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.)

 

[DavidC1]

TS score(Highest for the given TDP from notebookcheck database):
15W 8840U -> 2688 (100%)
17W LNL -> 3438 (128%)
64/20W MTL 185H -> 3537 (132%)

28W 7840U -> 3232 (100%)
64/28W MTL 155H -> 3710 (115%)
30W LNL -> 4151 (128%)

At low TDPs It looks the most impressive. If Intel will keep improving their drivers, then It could be interesting.

Xe2 with the hardware changes will do better than Xe. Xe's deficiency likely makes it harder for the driver team.

Look at what Tom Petersen has been saying:
-SIMD16 internals allow much more games to run OOB, meaning it won't need Day 1 changes just to make it work
-Draw/Execute Indirect boosts performance, especially in UE5 games
-Fast Clear native hardware for Xe2.

The strong demand on the iGPUs from synthetic benchmarks such as Time Spy makes Alchemist look better than it is. Realistically Xe2 will do better in games because of the changes. Supporting Fast Clear only now is a big surprise too. Someone said AMD/Nvidia had it for more than a decade or something? Fast Clear also means needing less memory BW per performance.

Also note that 155H's Arc 7 tops at 2.25GHz GPU Turbo, while Lunarlake is only 2.05GHz. So it's faster even though the top clocks are lower.

 

[TwistedAndy]

It's already kinda established that those 2 Crestmont LPE cores in MTL SoC tile aren't powerful enough to handle the background tasks. If the same cores are gonna feature in ARL, they're still not gonna be very useful. Or am I missing something crucial?

I guess we will get similar Crestmont cores. Probably, they won't be exposed to the system.

I think Intel would like to use the SoC from Meteor Lake in Arrow Lake, but it won't match Microsoft's 40 TOPs requirement. So, Intel probably has to make some changes to the SoC tile. Maybe we will get Skymont cores and the Side Cache there, but it's unlikely.

 

[TESKATLIPOKA]

Lunar Lake is going to be a big improvement for small form factor laptops. Imagine a low power laptop with 8 Raptor Cove cores. Pretty good. Now imagine 4 Lion Cove + 4 Raptor Cove. Even better. Top of the stack parts might well compete with the 14500.

I would wait for reviews, but so far nT performance doesn't look very strong against competition, which is kinda expected considering It has only 8 cores and no HT.
Compared to Ultra 5 135U 59/18W (2P+8E+2LP), which managed higher 9315pts in CB R23 this one should still be better in comparison from 1->8 threads.

It won't be a bad SoC or anything.
Yes, If you check raw performance of the CPU part then It will loose against Strix maybe even against Kraken at the same TDP, but mostly due to higher core count or SMT or both.
Still, I can agree that not many users would realistically need more than 8 cores or use more than 8 threaded Apps, so this limitation shouldn't be a real problem.
The question is at what price It will be sold compared to competition.

 

[Magio]

LNL has a new GPU architecture, so we don't know If TS score will be inflated or not or by how much compared to MTL.
On the other hand, even If It is inflated, It is not a low value.
If It was TS Graphics, It would be better, but whatever.

TS score(Highest for the given TDP from notebookcheck database):
15W 8840U -> 2688 (100%)
17W LNL -> 3438 (128%)
64/20W MTL 185H -> 3537 (132%)

28W 7840U -> 3232 (100%)
64/28W MTL 155H -> 3710 (115%)
30W LNL -> 4151 (128%)

At low TDPs It looks the most impressive. If Intel will keep improving their drivers, then It could be interesting.

Is the 15W 8840U score with a 15W PL1/2, or just a 15W PL1? Because the guy who gave the 17W LNL score confirmed it's 17W PL1/PL2.

 

[TESKATLIPOKA]

Xe2 with the hardware changes will do better than Xe. Xe's deficiency likely makes it harder for the driver team.

Look at what Tom Petersen has been saying:
-SIMD16 internals allow much more games to run OOB, meaning it won't need Day 1 changes just to make it work
-Draw/Execute Indirect boosts performance, especially in UE5 games
-Fast Clear native hardware for Xe2.

