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.)

 

[TwistedAndy]

You keep using the term Golden Cove. It's not. Intel slides say different.

That's because I consider Raptor Cove to be a slight refresh of Golden Cove. Yes, technically the cores are different and there's a 1-3% difference in terms of IPC, but it's minimal.

Shifting nodes during a refresh is unlikely afaik.

Intel did that with their Tick-Tock model for years to reduce the risks. Also, I don't think Intel is excited to pay TSMC to produce chips when they will have their good node

 

[SiliconFly]

That's because I consider Raptor Cove to be a slight refresh of Golden Cove. Yes, technically the cores are different and there's a 1-3% difference in terms of IPC, but it's minimal.

You may use Sunny Cove, Willow Cove or Golden Cove. But for all practical purposes, we should go with the source. Thats accurate. Otherwise, it doesn't sounds real.

Intel did that with their Tick-Tock model for years to reduce the risks. Also, I don't think Intel is excited to pay TSMC to produce chips when they will have their good node

You're expecting ARL-R to be a tick (node shrink). But leaks suggest that it's just a refresh (same design on the same node with a few tweaks). I think it's time for Intel to push the pedal to the metal, instead of churning out ungodly refreshes.

 

[SiliconFly]

This just in from Raichu...



It seems Panther Lake -> Darkmont is getting some awesome updates.

 

[TwistedAndy]

I think it's time for Intel to push the pedal to the metal, instead of churning out ungodly refreshes.

Yep. I hope the competition from AMD, Qualcomm, and other companies will make this process faster

 

[Geddagod]

This just in from Raichu...

Thought since the Xino purge, people would be a lot more respectful of what a locked account means...

It seems Panther Lake -> Darkmont is getting some awesome updates.

Personally, I expect something similar to Crestmont< Gracemont

Plus the 8+32 sku will come

Thought that got canned

 

[static shock]

8+32 core? Is a Xeon?

 

[Geddagod]

8+32 core? Is a Xeon?

No, the rumored i9 ARL-R sku, which I think got canned? Idk where that rumor is from tho, so don't take my word for it.

 

[Henry swagger]

Thought since the Xino purge, people would be a lot more respectful of what a locked account means...

Personally, I expect something similar to Crestmont< Gracemont

Thought that got canned

For the arrow lake refresh its been activated again

 

[Geddagod]

For the arrow lake refresh its been activated again

Ik MLID said that in one of his videos a couple months ago, I'm referring to a rumor much more recent.
And besides, Idk how much sense it would even make tbh....

 

[Hulk]

Honestly, if Skymont is Raptor Cove from an IPC point of view, I don't see a lot of applications on the desktop that could use 8+32 (over 8+16) outside of CB benchmarking and DC. With Raptor Lake once all of the Raptor Cove physical threads are occupied there is a significant drop in performance on the Gracemont threads. We're looking at a nearly 50% IPC improvement from Gracemont to Skymont. It's massive.

If Lion Cove is +14% over Raptor Cove and can do 5.4GHz all-core and Skymont is delivered as promised Intel will be fine with ARL. More than fine actually. It looks like Intel and AMD have the tech, they just have to deliver it in a timely manner.

 

[Ghostsonplanets]

For the arrow lake refresh its been activated again

Afaik 8+32 was cancelled in favor of bigger NPU on ARL-R.

 

[Geddagod]

Honestly, if Skymont is Raptor Cove from an IPC point of view, I don't see a lot of applications on the desktop that could use 8+32 (over 8+16) outside of CB benchmarking and DC. With Raptor Lake once all of the Raptor Cove physical threads are occupied there is a significant drop in performance on the Gracemont threads. We're looking at a nearly 50% IPC improvement from Gracemont to Skymont. It's massive.

If Lion Cove is +14% over Raptor Cove and can do 5.4GHz all-core and Skymont is delivered as promised Intel will be fine with ARL. More than fine actually. It looks like Intel and AMD have the tech, they just have to deliver it in a timely manner.

Napkin math using 5.5GHz nT and 4.6 Ghz nT for LNC and SKMT suggests a ~20% uplift over RPL for INT workloads.

 

[The Hardcard]

Honestly, if Skymont is Raptor Cove from an IPC point of view, I don't see a lot of applications on the desktop that could use 8+32 (over 8+16) outside of CB benchmarking and DC. With Raptor Lake once all of the Raptor Cove physical threads are occupied there is a significant drop in performance on the Gracemont threads. We're looking at a nearly 50% IPC improvement from Gracemont to Skymont. It's massive.

