Discussion Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads

[Tigerick]

Wildcat Lake (WCL) Specs

Intel Wildcat Lake (WCL) is upcoming mobile SoC replacing Raptor Lake-U. WCL consists of 2 tiles: compute tile and PCD tile. It is true single die consists of CPU, GPU and NPU that is fabbed by 18-A process. Last time I checked, PCD tile is fabbed by TSMC N6 process. They are connected through UCIe, not D2D; a first from Intel. Expecting launching in Q1 2026.

		Intel Raptor Lake U	Intel Wildcat Lake 15W?	Intel Lunar Lake		Intel Panther Lake 4+0+4
Launch Date		Q1-2024	Q2-2026	Q3-2024		Q1-2026
Model		Intel 150U	Intel Core 7	Core Ultra 7 268V		Core Ultra 7 365
Dies		2	2	2		3
Node		Intel 7 + ?	Intel 18-A + TSMC N6	TSMC N3B + N6		Intel 18-A + Intel 3 + TSMC N6
CPU		2 P-core + 8 E-cores	2 P-core + 4 LP E-cores	4 P-core + 4 LP E-cores		4 P-core + 4 LP E-cores
Threads		12	6	8		8
Max Clock		5.4 GHz	?	5 GHz		4.8 GHz
L3 Cache		12 MB		12 MB		12 MB
TDP		15 - 55 W	15 W ?	17 - 37 W		25 - 55 W
Memory		128-bit LPDDR5-5200	64-bit LPDDR5	128-bit LPDDR5x-8533		128-bit LPDDR5x-7467
Size		96 GB		32 GB		128 GB
Bandwidth				136 GB/s
GPU		Intel Graphics	Intel Graphics	Arc 140V		Intel Graphics
RT		No	No	YES		YES
EU / Xe		96 EU	2 Xe	8 Xe		4 Xe
Max Clock		1.3 GHz	?	2 GHz		2.5 GHz
NPU		GNA 3.0	18 TOPS	48 TOPS		49 TOPS

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.

 

[DavidC1]

The die-waste unit now wastes 40% less of the die, and the fops(fake ops) is few % better than Lunarlake.

 

[511]

The Xe Media Engine with Panther Lake does add 10-bit AVC encode/decode, 10-bit AV1 encode and decode, and Sony XAVC-H / XAVC-HS / XAVC-S encode and decode.

 

[511]

also 18A looks fine to me

 

[poke01]

The Xe Media Engine with Panther Lake does add 10-bit AVC encode/decode, 10-bit AV1 encode and decode, and Sony XAVC-H / XAVC-HS / XAVC-S encode and decode.

Now that’s industry leading

 

[511]

QSV has been industry leading for a while it basically supports more stuff than NVENC 🤣

 

[DavidC1]

also 18A looks fine to me

5.1GHz Cougar Cove should beat 5.4GHz Lion Cove by few %, if their 5-10% per clock improvements are correct.

This also makes Novalake look much better.

 

[511]

5.1GHz Cougar Cove should beat 5.4GHz Lion Cove by few %, if their 5-10% per clock improvements are correct.

it matches ARL-H so i would expect a draw

 

[CouncilorIrissa]

Page 878 - Discussion - Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads

Page 878 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

This isn't Xe2 vs Xe3. It's Xe2 (Arv 140V) with the power tuning patch on MSI Claw.

Thanks for the correction! The slide deck is highly confusing.

 

[ToTTenTranz]

There's a whopping 16MB L2 in there meaning they increased the L2 per core by 50%, and then it can access up to 18MB L3 on the Compute Tile.

Adding to this, there's also 8MB memory-side cache which may increase effective bandwidth out of the 153.6GB/s LPDDR5X.

Page 878 - Discussion - Intel Meteor, Arrow, Lunar & Panther Lakes + WCL Discussion Threads

Page 878 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

This isn't Xe2 vs Xe3. It's Xe2 (Arv 140V) with the power tuning patch on MSI Claw.

Yes, that slide comes out of nowhere and even the coloring compared to previous slides seems to imply the rose bars are for Panther Lake, when in reality it's just about driver updates on Lunar Lake.

 

[Saylick]

also 18A looks fine to me View attachment 131712

Looks like it’s around N3E/N3P performance based on how it stacks up with N3B? Which has been what TSMC believed it would be from the beginning.

 

[DavidC1]

it matches ARL-H so i would expect a draw

That's not what the graph is saying.

 

[511]

That's not what the graph is saying.

yeah but if you look hard enough not much improvement

Looks like it’s around N3E/N3P performance based on how it stacks up with N3B? Which has been what TSMC believed it would be from the beginning.

