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.





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.)

 

[Win2012R2]

Yep, Panther Lake is officially "rescheduled" for 2026

So 18A isn't that healthy then.

 

[511]

Their CEO said Yesterday that Panther Lake is launching on 18A in H2 25 looked like the entire management is clueless 🤣 they should not be called management but clowns

 

[dullard]

18A is now in risk production: meaning designs for the initial chips are locked down and Intel is ramping up the ability to mass produce it (from hundreds of chips per day to tens of thousands per day).

Panther Lake will be sold by Intel to Intel's customers in H2 2025 (laptop manufacturers). But, since laptops take time to optimize they are likely not going to be released to the public until 2026. This timing is up to the manufacturers of the laptops--not Intel. So, while Intel will launch it in 2025, we won't likely get it in 2025. Exactly the same thing that happened many times over the last few years.

 

[511]

18A is now in risk production: meaning designs for the initial chips are locked down and Intel is ramping up the ability to mass produce it (from hundreds of chips per day to tens of thousands per day).

Panther Lake will be sold by Intel to Intel's customers in H2 2025 (laptop manufacturers). But, since laptops take time to optimize they are likely not going to be released to the public until 2026. This timing is up to the manufacturers of the laptops--not Intel. So, while Intel will launch it in 2025, we won't likely get it in 2025. Exactly the same thing that happened many times over the last few years.

WE will have 3-4 Panther Lake model to buy this year at best I guess.

 

[igor_kavinski]

WE will have 3-4 Panther Lake model to buy this year at best I guess.

Not like Lunar Lake is widely available. Most Amazon listings are for Raptor/Alder Lake laptops. They need maybe a year to ramp up production for LAST year's mobile chip. It's also possible that Microsoft bought up all the good dies for Lunar Lake for their Surface line, after the major disappointment of Snapdragon Elite.

 

[GTracing]

Not like Lunar Lake is widely available. Most Amazon listings are for Raptor/Alder Lake laptops. They need maybe a year to ramp up production for LAST year's mobile chip. It's also possible that Microsoft bought up all the good dies for Lunar Lake for their Surface line, after the major disappointment of Snapdragon Elite.

Lunar Lake is in over 40 different devices. https://www.ultrabookreview.com/69679-intel-lunar-lake-laptops/

How many options does it need to be considered widely available?

 

[igor_kavinski]

How many options does it need to be considered widely available?

It's not the models. It's the quantity. Meteor Lake is available almost like 100:1 compared to Lunar Lake. Something like 20:1 would be acceptable. And with reduced availability comes the artificially high prices.

 

[Markfw]

It's not the models. It's the quantity. Meteor Lake is available almost like 100:1 compared to Lunar Lake. Something like 20:1 would be acceptable. And with reduced availability comes the artificially high prices.

I could be wrong, but is it something like "Intel wants to dump all meteor lake first before introducing new models that are better in all regards" ??

 

[GTracing]

It's not the models. It's the quantity. Meteor Lake is available almost like 100:1 compared to Lunar Lake. Something like 20:1 would be acceptable. And with reduced availability comes the artificially high prices.

You got a source on that 100:1 number? That's crazy if true.

And pricing on Lunar Lake devices is good in the US at least. Better than X Elite laptops.

 

[igor_kavinski]

You got a source on that 100:1 number? That's crazy if true.

Just what I'm seeing searching on my local Amazon.ae

 

[DKR]

Interesting conversation here

 

[Josh128]

Intel and TSMC tentative joint venture announced!

https://finance.yahoo.com/news/intel-tsmc-tentatively-agree-form-185938740.html

 

[poke01]

lol Intel is bad at design now
Removed SMT to save 15% die area but it’s still a massive core.

 

[adroc_thurston]

lol Intel is bad at design now
Removed SMT to save 15% die area but it’s still a massive core.

that's not the real reason

 

[Markfw]

lol Intel is bad at design now
Removed SMT to save 15% die area but it’s still a massive core.

and still loses to its competition. Yes, bad at design. I hope they abandon P and E cores, and start from scratch with the new management. I would actually like to see them beat Zen, it will create even more competition.

 

[poke01]

that's not the real reason

Well that’s the reason Intel gave. These Intels engineers are just marketing

 

[511]

and still loses to its competition. Yes, bad at design. I hope they abandon P and E cores, and start from scratch with the new management. I would actually like to see them beat Zen, it will create even more competition.

The P core is bad not the E core lol it's been using large building blocks for a while

With Zen 5 presentation at ISSCC it's clear Zen's RTL/Physical Layout is just better than Intel's P core on N3B which is most likely using 3-2 Fin to get the performance while Zen 5 is using N4P 2-2 Fin with 6 tracks and more metal layers to hit the frequency. N4P 2-2 fin is more denser than the N3B 3-2 fin Intel is using

 

[511]

Intel engineer said Panther Cove in the interview with kitguruTech 🤣

 

[Fjodor2001]

I just wonder if someone has found any CPU performance scaling graph for 285K? I'd like to see how well it performs at lower TDP levels (~65-120W) compared to e.g. 9950X(3D) and 14900K.

I found this one here:

From this post. But in that post there is no link or reference to the source where it comes from.

So I wonder if someone has found anything similar, and can link to some official source? Thanks!

 

[coercitiv]

From this post. But in that post there is no link or reference to the source where it comes from.

Reference.

 

[Fjodor2001]

So based on this, if one is looking for a CPU that:

* Performs well when TDP limit configured to 100 W (since after that limit you don't get much extra perf for the additional Watts)
* Performs well in MT use cases
* Has low idle power consumption (since turned on 24 h, but often idling), and 285K has lower idle power consumption than 9950X(3D)

then 285K would currently be the best option on desktop?

 

[gdansk]

So based on this, if one is looking for a CPU that:

* Performs well when TDP limit configured to 100 W (since after that limit you don't get much extra perf for the additional Watts)
* Performs well in MT use cases
* Has low idle power consumption (since turned on 24 h, but often idling), and 285K has lower idle power consumption than 9950X(3D)

then 285K would currently be the best option on desktop?

If your workload resembles CBr24 and doesn't mind p/e core dichotomy then possibly. It has a more efficient manufacturing process, lower pj/b interconnect and faster memory controller which all help at lower power CBr24.

The disadvantages, however, may appear suddenly in other workloads (even SPEC nT).

 

[Fjodor2001]

If your workload resembles CBr24 and doesn't mind p/e core dichotomy then possibly. It has a more efficient manufacturing process, lower pj/b interconnect and faster memory controller which all help at lower power CBr24.

The disadvantages, however, may appear suddenly in other workloads (even SPEC nT).

The intention is to use it mainly for SW compilation and video transcoding.

 

[igor_kavinski]

The intention is to use it mainly for SW compilation and video transcoding.

285K is badly tuned at stock settings. If you want to save yourself the headache, just go with 9700X or 9950X. If you still want Arrow Lake, try going with this: https://www.newegg.com/intel-core-ultra-5-225-arrow-lake-lga-1851-processor/p/N82E16819118582

Then you can upgrade to 385K later which hopefully should have some much needed fixes.

 

[Fjodor2001]

285K is badly tuned at stock settings.

Could you please clarify what you mean by this? Also, note that the intention is to run it with TDP max limit set to approximately 100 W.

Regarding 385K, what fixes are expected?