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.)

 

[Hulk]

1.095V for 2000 Mhz is a horribly high voltage. My A380 runs at 0.985V for 2450 Mhz and MTL tGPU requires 0.7xV for 2000 Mhz. No wonder Xe LP didn't run high clocked.

I ran Topaz Photo AI and PureRaw3 just to see what the igpu was drawing. About 12 watts for both of them. More than the GPU-Z stress test.
How much more compute does the A380 have over the igpu in my 13600K? Would be nice to speed this operations up a bit.
Do you use the A380 for gaming or compute?

 

[igor_kavinski]

My Iris Xe on the 11370H pulls 0.15 watts during this test.

Proofy proofy sensor screenshot, where is it?

 

[mikk]

I ran Topaz Photo AI and PureRaw3 just to see what the igpu was drawing. About 12 watts for both of them. More than the GPU-Z stress test.
How much more compute does the A380 have over the igpu in my 13600K? Would be nice to speed this operations up a bit.
Do you use the A380 for gaming or compute?

Try Furmark or Furry Cube in CapsViewer, power consumption should be higher there. A380 has 5x the compute, 1 tflop vs 5 tflops.

 

[SiliconFly]

Yeah, it did doesn't really make sense to keep those since there's NPU. I do wonder if XeSS will use DP4a path then... Or if it would be possible to make it run on NPU.

Not sure whether MTL supports DP4a natively, but it does have support for either DP4a and/or XMX functionality in NPU in some fashion (source: Intel demo), thereby negating the requirement of XMX in the GPU tile. Not an ideal solution when it comes to performance/power savings, but it does reduce tGPU die area thereby lowering MTL cost. Thats probably the reason they're trying to put XMX back in ARL's GPU tile.

 

[tamz_msc]

With Topaz Photo AI:



Running GPU-Z stress test:

 

[Hulk]

With Topaz Photo AI:

View attachment 89027

Running GPU-Z stress test:

View attachment 89028

After looking at these results I don't think the GPU-Z "stress test" is a stress test at all. Or I should say for a weak gpu like the igpu in my 13600K it's quite taxing as indicated by the power required. But your A380 is only loafing to run this routine as evidenced by the low clock and power usage. Your gpu is way down the v/f curve, mine way up. Interesting bench as it doesn't run "as fast as the gpu can run" it but the same for every gpu. It's not a race to the finish like most benches.

 

[tamz_msc]

After looking at these results I don't think the GPU-Z "stress test" is a stress test at all. Or I should say for a weak gpu like the igpu in my 13600K it's quite taxing as indicated by the power required. But your A380 is only loafing to run this routine as evidenced by the low clock and power usage. Your gpu is way down the v/f curve, mine way up. Interesting bench as it doesn't run "as fast as the gpu can run" it but the same for every gpu. It's not a race to the finish like most benches.

By stress test, I mean that it simply stresses the internal links between the iGPU and the rest of the chip. It is essentially running the data paths at full tilt. I agree that it doesn't stress anything in the traditional sense.

But your idle power consumption is way too high.

 

[Hulk]

By stress test, I mean that it simply stresses the internal links between the iGPU and the rest of the chip. It is essentially running the data paths at full tilt. I agree that it doesn't stress anything in the traditional sense.

But your idle power consumption is way too high.

Got it. But my idle power is zero?

 

[tamz_msc]

The minimum "GT cores power" was 10.4 watts according to your first screenshot.

 

[Henry swagger]

https://twitter.com/x/status/1725921039069462846

Meteor lake igpu looking good.. battlemage will be evem better

 

[Hulk]

I must have opened the app while the gpu was loaded. It's goes to 0 when not used. Or almost 0.

The minimum "GT cores power" was 10.4 watts according to your first screenshot.

 

[SiliconFly]

https://twitter.com/x/status/1725921039069462846

Meteor lake igpu looking good.. battlemage will be evem better

Arrow Lake has Alchemist++ while Lunar Lake has Battlemage, and both ARL & LNL are launching side by side. This is nuts. Why on earth would Intel choose to put Alchemist++ in ARL instead of the newer Battlemage? Is it just Intel being Intel... stuck in the same old +++ c**p?

