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.

 

[Joe NYC]

Intel problem is that they deprecated a chunk of i7 capacity which they now need.

I was wondering about that. Intel must have taken some of that capacity down, for unknown reasons.

I seem to recall the CFO some quarters ago talking (bragging) about some good deal they got on some used equipment. I wonder if it is related.

 

[jpiniero]

I was wondering about that. Intel must have taken some of that capacity down, for unknown reasons.

Fab Space ain't cheap. Take the space used up by 10 nm for newer nodes.

 

[OneEng2]

The compute tile in PTL also has an image processor, a 50 TOPs NPU, display outputs, memory controller and a fat bus that reaches the iGPU chiplet.
You can't compare it to the Zen 5 CCDs that only have the CPU cores and IF.

Fair.

client is getting N2 piles and 2.5D slabs just the same.

Doesn't make NVL architecture any more suited to win server business over the competition. You can waste lots of wafers when you get a fortune for any good die you happen to make. Not so much when you need to offer the chips near commodity pricing (like client).

Fab Space ain't cheap. Take the space used up by 10 nm for newer nodes.

10nm isn't exactly "old". I think that if Intel could get design wins on more commodity chips they could turn that into a money maker vs. an endless stream of debt for new nodes.

 

[adroc_thurston]

Doesn't make NVL architecture any more suited to win server business over the competition

Well, DMR is kind of that.
It's the same AMD bootleg of CCDs + NUCA, just without b.L. it sucks.

 

[poke01]

The compute tile in PTL also has an image processor, a 50 TOPs NPU, display outputs, memory controller and a fat bus that reaches the iGPU chiplet.
You can't compare it to the Zen 5 CCDs that only have the CPU cores and IF

Finally some one actually responds with sense. People keep on comparing PTL or M5 SoC to a Zen5 CCD when the CCD is bare bones

 

[gdansk]

Finally some one actually responds with sense. People keep on comparing PTL or M5 SoC to a Zen5 CCD when the CCD is bare bones

It's a bad technical comparison
But regarding economics... maybe logical?
The barebones CCD approach seems - ironically - to be the best approach for manufacturing processes with bad yields (like 18A, allegedly).

Related: Did we get approximate sizes for WCL yet?

 

[Joe NYC]

Fab Space ain't cheap. Take the space used up by 10 nm for newer nodes.

Fab52 for 18A is a new fab. Only Ireland fab may have been converted from 10nm (non-EUV) to EUV.

 

[Geddagod]

Edit: NVM I burned a few trees and figured it out.

is this a euphemism for searching it up on chat gpt? Cuz if so that's fire lol

 

[Thunder 57]

is this a euphemism for searching it up on chat gpt? Cuz if so that's fire lol

I'm not sure what it means but I have a pretty good idea and that isn't what came to mind.

 

[Geddagod]

I'm not sure what it means but I have a pretty good idea and that isn't what came to mind.

what's your idea lol

 

[techjunkie123]

is this a euphemism for searching it up on chat gpt? Cuz if so that's fire lol

Yup that's exactly what I meant.

I'm not sure what it means but I have a pretty good idea and that isn't what came to mind.

Trees, not plants 😉.

 

[511]

Related: Did we get approximate sizes for WCL yet?

75-80mm2

 

[OneEng2]

It's a bad technical comparison
But regarding economics... maybe logical?
The barebones CCD approach seems - ironically - to be the best approach for manufacturing processes with bad yields (like 18A, allegedly).

Related: Did we get approximate sizes for WCL yet?

That was my thought too. Smaller die give you better yields. The trick is to keep performance up when you have so many die to die interfaces to maneuver around. I agree that the highest performant solution is monolithic, it just doesn't seem to be the right answer today if you want your company to remain solvent .

 

[Tarkin77]

first PTL "Review" online: https://www.wired.com/story/intel-panther-lake-core-ultra-series-3-review/

 

[DavidC1]

first PTL "Review" online: https://www.wired.com/story/intel-panther-lake-core-ultra-series-3-review/

Man. Those are decent CPU scores actually.

The highest Arrowlake-H score in ST is 130, same as the Pantherlake one. And 1285 points in MT is beating the best ARL-H by 7%, which needs 80W/100W settings to do so. Highest AMD HX 370 is 121 points for ST, and 1299 points for MT. That means it's 3x the MT performance of Lunarlake. Looks like Techpowerup saying 5-10% per clock for Cougar Cove over Lion Cove is right.

 

[regen1]

https://www.notebookcheck.net/Asus-ZenBook-Duo-UX8407-convertible-review-Intel-Panther-Lake-is-a-game-changer.1212285.0.html

 

[mvprod123]

Blender GPU score

 

[techjunkie123]

first PTL "Review" online: https://www.wired.com/story/intel-panther-lake-core-ultra-series-3-review/

Man. Those are decent CPU scores actually.

The highest Arrowlake-H score in ST is 130, same as the Pantherlake one. And 1285 points in MT is beating the best ARL-H by 7%, which needs 80W/100W settings to do so. Highest AMD HX 370 is 121 points for ST, and 1299 points for MT. That means it's 3x the MT performance of Lunarlake. Looks like Techpowerup saying 5-10% per clock for Cougar Cove over Lion Cove is right.

If anything, the 14 inch chassis achieves ST scores way slower than the M5, and comparable MT and GPU performance. The M5 is a 150 mm^2 chip. And the CPU is supposed to be on a more advanced process.