The strong demand on the iGPUs from synthetic benchmarks such as Time Spy makes Alchemist look better than it is. Realistically Xe2 will do better in games because of the changes. Supporting Fast Clear only now is a big surprise too. Someone said AMD/Nvidia had it for more than a decade or something? Fast Clear also means needing less memory BW per performance.

Also note that 155H's Arc 7 tops at 2.25GHz GPU Turbo, while Lunarlake is only 2.05GHz. So it's faster even though the top clocks are lower.

155H uses ARC 8 at 2.25GHz and even If LNL has lower turbo, It's highly possible It can keep higher sustained clocks than MTL, so I wouldn't make conclusions based on It.

So the question is how much better will Battlemage do than Alchemist in games. We don't know that yet.

@Magio I edited my original post, It is 15W PL1/PL2 GPD Win Mini Zen 4

 

[DavidC1]

155H uses ARC 8 at 2.25GHz and even If LNL has lower turbo, It's highly possible It can keep higher sustained clocks than MTL, so I wouldn't make conclusions based on It.

So the question is how much better will Battlemage do than Alchemist in games. We don't know that yet.

@Magio I edited my original post, It is 15W PL1/PL2 GPD Win Mini Zen 4

Alchemist is shown by C&C to have significant flaws that it can't take advantage of neither the 512GB/s memory nor the full architecture unless it has higher workload, which is why it's significantly better on 4K and high setting setups. It sounds like you didn't read all of my post.

Also without the Day 1 driver optimizations it often does not work out of the box which Peterson said the SIMD16 will make it have many more games work without changes.

The changes highlighted doesn't sound like something trivial. It sounds like it'll be quite a bit better on games too. Like I said, 3DMark benchmarks pummel iGPUs with mostly unplayable frame rates, which favors Alchemist. Battlemage will do better without some unforseen errors.

Look at the graph where Intel says it's "1.5x" performance. It calculates out to be 1.3x at the high end, so if you assume similar clocks it's 1.5x faster.

 

[TESKATLIPOKA]

Look at the graph where Intel says it's "1.5x" performance. It calculates out to be 1.3x at the high end, so if you assume similar clocks it's 1.5x faster.

1.5x performance was calculated based on TS and against MTL 165U and that one has 4 XEs at 2GHz.
And you have an actual TS score for LNL, unless that is a fake.

 

[DavidC1]

1.5x performance was calculated based on TS and against MTL 165U and that one has 4 XEs.
And you have an actual TS score, unless that is a fake.

Take a look at the Lunarlake presentation.

1.5x is against the -U but at far lower power. It performs 2.25x as fast in it's full envelope. It's more than 1.3x over -H. They sandbagged it. I bet that's not over.

 

[TESKATLIPOKA]

Take a look at the Lunarlake presentation.

1.5x is against the -U but at far lower power. It performs 2.25x as fast in it's full envelope. It's more than 1.3x over -H. They sandbagged it. I bet that's not over.

I don't need to look at anything from Intel's presentation, we already have LNL scores in TS at 17W and 30W unless they are fake.
We just don't know how It performs in actual games and this we won't know for some time unless another leak is out.

 

[dullard]

But aren't they reusing the exact same TSMC N6 SoC tile from MTL? If thats the case, how can we expect any improvement?

I didn't realize there were no changes. If so, then we'd only get scheduling improvements (and possibly different power levels available but that wouldn't change much).

 

[Philste]

https://tieba.baidu.com/p/9073067021#/

 

[Henry swagger]

https://twitter.com/x/status/1807808541010591796

27% single core improvement 😁.. lion cove is roaring up the charts

 

[SiliconFly]

https://twitter.com/x/status/1807808541010591796

27% single core improvement 😁.. lion cove is roaring up the charts

OMG!!! Is this even REAL??? or FAKE?

Thats a massive 27% single-core uplift. That too against RPL. Against MTL, it'll be well above 30%.

Assuming it's real, it's time for me say "I told you so" to around a dozen people here.

Hold on! @SiliconFly is typing...

