Render+. This basically means the "assembled" timeline is frameserved to what is essentially the Handbrake encoding engine.

I have volts manually at 1.3 and with droop during the load it's at about 1.06V.
HT is off.
All 24 cores are slammed.
About 5.3/4.2 average.
Max watts is set to 200 and max temp (on air) is 81C.
This is how I run my rig day-to-day, sometimes rendering for hours with complete stability.
This is my third Raptor Lake. 13900K degraded in 3 months on "Auto" BIOS settings.
Then a 13600K to hold me over during the Intel warranty refund process. Ran that one on "Auto" but stock frequencies are reasonable so no issues.
Finally the 14900K using 200W cap, no HT, manual Vcore of 1.3V and LLC 4.
It's all good now but there was obviously a learning curve.

 

[Thunder 57]

How will that happen when the 3D V cache stack is so heat sensitive. No way it will be able to clock much higher than 7800X3D.

AMD promised the same thing with 7800X3D over 5800X3D and it was multipler locked when they stated way before hand it owuld be unlocked. And it is as clock limited compared to vanilla Zen 4 as 5800X3D is limited compared to vanilla Zen 3.

And since vanilla Zen 5 is barley any faster than vanilla Zen 4 and way behind 7800X3D , hard to imagine 9800X3D will be much if any faster than 7800X3D with a clock bump that will not exist as per reasons and evidence above.

This post won't age well. It is not heat sensitive but voltage sensitive. Just wait for Bartlett Lake since that is your dream CPU.

 

[igor_kavinski]

View attachment 107237

You have kind of a golden sample, at least as far as your Raptormont clusters are concerned. 4.2 GHz all core and boosts of 4.4! My lousy single Gracemont cluster gets unstable if I force it to run at 4.1 GHz all core.

 

[cannedlake240]

lead Xeon on 18A

Pretty sure they said it was lead 18A vehicle... Still Clearwater will be first to market it seems. It's tile is both smaller and simpler. PTL houses CPU, NPU, memory controller, media etc. Clearwater tile only has the 6x 4C Darkmont + L2. L3, interconnect are all on the Intel 3 base tile.

Falcon shores being on N3 is a big fail too

That's true. Maybe it's because Falcon launches in 2026 1H, but 18A-P products only start appearing by end of year. Something about the node not being "optimized" for GPUs... Their Technology Development lead also hinted in an interview that up to 18A, it was traditional development with CPU being the main target. Only starting from 'Intel next', or now known as 14A, they will be focusing on other types of product.

PTL tGPU using N3E is a clear indication that 18A is behind in that aspect. Wildcat lake apparently has an 18A GPU, but that's low end. If they wanted to save as much money by going full in house production, why wouldn't they use 18A for the tGPU too? Or at least by using both IFS and external to manage risk

 

[dullard]

PTL tGPU using N3E is a clear indication that 18A is behind in that aspect. Wildcat lake apparently has an 18A GPU, but that's low end. If they wanted to save as much money by going full in house production, why wouldn't they use 18A for the tGPU too? Or at least by using both IFS and external to manage risk

Intel doesn't have the EUV capacity--yet.

 

[cannedlake240]

Intel doesn't have the EUV capacity--yet.

Nah they probably bought up plenty. Nowadays they even have a high NA machine. Foundry R&D center is a massive fab space, production can start there and eventually transfered to the newly built ones that should be ready by then, even factoring delays, as long as Intel isn't collapsing. They did the same with MTL, SRF. Plus Isn't it better to have the existing fabs fully utilized?

 

[9949asd]

Memory latency may negate whatever IPC uplift ARL has. It may need DDR5-8400 or higher to realize decent gains over 14900KS.

CUDIMMs may save ARL-S, if DDR5-9200 becomes available before year end and actually possible on 285K.

There are already two vendors announced cudimm xmp9200, and it’s for z890 run at gear2. I think manual OC can get more, but for the gaming performance improvement is unknown.