If macs could run windows properly, I know what hardware I'd be getting. I know it's not happening, but the gap is widening rather than shrinking. edit: AMD is in a similar position at least right now.

 

[DavidC1]

https://www.notebookcheck.net/Asus-ZenBook-Duo-UX8407-convertible-review-Intel-Panther-Lake-is-a-game-changer.1212285.0.html

Wow. It's beating Arrowlake average by not a small amount either in CPU.

If anything, the 14 inch chassis achieves ST scores way slower than the M5, and comparable MT and GPU performance. The M5 is a 150 mm^2 chip. And the CPU is supposed to be on a more advanced process.

If macs could run windows properly, I know what hardware I'd be getting. I know it's not happening, but the gap is widening rather than shrinking. edit: AMD is in a similar position at least right now.

The CPU divisions are in a turmoil right now, but this is a clear improvement despite not expecting to be more than basically a Tick in terms of CPU uarch. It's beating the predecessor Zenbook Duo in MT anywhere from 15-50%, primarily because PTL is super stable while ARL-H drops after few runs.

and comparable MT and GPU performance.

30-40% is not "comparable" MT performance.

 

[511]

If anything, the 14 inch chassis achieves ST scores way slower than the M5, and comparable MT and GPU performance. The M5 is a 150 mm^2 chip. And the CPU is supposed to be on a more advanced process.

We can't really do a fair comparison cause PTL has way more IO than M5 and is Chiplet as for ST i agree both Intel/AMD have not imrpoved ST performance by a bit also M4 was 168mm2 where did you get 150mm2 for M5?

If macs could run windows properly, I know what hardware I'd be getting. I know it's not happening, but the gap is widening rather than shrinking. edit: AMD is in a similar position at least right now.

Mac can't support long term software their chips are tailor made for their own software true SW/HW Co design

 

[DavidC1]

We can't really do a fair comparison cause PTL has way more IO than M5 and is Chipltet as for ST i agree both Intel/AMD have not imrpoved ST performance by a bit.

You can't do ISO comparisons mm2 for Intel as it's more important for them to fill their fabs. Even if it costs twice the area it should still be on their fabs.

They aren't out of the gutter but Pantherlake is a clear improvement over their predecessor in all areas, and right now the best x86 laptop chip.

 

[511]

You can't do ISO comparisons mm2 for Intel as it's more important for them to fill their fabs. Even if it costs twice the area it should still be on their fabs.

we can only do that on fully monolithic chip on the same node and this aint that

They aren't out of the gutter but Pantherlake is a clear improvement over their predecessor in all areas, and right now the best x86 laptop chip.

the Single Core performance is my only issue with these rn no improvement

 

[DavidC1]

we can only do that on fully monolithic chip on the same node and this aint that

the Single Core performance is my only issue with these rn no improvement

It's same or slightly faster, meaning we got the more important part, which is uarch gains. TPU is correct with 5-10% gains for Cougar Cove.

Computerbase test is showing the iGPU is beating Strix Point 50-70% at same power envelopes.

Have you noticed that the memory bandwidth and latency has improved greater than memory speed changes? They improved that a noticeable amount. Memory bandwidth went up 30%, and latency went down by 10ns. They executed on this.

 

[511]

Have you noticed that the memory bandwidth and latency has improved greater than memory speed changes? They improved that a noticeable amount. Memory bandwidth went up 30%, and latency went down by 10ns. They executed on this.

Well it's even greater than that

Ultra X9 388H性能分析：重铸Intel荣光_哔哩哔哩_bilibili

在刚刚过去的CES上，Intel放出豪言壮语：第三代酷睿Ultra处理器（代号Panther Lake）的多核性能比ARL-H更强，续航表现比LNL更好，集显性能直逼4050！这要真能落实下来，那可太厉害了，今天我拿到了这台搭载第三代英特尔酷睿Ultra处理器的联想 小新Pro16 GT AI元启版，不仅散热给力，内存频率拉满，电池还升级到了99.9Wh，可以好好测测看Intel到底是不是吹牛了！, 视频播放量 61041、弹幕量 608、点赞数 4503、投硬币枚数 1232、收藏人数 870、转发人数 860, 视频作者 笔吧评测室, 作者简介...

www.bilibili.com

 

[DavidC1]

Well it's even greater than that

Ultra X9 388H性能分析：重铸Intel荣光_哔哩哔哩_bilibili

在刚刚过去的CES上，Intel放出豪言壮语：第三代酷睿Ultra处理器（代号Panther Lake）的多核性能比ARL-H更强，续航表现比LNL更好，集显性能直逼4050！这要真能落实下来，那可太厉害了，今天我拿到了这台搭载第三代英特尔酷睿Ultra处理器的联想 小新Pro16 GT AI元启版，不仅散热给力，内存频率拉满，电池还升级到了99.9Wh，可以好好测测看Intel到底是不是吹牛了！, 视频播放量 61041、弹幕量 608、点赞数 4503、投硬币枚数 1232、收藏人数 870、转发人数 860, 视频作者 笔吧评测室, 作者简介...

www.bilibili.com

Lunarlake is lower latency than Arrowlake, that's why I said 10ns.

Look at that, Darkmont is beating Zen 5 per clock, and is near Lion Cove. Within 1%. The LPE cores are faster than Lunarlake too per clock.