If Lion Cove is +14% over Raptor Cove and can do 5.4GHz all-core and Skymont is delivered as promised Intel will be fine with ARL. More than fine actually. It looks like Intel and AMD have the tech, they just have to deliver it in a timely manner.

If they get the power and thermals under control, an 8+32 might appeal to the low end workstation market. Especially if they can put it into a Mac Studio type box that is portable and quiet, but with enough power for say video, editors, and audio engineers, as well as others who could use that kind of oomph in an easy to move device.

Put in a 5080 GPU with the easy option of lower clocks so it also doesn’t need fans blasting, it will appeal to some.

 

[TESKATLIPOKA]

Napkin math using 5.5GHz nT and 4.6 Ghz nT for LNC and SKMT suggests a ~20% uplift over RPL for INT workloads.

That napkin math is with SMT enabled on LNC or not?

 

[Geddagod]

That napkin math is with SMT enabled on LNC or not?

Not enabled. I don't think we are getting SMT on ARL either...

If they get the power and thermals under control, an 8+32 might appeal to the low end workstation market. Especially if they can put it into a Mac Studio type box that is portable and quiet, but with enough power for say video, editors, and audio engineers, as well as others who could use that kind of oomph in an easy to move device.

Put in a 5080 GPU with the easy option of lower clocks so it also doesn’t need fans blasting, it will appeal to some.

Power, mem bandwidth, and sheer die size and cost all start becoming bigger and bigger issues...

 

[SiliconFly]

Afaik 8+32 was cancelled in favor of bigger NPU on ARL-R.

Actually, there were no good leaks regarding ARL-R 8+32 being cancelled. I still think it's a possibility.

 

[Elfear]

Actually, there were no good leaks regarding ARL-R 8+32 being cancelled. I still think it's a possibility.

Based on @Hitman928's chip area calculations, Skymont is ~40-50% the size of Lion Cove. So an 8+32 would be like 21-24 P-cores on one chip area-wise. Is that feasible? It would certainly be a MT monster.

 

[SiliconFly]

Based on @Hitman928's chip area calculations, Skymont is ~40-50% the size of Lion Cove. So an 8+32 would be like 21-24 P-cores on one chip area-wise. Is that feasible? It would certainly be a MT monster.

The P core to E core ratio is close to 1:3. So, roughly 18-19. And yes, it'd be a monster. And very much suitable for HEDT (like threadrippers).

 

[Hulk]

6+24 would be only slightly larger area-wise than 8+16 and deliver nearly the same ST performance a a significant boost in MT now that the E's are so performant.

 

[Hitman928]

The P core to E core ratio is close to 1:3. So, roughly 18-19. And yes, it'd be a monster. And very much suitable for HEDT (like threadrippers).

It’s not, at least not in LNL. Maybe in ARL but that’s TBD, though I doubt it.

 

[dullard]

6+24 would be only slightly larger area-wise than 8+16 and deliver nearly the same ST performance a a significant boost in MT now that the E's are so performant.

I've been calling for that 6+24 solution for many years. One example:

...I think Intel is missing out on a potential multithreaded monster: 6 P cores, 24 E cores. Instead they went with 8 P Cores and 16 E cores. Both would use about the same amount of power and about the same amount of silicon area. But 6+24 would blow these away in multi-threaded applications.

 

[Wolverine2349]

6+24 would be only slightly larger area-wise than 8+16 and deliver nearly the same ST performance a a significant boost in MT now that the E's are so performant.

Well wouldn't it be just as performant single thread as doesn't single thread mean only 1 core? Or does single thread also mean more than 1 core, but limited amount beyond 1 core for SMT?

 

[ondma]

I've been calling for that 6+24 solution for many years. One example:

Come on guys!!! Please don't give Intel any more ideas on how to span E cores. I could see 8+24, perhaps for (uggh) ARL-R, but I really would hate to see them delete P cores in order to include more E cores.

 

[dullard]

Come on guys!!! Please don't give Intel any more ideas on how to span E cores. I could see 8+24, perhaps for (uggh) ARL-R, but I really would hate to see them delete P cores in order to include more E cores.

More performance, less power. Why wouldn't you want that. Unless you fall into the false belief that you want P cores for things like gaming.

 

[Wolverine2349]

More performance, less power. Why wouldn't you want that. Unless you fall into the false belief that you want P cores for things like gaming.

Well more performant lower power P cores so no hybrid scheduling quirks and yes better for gaming for sure.

Maybe the e-cores become the new P cores anyways if the Austin team, is really doing that well and Israel design center is not sometime in a couple or few years???