 

[dullard]

Nah they probably bought up plenty. Nowadays they even have a high NA machine.

1) They aren't using the high-NA machine for plain jane 18A. That is being used for 14A development.

2) They didn't buy up plenty of machines. That is ultimately the biggest mistake Intel made. They skipped buying EUV when it first came out and when they finally wanted some equipment, they were at the back of the long line. It is why they did overly complex multi-patterning that delayed Intel 7 and Intel 4. It is also a big reason why they are using TSMC. Intel simply doesn't have the capacity on their equipment. It is maxed out. Intel said so themselves. Maybe they'll have sufficient EUV capacity in 2026 to do all they want. But not now.

 

[alcoholbob]

If these leaks are legit it's gonna be Zen 5% vs NArrow Gains Lake in 2024...

At least AMD doesn't require a platform update. Hopefully Intel doesn't pull a fast one and require a new socket again for Nova Lake...

 

[Josh128]

This post won't age well. It is not heat sensitive but voltage sensitive. Just wait for Bartlett Lake since that is your dream CPU.

Its likely related to heat as well, as Zen 4 X3Ds are thermally limited to 89C where non X3D can go to 95 and run there all day.

Regardless, I would take Donny Woligroski's "cool differentiators" comment about Zen 5 X3D with truckload of salt before assuming that means its going to hit or come close to hitting non-X3D frequencies. The same guy lied through his teeth and said the following couple gems of quotes:

"Our cores are so powerful we dont need 105W TDP" - speaking about AMDs reasoning behind 9700X's 65W TDP.

...and this doozy of a quote:

"Is it the fastest in gaming? It's faster than the competition in our tests. X3D is still the king of the hill, but by a much smaller margin than typically between X3D and non-X3D" - speaking about Zen 5 vs Zen 4 X3D.


I dont guess I need to remind anybody that 7600X, let alone the 7700X is actually faster on average than 5800X3D according to HWUB's 50+ game shootout. That absolutely negates what Donny claims above. With AMDs recent flat out lies from marketing and him being the head of that marketing, I learned very quickly that he is full of shat. I could see Zen 5 squeezing out 200 or maybe even 300MHz vs 7800X3D, but unless they made some miraculous discovery since the design of Zen 4 X3D, 5.4 or 5.5 isnt happening. Regarding voltage, is there any evidence that Zen 5 cores require less voltage than Zen 4 to hit a given boost? If so, I havent seen it.

 

[dullard]

Hopefully Intel doesn't pull a fast one and require a new socket again for Nova Lake...

Current rumors are for Nova to share the same socket. But, who knows. LGA 1700 ends up being used with 4 generations once you add in Bartlett Lake. Rumors are for similar lifespan of LGA 1851.

Intel's Next Consumer Desktop Platform Will Be Nova Lake-S, Panther Lake Reportedly Mobile Only

New rumor suggests that Intel's Nova Lake-S CPUs will be the desktop follow-up to Arrow Lake-S while Panther Lake will be mobile only.

wccftech.com

 

[OriAr]

Current rumors are for Nova to share the same socket. But, who knows. LGA 1700 ends up being used with 4 generations once you add in Bartlett Lake. Rumors are for similar lifespan of LGA 1851.

Intel's Next Consumer Desktop Platform Will Be Nova Lake-S, Panther Lake Reportedly Mobile Only

New rumor suggests that Intel's Nova Lake-S CPUs will be the desktop follow-up to Arrow Lake-S while Panther Lake will be mobile only.

wccftech.com

Nova Lake is supposedly a new socket, it'll also have beyond just a new compute tile, which signifies a pretty big platform upgrade.
(Unrelated, I bet 285K does 46-48K on CB23, and if there is a KS variant it'll do 50K).

 

[alcoholbob]

Nova Lake is supposedly a new socket, it'll also have beyond just a new compute tile, which signifies a pretty big platform upgrade.
(Unrelated, I bet 285K does 46-48K on CB23, and if there is a KS variant it'll do 50K).