 

[Henry swagger]

OMG!!! Is this even REAL??? or FAKE?

Thats a massive 27% single-core uplift. That too against RPL. Against MTL, it'll be well above 30%.

Assuming it's real, it's time for me say "I told you so" to around a dozen people here.

Hold on! @SiliconFly is typing...

I.ve already said desktop lion cove will be 25% or 30% plus over raptor cove.. in the forum they said its using 20A node 🤔

 

[SiliconFly]

I.ve already said desktop lion cove will be 25% or 30% plus over raptor cove.. in the forum they said its using 20A node 🤔

Just went thru it using translate. 20A ROCKS!!!

After being disappointed for so many years, I'm finally happy to see Intel firing on all cylinders!!!

 

[SiliconFly]

Also, at this stage in the cycle, it has to be an ES. If this has this level of performance, what about ARL final?

30% ST perf over RPL and 33%-35% ST perf over MTL?

 

[Hitman928]

All aboard!!

I guess some people won’t learn.

 

[whoshere]

• CPU-Z has never been considered a valid/good/decent benchmark.
• Arrow Lake is at least Q4/2024, it's too early for such leaks. Intel last weak announced they had starting to tap out Lunar Lake CPUs.
• Such a screenshot is too easy to photoshop. Don't get your hopes high. We've already seen fabricated Zen 5 screenshots from China.

GB6/GB5 ST results will show a much better picture. MT not so much.

 

[SiliconFly]

...Such a screenshot is too easy to photoshop...

😭

 

[AcrosTinus]

• CPU-Z has never been considered a valid/good/decent benchmark.
• Arrow Lake is at least Q4/2024, it's too early for such leaks. Intel last weak announced they had starting to tap out Lunar Lake CPUs.
• Such a screenshot is too easy to photoshop. Don't get your hopes high. We've already seen fabricated Zen 5 screenshots from China.

GB6/GB5 ST results will show a much better picture. MT not so much.

It is a alright benchmark that at least scales compared to GB6. CPU-Z is a good indicator if compared against other Intel CPUs.
Chips and Cheese analyzed the bench and came to the conclusion that the instruction mix does not accurately resemble consumer workloads.
I call half BS on that, which bench really resembles consumer workloads SPEC - FP or INT, Cinebench, all the other crazy SIMD benches ?
I think the bench got overcriticized because AMD is slow in it, showing that the Zen4 core is less wide AND SLOWER, which it is.

I cannot make this sh... up, wait for GB6 benches and don't trust CPU-Z.....

The leaks seem possible, E-Cores are buffed beyond 50% and the P core are min +15% faster with additional news of the new layout.
Current System (13700K@1,34V 2P@5,8 6P@5,7 8P@5,5 8E@4,4 = CPU-Z : ST=938, MT= 13100), so quite realistic.

 

[H433x0n]

Can we please not hype an easily faked benchmark? Even if it somehow is not faked, CPU-Z is not representative of anything.

 

[DrMrLordX]

Because in reality there was no IPC increase from Zen3 to Zen4.

Wow that's some . . . interesting analysis there!

 

[AcrosTinus]

Wow that's some . . . interesting analysis there!

I am right, the bench showed nearly zero IPC gain (CPU-Z), the performance increase in CPU-Z was due to clocks. I think other benches on average showed 10% or 13% increase from Zen3 to Zen4 or is that wrong ?

 

[DrMrLordX]

I am right, the bench showed nearly zero IPC gain (CPU-Z), the performance increase in CPU-Z was due to clocks. I think other benches on average showed 10% or 13% increase from Zen3 to Zen4 or is that wrong ?

And yet other benches showed IPC gains for Zen4. Have you ever considered the possibility that CPU-z benchmark is flawed?

 

[Abwx]

All conspiracy with no evidence. SKL is faster than Zen1 in all cases.

Lol, it s documented by CPU Z themselves, they explained the "reasons" for such a downgrade after they revised the 1.73 version wich was released in december 2016 and "updated" in may 2017 after they noticed the surprisingly good score of the 1800X.