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.)

 

[Goop_reformed]

The N3B is not the best node. And SRAM scaling is very low there. Intel could use either one of the nodes I mentioned, even if it is a cost of a percentage decrease in performance or power.

Capacity (especially EUV capacity) may be an issue Intel is facing, but TSMC slowed down delivery of some equipment, so Intel could grab some of that, if the capacity was the issue.

Or, it could be (as Hillary Clinton famously said) a "public" position on an issue and "private" position.

Public being the one Gelsinger uses as a pep-talk (to customers, Wall Street, employees) and private position is what Intel really believes about readiness and competitiveness of the new nodes.

>Public being the one Gelsinger uses as a pep-talk (to customers, Wall Street, employees) and private position is what Intel really believes about readiness and competitiveness of the new nodes.

I agree, Intel would definitely fab granites at TSMC if foundry business is a separate entity.

 

[FlameTail]

The difference between N4P and N3B/N3E is small. I can’t find TSMC marketing material that compares N4P vs N3B or N3E directly. It’s like ~5-8% performance improvement with better density. If Intel 3 matches or exceeds N4P performance then practically speaking it’s comparable.

Even factoring in sram density assuming that Intel 3 has no improvements over Intel 4 sram cells then it’s 0.0240 vs 0.0210 (N3E) - it’s not that far off.

 

[H433x0n]

View attachment 89147View attachment 89148

That’s first generation N5.

If you compare N4P vs N5 (Gen 1) you get +11% better perf isopower. That same comparison with N3B is 10-15% better perf at isopower which is nearly identical performance. The comparison with N3E gets +18% better perf at isopower which puts N3E ahead of N4P by 7%.

Generally speaking the performance improvement N3 provides over the latest/greatest N4P is <10%.

 

[ondma]

Uh, I think people don't like linking twitter stuff here, so here:
View attachment 89074
View attachment 89075
View attachment 89076
View attachment 89077View attachment 89078

So ARL is supposed to be 20A and they are using N3B for Lunar Lake? That doesnt make any sense, and seems like a big blow to the foundry aspirations.

 

[igor_kavinski]

So ARL is supposed to be 20A and they are using N3B for Lunar Lake? That doesnt make any sense, and seems like a big blow to the foundry aspirations.

Maybe coz 20A isn't meeting their power efficiency targets? Also, they probably decided to save time and go with a proven power efficient node rather than try to make their frequency optimized node consume less power.

 

[FlameTail]

Maybe coz 20A isn't meeting their power efficiency targets? Also, they probably decided to save time and go with a proven power efficient node rather than try to make their frequency optimized node consume less power.

Yeah I have thought about this. Intel nodes are good at reaching high performance levels, but ghjkfdghji at low performance efficiency.

I guess that's why Qualcomm ditched their plans with IFS.

 

[DrMrLordX]

What happened to 5 nodes in 4 years?

It's easy if it's all the same node (more or less), or maybe two nodes.

So ARL is supposed to be 20A and they are using N3B for Lunar Lake? That doesnt make any sense, and seems like a big blow to the foundry aspirations.

20a isn't one of Intel's foundry nodes. And all indications are that Intel 3 is going to wind up being niche like 22FFL and Intel 16 (as a foundry node). 18a is the proving ground for IFS.

Also now that we have at least semi-confirmation that Intel is still taking N3B wafers, we may also see Arrow Lake compute tiles on TSMC N3B.

 

[dullard]

I'm confused as to why Intel is using TSMC for the compute tile. I very much expected graphics and/or I/O dies to be made by TSMC. I don't get how people are spinning this as being just fine. What happened to 5 nodes in 4 years?

Intel has been saying external nodes will be used. They have been saying that for years. Yes, even compute tiles could be external. Intel has moved to node agnostic designs that can be fabbed elsewhere. Remember when this decision was made, (1) there was a major chip shortage with Intel running at full capacity, (2) the 20A/18A nodes were not yet proven, and (3) Intel was falling further and further behind TSMC. So, Intel took the conservative approach to solve problems #1, #2, and #3 by using external fabs.

Node 1) Intel 7, Alder Lake/Raptor Lake 2021 to 2024
Node 2) Intel 4, Meteor Lake 2023 (Barely, most in 2024)
Node 3) Intel 3, Granite Rapids 2024. Synopsys is an Intel 3 foundry customer: https://www.forbes.com/sites/tirias...l-for-intel-foundry-services/?sh=2efac6a5fa8b
Node 4) Intel 20A, Arrow Lake 2024 (Probably)
TSMC, Lunar Lake
Node 5) Intel 18A, Panther Lake 2025 (goal), Clearwater Forest, many external customers. This is where Intel thinks they'll finally have leadership chips. Meaning Intel will be behind TSMC's capabilities until 2025.

Looks like using 5 different Intel nodes in 3 years to me is a distinct possibility (2023, 2024, 2025). That is true even with using TSMC.

 

[Khato]

Remember when this decision was made, (1) there was a major chip shortage with Intel running at full capacity, (2) the 20A/18A nodes were not yet proven, and (3) Intel was falling further and further behind TSMC. So, Intel took the conservative approach to solve problems #1, #2, and #3 by using external fabs.