On how many watts? Even a 9950X can go nuclear if you turn on PBO and remove the power limits

 

[OriAr]

On how many watts? Even a 9950X can go nuclear if you turn on PBO and remove the power limits

At its stock boosts (5.4 all core for the P cores, 4.6 all core for the E cores)
KS variant will probably do something like 5.6 all core for the P cores and 4.8 for the E cores. (I can see Intel going 5 all core for the E cores and 5.8 all core for the P cores if they really want to make it a super halo SKU and for the marketing)

 

[alcoholbob]

At its stock boosts (5.4 all core for the P cores, 4.6 all core for the E cores)
KS variant will probably do something like 5.6 all core for the P cores and 4.8 for the E cores. (I can see Intel going 5 all core for the E cores and 5.8 all core for the P cores if they really want to make it a super halo SKU and for the marketing)

You would need way over 20% IPC increase on the P-cores to achieve that without hyperthreading

 

[OriAr]

You would need way over 20% IPC increase on the P-cores to achieve that without hyperthreading

HT yields on Intel are usually in the 20-25% range, not ~30% like it is for AMD.
And Skymont is gonna be a huge increase in perf in CB which will more than make up for the lack of HT.

 

[9949asd]

On how many watts? Even a 9950X can go nuclear if you turn on PBO and remove the power limits

285k stock r23 45k at 220w

 

[9949asd]

You would need way over 20% IPC increase on the P-cores to achieve that without hyperthreading

You don’t need P core over 20% ipc, because the e core get 50% ipc uplift and there 16 of them lol.

 

[ondma]

In which world will ZEN5 X3D be 10-15% faster than ZEN4 X3D? ZEN5 is about 5% faster than ZEN4 in Gaming with same RAM. All Tests that show more is 5200MT vs 5600MT. So 5% over 7800X3D is what we should expect.

I agree, but I think AMD stated that 9xxx X3D has some specific improvements vs previous gen X3D chips. Perhaps they will be able to come closer to the clock speed of the vanilla chips, or make it overclockable? Perhaps the extra cache will mitigate whatever is seemingly holding back vanilla Zen 5 gaming performance. Bottom line is, it is way too soon to say what kind of performance 9xxx X3D will show.

 

[ondma]

At its stock boosts (5.4 all core for the P cores, 4.6 all core for the E cores)
KS variant will probably do something like 5.6 all core for the P cores and 4.8 for the E cores. (I can see Intel going 5 all core for the E cores and 5.8 all core for the P cores if they really want to make it a super halo SKU and for the marketing)

Didn't they learn anything about pushing their chips too far from the Raptor Lake stability debacle? They might be able to push the P cores to 5.8, but I think 5.0 for the E cores is totally out of sight.

 

[MarkPost]

HT yields on Intel are usually in the 20-25% range, not ~30% like it is for AMD.
And Skymont is gonna be a huge increase in perf in CB which will more than make up for the lack of HT.

Actually its ~35% on Intel. I tested it some time ago with my 13900K:

P-CORE RPL (5.0)
1T 1985
8T 15531 (7.82x)
8T+8(HT) 21005 (10.59x)

P-CORE RPL (5.4)
1T 2139
8T 16769 (7.84x)
8T+8(HT) 22626 (10.56x)

These are facts.

And about the rest of your really optimistic especulation, I dont think so, but we will see

 

[Hulk]

I wonder how what percentage of people who buy thin and light laptops get them with 64GB of RAM?

Why I won't buy an Intel Lunar Lake-powered laptop

Intel’s new mobile chips promise better performance and longer battery life, but they snub power users like me.

www.tomshardware.com

 

[poke01]

I wonder how what percentage of people who buy thin and light laptops get them with 64GB of RAM?

Why I won't buy an Intel Lunar Lake-powered laptop

Intel’s new mobile chips promise better performance and longer battery life, but they snub power users like me.

www.tomshardware.com

I could count the %age on my hand. clickbait article