 

[Thibsie]

Arrow Lake has Alchemist++ while Lunar Lake has Battlemage, and both ARL & LNL are launching side by side. This is nuts. Why on earth would Intel choose to put Alchemist++ in ARL instead of the newer Battlemage? Is it just Intel being Intel... stuck in the same old +++ c**p?

I wouldn't take this guy any more seriously than MLID.

 

[mikk]

Arrow Lake has Alchemist++ while Lunar Lake has Battlemage, and both ARL & LNL are launching side by side. This is nuts. Why on earth would Intel choose to put Alchemist++ in ARL instead of the newer Battlemage? Is it just Intel being Intel... stuck in the same old +++ c**p?

Because Arrow Lake was planned for late 2023 when Battlemage was no option. Intel isn't flexible enough to change the graphics tile completely in the middle of a chip development. Lunar Lake was planned for 2025, so they were able to use Battlemage from the beginning.

 

[SiliconFly]

Because Arrow Lake was planned for late 2023 when Battlemage was no option. Intel isn't flexible enough to change the graphics tile completely in the middle of a chip development. Lunar Lake was planned for 2025, so they were able to use Battlemage from the beginning.

I'm actually a bit excited about ARL, considering it has a brand spanking new process 20A (with GAA & BPD). It also has a brand new core architecture LNC which I hope will finally address the issue of P cores being super fat. If LNC P cores are leaner than existing ones with good performance, it'll be Intel's Zen moment. Also, their tile architecture with Foveros is far ahead of competition already. The only hiccup for ARL at this point is Alchemist++ imo.

 

[Hulk]

The minimum "GT cores power" was 10.4 watts according to your first screenshot.

GPU was loaded when HWinfo was opened.
Idle @ 0.1W

 

[Hulk]

Arrow Lake has Alchemist++ while Lunar Lake has Battlemage, and both ARL & LNL are launching side by side. This is nuts. Why on earth would Intel choose to put Alchemist++ in ARL instead of the newer Battlemage? Is it just Intel being Intel... stuck in the same old +++ c**p?

Past history has shown us Intel puts little effort into desktop iGPU's. They have tremendous sales for business desktop that simply need the gpu to drive the monitor to view spreadsheets and websites. If anyone needs more for the desktop they can add in a discrete GPU.

On the other hand you get what you get with mobile graphics (no upgrade path) so they put more effort into that segment.

It is possible it was easier to get ARL moving along with the older tech and who knows the tile might even be smaller/cheaper to produce?

 

[Joe NYC]

Looks promising:

https://twitter.com/x/status/1726438814430937187

https://twitter.com/x/status/1726438872819913050

 

[Geddagod]

Uh, I think people don't like linking twitter stuff here, so here:

 

[igor_kavinski]

OK, WOAH!!! The thing that jumps out at me is real time raytracing! What the heck is that??? Does that mean zero cost raytracing in any scene????

 

[Joe NYC]

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

Deleted from Twitter. I was just saving the images, but you beat me to it.

 

[igor_kavinski]

Everyone cheer: Gedda! Gedda! GEDDA!!!!

 

[Geddagod]

40% scenario power reduction seems pretty interesting. Of course it's very vague, but if it's talking about CPU perf, that actually sounds not bad for a jump from Intel 4 to N3B, with a brand new arch.

 

[Tigerick]

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

Nice catch, saving now ...

I am more interested on the last slide which mentioned 8 Xe LPG cores equal to 1024 ALU. And based on 3.8TF, we could estimate the iGPU clock speed to be around 1.85GHz...

FYI, M2's iGPU is having 3.6TF, slightly lower than LNL. We don't know the TF of M3 but should be slightly higher than M2; of course the real upgrade on M3 is hardware RT support. However, the real competitor is actually upcoming M4 from Apple cause Intel will most likely ship LNL in early 2025.

Well, at least Intel managed to have Apple's competitor after so many years of talking, let's not forget that LNL is based on N3B process, where is 18A's CPU that offer better PPA??? Still waiting

 

[tamz_msc]

So Lunar Lake MX will integrate the CPU and GPU on one tile?