Indeed, 2-3 years ago when the decision was made to use N3B Intel may not have had high confidence in 20A/18A.

 

[desrever]

Intel is getting 5 nodes in 4 years cause they are counting TSMC's nodes too.

 

[SiliconFly]

So LNL is using a 4P+4E configuration.

Rather unusual. That wouldn't be very strong in Multithreaded. Are we expecting huge ST gains?

Maybe huge ST gains. Maybe huge efficiency gains. Maybe significantly smaller die. Maybe all three.

Or, Intel being Intel, none of the above!

 

[dullard]

Intel is getting 5 nodes in 4 years cause they are counting TSMC's nodes too.

Really? You had to post thus just after I listed the nodes?
1) Intel 7
2) Intel 4
3) Intel 3
4) Intel 20A
5) TSMC N3
6) Intel 18A

For some reason, my schooling said that 6 is more than 5.

You could make a valid argument that Intel 3 and Intel 18A are only half node steps and that really only makes 3 full nodes. But that isn't the argument you chose to make.

 

[cebri1]

Intel Core Ultra 7 155H Arc iGPU surpasses Radeon 780M, new leaked OpenCL result shows - VideoCardz.com

ASUS laptop with Core Ultra 7 155H and ARC iGPU surpasses Radeon 780M in Geekbench OpenCL test Intel is set to unveil the upcoming Core 100 series in three weeks. The integrated graphics department will receive one of the biggest upgrades. The Core series will soon include Intel Arc GPUs. The...

videocardz.com

 

[uzzi38]

So LNL is using a 4P+4E configuration.

Rather unusual. That wouldn't be very strong in Multithreaded. Are we expecting huge ST gains?

Cores are the same as ARL I'm pretty sure.

LNL is a genuinely very interesting product though, and that's because of everything else.

 

[uzzi38]

So ARL is supposed to be 20A and they are using N3B for Lunar Lake? That doesnt make any sense, and seems like a big blow to the foundry aspirations.

No, ARL is a mix of N3B and 20A depending on the product.

 

[mikk]

Meteor Lake – GPU Architecture

 

[poke01]

Cores are the same as ARL I'm pretty sure.

LNL is a genuinely very interesting product though, and that's because of everything else.

It will be the first Intel product that will be powerful in a fanless laptop. The past ones were crap and underpowered.

Those Memory modules being on package will benefit handhelds and laptops. Especially with 8000+ MT/s.

 

[Abwx]

Meteor Lake – GPU Architecture

Dunno for the actual power used for thoses tests but if frequency is to be increased by 50% and exe units by 33% then the process should provide roughly 0.5x lower power to be at same powers than previous 96EUs iGPUs, that s 100% better perf/Watt at isofrequency and quite doubtfull.

Indeed first models with a battery as big as 68Wh for a mere 14" laptop are not indicative of huge efficency improvements, guess that we re good for another big round of deceptive TDPs, an habit at Intel.

HP Spectre x360: Upcoming Meteor Lake laptop revealed in full by early website listing

HP has accidentally listed its upcoming Spectre x360 laptop ahead of time. It comes in two flavours, Intel Core i7-155H and Intel Core i5-125H, and offers next-gen features such as Wi-Fi 7.

www.notebookcheck.net

 

[igor_kavinski]

HP Spectre x360: Upcoming Meteor Lake laptop revealed in full by early website listing

HP has accidentally listed its upcoming Spectre x360 laptop ahead of time. It comes in two flavours, Intel Core i7-155H and Intel Core i5-125H, and offers next-gen features such as Wi-Fi 7.

www.notebookcheck.net

Huge touchpad. Huge keys. I guess the average consumer has fat fingers nowadays.

 

[tamz_msc]

Indeed first models with a battery as big as 68Wh for a mere 14" laptop are not indicative of huge efficency improvements, guess that we re good for another big round of deceptive TDPs, an habit at Intel.

Ok so going by your logic Apple silicon MacBook pros with 99 Wh batteries are inefficient as well.

And deceptive TDPs? Please show me where AMD lists the short duration power limits in their product description webpages.

 

[mikk]

Indeed first models with a battery as big as 68Wh for a mere 14" laptop are not indicative of huge efficency improvements, guess that we re good for another big round of deceptive TDPs, an habit at Intel.

Lenovo Yoga with Phoenix has a bigger battery than that. Poor efficiency according to you.

 

[jpiniero]

I imagine the PL1/PL2 on Meteor Lake H laptops will be no different than the OEMs Alder/Raptor ones.

 

[FlameTail]

Will Meteor Lake be found in sub-$1000 laptops?

 

[jpiniero]

Will Meteor Lake be found in sub-$1000 laptops?

There will be at least one "Intel Processor" U part, I imagine. But for the most part Raptor Lake Refresh will be what is used in the cheaper, mass market stuff.

 

[tamz_msc]

Will Meteor Lake be found in sub-$1000 laptops?

Yes. Newegg pre-order prices were a little above thousand for the base i7 model.