I was always expecting between N3P and N2 tbh and it seems like that N3B was between N5 and N3E so this means that which is far better tbh and perfectly servicable

 

[DavidC1]

yeah but if you look hard enough not much improvement

That's still a few %, which is really good if the claims of Lunarlake efficiency is true. They can replace both effectively, while keeping the P cores at 4. Unless the clock speeds are higher than 5.1GHz, 5-10% perf/clk improvements are very reasonable. It's between Tick and Tock, Tick never got branch prediction improvements.

I expect a bit less for Darkmont, in the low single digit, but probably improved power efficiency. I don't think beating 6+8 Arrowlake parts are possible without that. This will probably improve P core's area efficiency because it sounds like the gains are rather big for comparatively small changes, meaning it was a targetted change fixing problems in Lion Cove.

It looks like Stephen Robinson is talking for both cores? Maybe they are already taking some responsibilities away.

Adding to this, there's also 8MB memory-side cache which may increase effective bandwidth out of the 153.6GB/s LPDDR5X.

I think this is just low power play again. Lunarlake's MLC offered almost nothing in terms of performance.


-Darkmont's power/perf graph is steeper compared to Skymont in the Raptorlake comparison. Meaning higher performance at lower power.
-40% perf/W improvement for graphics is good, since N3E isn't a huge improvement over N3B. Architecturally it's 25-30% better ISO-process over Xe2.

 

[Magio]

also 18A looks fine to me View attachment 131712

Does seem good from an efficiency POV, but looking at how the lines taper out after that, peak performance is likely not much of an upgrade.

 

[MoogleW]

5.1GHz Cougar Cove should beat 5.4GHz Lion Cove by few %, if their 5-10% per clock improvements are correct.

This also makes Novalake look much better.

Is that per clock or number of clucks at a specific voltage?

 

[DavidC1]

Is that per clock or number of clucks at a specific voltage?

Read again. 5.1GHz Pantherlake P cores should beat 5.4GHz Arrowlake-H by few single digit %. Probably 3-5%.

Is that per clock or number of clucks at a specific voltage?

Well, it's 4P + 8E + 4LPE vs 6+8. The gains are very respectable considering that it also clocks lower.

 

[Covfefe]

Looks like it’s around N3E/N3P performance based on how it stacks up with N3B? Which has been what TSMC believed it would be from the beginning.

Depends on what you mean by performance. The clock ceiling of panther lake is no better than lunar lake. Which implies 18A is slightly worse than N3E/N3P for high clock speeds.

However, the power consumption of panther lake is reportedly >30% lower than arrow lake at the same performance. Which implies that 18A is more efficient than N3E/N3P.

 

[DavidC1]

Looks like Pantherlake is 10-15% better in MT perf over ARL-H. Very respectable. I reckon it would be ~3-5% without efficiency improvements.

 

[511]

Well, it's 4P + 8E + 4LPE vs 6+8. The gains are very respectable considering that it also clocks lower.

the perf/watt improvements are really nice

 

[regen1]

Branding Xe3 under "B-series" can be somewhat confusing.

For the upcoming client products PTL, WCL, NVL and potential dGPUs:

Vanilla Xe3 is supposed to be used in iGPUs of
Panther Lake, WildCat Lake, NVL Desktop and -HX SKUs,
NVL-(U?) 4+0+4 compute tile SKUs and
may be some NVL-H(4+8+4 compute tile) SKUs in a similar fashion to PTL.

The 4Xe3 GPU tile(Intel 3 node) used in PTL line should be re-used in NVL Desktop and HX SKUs, NVL-(U?) 4+0+4 compute tile and may be some NVL-H(4+8+4 compute tile) SKUs

Xe3P is supposed to come with some NVL-H SKUs(up to 12Xe3P),
NVL-AX(Halo, if it launches), and discrete GPUs(if it launches).

Xe3P should be a significant upgrade over vanilla Xe3.

 

[511]

 

[Geddagod]

if their 5-10% per clock improvements are correct.

Where did they say this? I find it incredibly hard to believe that they squeezed that much out of esentially the same core lol.

 

[DavidC1]

Where did they say this? I find it incredibly hard to believe that they squeezed that much out of esentially the same core lol.

TPU.

Unless you think Pantherlake clocks at 5.6GHz, the performance graphs also support this.

 

[511]

TPU.

Unless you think Pantherlake clocks at 5.6GHz, the performance graphs also support this.

5.1 Ghz

 

[Win2012R2]

So Xe3 isn't Celestial then