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.)

 

[adroc_thurston]

+/- 10% margin as well, lol.

margin is fine but a "faster" side by side with two real delta numbers is hilarious.
it's like AMD glossing over PHX iGP existence on CES'23 for obvious reasons.

 

[CouncilorIrissa]

View attachment 99308
"faster".
Well, ugh, that's a metric I guess!

One hopes that it's faster while also not crashing upon running a benchmark.

 

[jpiniero]

it's like AMD glossing over PHX iGP existence on CES'23 for obvious reasons.

The entire industry only cares about AI?

(I bet Intel wanted three squares, and the "faster" one is the only thing else they could think of)

 

[adroc_thurston]

One hopes that it's faster while also not crashing upon running a benchmark.

oh yeah MTL was funny. but still, "faster" is really one way to put it.

The entire industry only cares about AI?

MS does, no one else really.
MS wants to get sub revenue in windows BU, pretty simple.

 

[jpiniero]

MS does, no one else really.

I'm not joking that Lisa's keynote at Computex is going to be nonstop AI. You'll be lucky if she spends more than a minute on anything else.

 

[adroc_thurston]

I'm not joking that Lisa's keynote at Computex is going to be nonstop AI. You'll be lucky if she spends more than a minute on anything else.

of course, gotta keep on the message or else The Street gonna riot.
but Zen5 is the selling point.

 

[Joe NYC]

The entire industry only cares about AI?

(I bet Intel wanted three squares, and the "faster" one is the only thing else they could think of)

Three squares are clearly better than two.

 

[controlflow]

Only 30% lower against 7840U certainly isn’t that impressive. Expected much more given it's a specialized part. But at least it's shaping up to be a good enough.

Keep in mind the LNL package power includes the DRAM.

 

[Ghostsonplanets]

Keep in mind the LNL package power includes the DRAM.

I had forgotten about the 2W penalty from the integrated PoP. That makes things a little better.

 

[poke01]

View attachment 99308
"faster".
Well, ugh, that's a metric I guess!

that’s worse than Apple. At least they provide some graphs. 😂

 

[mikk]

50% faster than 165U in timespy at 17W roughly makes it 3000 points, I've heard 5000 points at 30W. But how well does it in real games, 3dmark never was the problem for MTL or ALchemist. I hope Xe2 does better in real games.

 

[uzzi38]

Only 30% lower against 7840U certainly isn’t that impressive. Expected much more given it's a specialized part. But at least it's shaping up to be a good enough.

Only 50% faster than MTL-U7 165U is certainly a bit disappointing compared to the 2x expectations. Specially given it's a 1024 ALU part at >=1.5GHz.

MS Teams video conferencing is literally best case scenario for Phoenix, it's like the one workload where AMD had a solid lead in battery for a while. This is genuinely a really damn good result.

 

[adroc_thurston]

MS Teams video conferencing is literally best case scenario for Phoenix, it's like the one workload where AMD had a solid lead in battery for a while. This is genuinely a really damn good result.

Yeah it's a pretty good BL part.
It doesn't compete with HPT tho.

 

[poke01]

MS Teams video conferencing is literally best case scenario for Phoenix, it's like the one workload where AMD had a solid lead in battery for a while. This is genuinely a really damn good result.

Could the NPU played a part, since it was a AI focused Teams benchmark?

 

[Joe NYC]

MS Teams video conferencing is literally best case scenario for Phoenix, it's like the one workload where AMD had a solid lead in battery for a while. This is genuinely a really damn good result.

Explains then why Intel is not comparing Lunar Lake to MTL or RPL, at the same time as it is comparing with QCOM and AMD...

 

[Ghostsonplanets]

MS Teams video conferencing is literally best case scenario for Phoenix, it's like the one workload where AMD had a solid lead in battery for a while. This is genuinely a really damn good result.

AMD stated that 7840 Pro gets 6:48 hours in MS Teams. If we go by Intel quoted figure, that means >8+ hours for Lunar Lake devices. Fairly impressive then.

 

[poke01]

Explains then why Intel is not comparing Lunar Lake to MTL or RPL, at the same time as it is comparing with QCOM and AMD...

It was an NPU based test. Qualcomm did the basically the same test with MTL with AI effects and the X Elite had a huge lead in Teams in regards to battery life.

 

[Philste]

50% faster than 165U in timespy at 17W roughly makes it 3000 points, I've heard 5000 points at 30W. But how well does it in real games, 3dmark never was the problem for MTL or ALchemist. I hope Xe2 does better in real games.

At first I was very underwelmed with the 1.5× figure, but then I saw it was the 17W one. If intel can get their stuff together and launch a XeSS 2.0 with Frame Gen that is implemented in many games, it could be a good iGPU. It also has the XMX Units for the good Version of XeSS, which already got praised when the DP4a Version was criticised. And even the latter seems to be really good now. If 5000 Timespy at 30W is true, it may even come close to Strix in that department.

 

[uzzi38]

50% faster than 165U in timespy at 17W roughly makes it 3000 points, I've heard 5000 points at 30W. But how well does it in real games, 3dmark never was the problem for MTL or ALchemist. I hope Xe2 does better in real games.

50% faster at 17W vs MTL-U at 15W still unfortunately puts it behind Phoenix at 15W in actual games, at that power config MTL was often only half as fast as Phoenix/Hawk Point. I'd seen performance estimations for LNL over MTL/ADL in a couple of (synthetic) workloads but they were very large ranges and 50% does end up on the lower end of the scale. I'm actually still holding out a little hope we see larger gains in real workloads.

But even assuming we're looking at 50% it's now close enough that if LNL scales down to lower power levels better than PHX/HPT (AKA doesn't completely fall apart below 12W) that still makes LNL a very viable part for gaming handhelds. Fingers crossed, I do want viable Intel parts for handhelds.

 

[Saylick]

Robert Hallock says the LP Island has 2x the compute over MTL, which is kinda surprising to me considering LNL has 2x as many LP cores as MTL already. I was expecting more than 2x because the LP cores should be beefier on an individual basis due to them being Skymont. That is, unless LNL's LP cores are clocked lower than MTL's...

 

[adroc_thurston]

That is, unless LNL's LP cores are clocked lower than MTL's...

yea.

 

[poke01]

It was an NPU based test. Qualcomm did the basically the same test with MTL with AI effects and the X Elite had a huge lead in Teams in regards to battery life.

Here it is.

 

[Kepler_L2]

I hope they are sandbagging GPU perf or Timespy is a worst case scenario for BMG (as it was a best case scenario for ACM), because 1.5x with double Xe Cores/ALUs is a bigger disaster than even RDNA3.

 

[adroc_thurston]

because 1.5x with double Xe Cores/ALUs is a bigger disaster than even RDNA3.

Well it clocks lower.
I think.
Oh wait, they're comparing it to the 64EU abortion?

 

[Kepler_L2]

Well it clocks lower.
I think.
Oh wait, they're comparing it to the 64EU abortion?

Yep