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

[Tigerick]

Wildcat Lake (WCL) Preliminary Specs

Intel Wildcat Lake (WCL) is upcoming mobile SoC replacing ADL-N. 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 Q2/Computex 2026. In case people don't remember AlderLake-N, I have created a table below to compare the detail specs of ADL-N and WCL. Just for fun, I am throwing LNL and upcoming Mediatek D9500 SoC.

		Intel Alder Lake - N	Intel Wildcat Lake	Intel Lunar Lake		Mediatek D9500
Launch Date		Q1-2023	Q2-2026 ?	Q3-2024		Q3-2025
Model		Intel N300	?	Core Ultra 7 268V		Dimensity 9500 5G
Dies		2	2	2		1
Node		Intel 7 + ?	Intel 18-A + TSMC N6	TSMC N3B + N6		TSMC N3P
CPU		8 E-cores	2 P-core + 4 LP E-cores	4 P-core + 4 LP E-cores		C1 1+3+4
Threads		8	6	8		8
Max Clock		3.8 GHz	?	5 GHz
L3 Cache		6 MB	?	12 MB
TDP		7 W	Fanless ?	17 W		Fanless
Memory		64-bit LPDDR5-4800	64-bit LPDDR5-6800 ?	128-bit LPDDR5X-8533		64-bit LPDDR5X-10667
Size		16 GB	?	32 GB		24 GB ?
Bandwidth			~ 55 GB/s	136 GB/s		85.6 GB/s
GPU		UHD Graphics		Arc 140V		G1 Ultra
EU / Xe		32 EU	2 Xe	8 Xe		12
Max Clock		1.25 GHz		2 GHz
NPU		NA	18 TOPS	48 TOPS		100 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.

 

[lopri]

70C load temperature at 250W seems a bit unrealistic to me. I wonder what part of die software is reporting.

 

[Thunder 57]

For the moment, it's DOA for gamers. There has to be significant gains waiting on microcode and OS updates, because something is as broken as ARC at the moment. I dislike using geomean, here's why -

View attachment 110261

20 games with double digit losses, 15 where it's basically 20-50%. That's- "Yes police, I'd like to report a murder" territory. If she'd added F1 '24 that's another 20%+ beatdown. This can't be all there is without overclocking and budget breaking stupidly expensive ram, it can't be...

You should list your sources. As soon as I read "she" I knew it was likely Nada at Techtesters. You are right about F1 though. And yes, geomean can "lie". What's the saying? "Lies, damned lies, and statistics"?

Link to video

 

[poke01]

https://twitter.com/x/status/1849667110277959745

is this good compared to the 9950X? Has your 9950X reached 50K?
While it does lose HT, 285K is on a newer node and uses more power to achieve this.
@Det0x

 

[DrMrLordX]

What if neither? Those aren't my main apps I use daily.

Then you're using your PC wrong!

...

ahem, sorry. Where was I?

@Hulk
And here we are with the first round of reviews where, unfortunately, it looks like some of those early GB6 leaks showing Arrow Lake to be putrid may have been accurate. Arrow Lake is all over the place, and we have no idea how much of that is Windows being stupid (except for the Phoronix review which still shows some signs of the Linux scheduler doing things in a less-than-optimal way). Plus MLID was right about instability on the 285k.

Needs updates and more testing, but for now Arrow Lake looks pretty bad unless for some reason it's really good in one application or game where it seems to perform well. Overall, hard pass, can't see why anyone would be picking Intel for DiY desktop right now, especially after all the Raptor Lake nonsense.

 

[controlflow]

https://twitter.com/x/status/1849667110277959745

is this good compared to the 9950X? Has your 9950X reached 50K?
While the it does lose HT, 285K is on a newer node and uses power to achieve this.
@Det0x

IMO this tuned result of 47k @ 215W is more impressive. 61 ns memory latency too (not that it is relevant to CB).

Intel Core Ultra 9 285k 'Arrow Lake' Discussion/News ("15th gen") on LGA-1851

it is free basically. i just use fake dollars conjured up out of nowhere. just add it to the global credit card bill. Global public debt is projected to go above $100 trillion this year. :cry: I think loyalty was buying a bulldozer chip, back in the day. but that time is in the past. it...

forums.overclockers.co.uk

 

[DAPUNISHER]

Still there are going to be people who watched de8bauer's video and will go out and drop several grand on a new build just to see if they can tweak this thing to the high heavens just to match last gen performance lmao.

It's a hobby after all, so more power to them I guess? I call them the hot rod crowd. It is going to do for raptor and alder what Zen 5 did for Zen 4 i.e. increase sales. Osborne effect is gone for Intel gamers

You should list your sources. As soon as I read "she" I knew it was likely Nada at Techtesters. You are right about F1 though. And yes, geomean can "lie". What's the saying? "Lies, damned lies, and statistics"?

Link to video

Counter point: Evidently, I have people for that?

Throw in Assetto Corsa and that's a 35% thrashing. Her Borderlands 3 numbers on high need to be rerun because Tom's has the 3D destroying arrow at Ultra by 30%. That's a 55% swing; makes me wonder if there is a windows issue on her testbed. I thought it has been well established BL3 likes the v-cache too.

 

[Thunder 57]

It's a hobby after all, so more power to them I guess? I call them the hot rod crowd. It is going to do for raptor and alder what Zen 5 did for Zen 4 i.e. increase sales. Osborne effect is gone for Intel gamers

Counter point: Evidently, I have people for that?

Throw in Assetto Corsa and that's a 35% thrashing. Her Borderlands 3 numbers on high need to be rerun because Tom's has the 3D destroying arrow at Ultra by 30%. That's a 55% swing; makes me wonder if there is a windows issue on her testbed. I thought it has been well established BL3 likes the v-cache too.

I pretty much knew since I skipped through her video earlier. She was comparing the 265k while everyone else was testing the 285k so I thought i'd take a look. When considering cost the 265k doesn't seem nearly as bad as the 285k. Then again, I haven't looked into motherboard prices but have read they are up there.

 

[gdansk]

https://twitter.com/x/status/1849667110277959745

is this good compared to the 9950X? Has your 9950X reached 50K?
While it does lose HT, 285K is on a newer node and uses more power to achieve this.
@Det0x

Looks like it https://hwbot.org/submission/5653223

Edit: Also, I'm not sure how having 50% more cores is a disadvantage in this workload. When did SMT become that much of an impact?

 

[TESKATLIPOKA]

With SMT they would had gained the equivalent of 2P assuming a paltry 25% SMT gain, that s not negligible, eventualy they ll get back to this approach in their next designs as the bigger a core the more the SMT relevance.

I think we can agree CB 2024 is the best case for Arrow Lake.

285K P-core at 4GHz -> 96 points
285K E-core at 4GHz -> 84 points ((1444-96)/16)
14900K P-core at 4GHz -> 90 points
14900K P-core at 4GHz + HT -> 117 points
14900K E-core at 4GHz -> 63 points ((1100-90)/16)
9700X 1C1T at 4GHz -> 94 points
9700X 1C2T at 4GHz -> 125 points

285K P-core with HT at 4GHz -> ~125 points
8 cores with enabled HT at 4GHz would provide additional 231 points. I won't bother calculating the gains at full speed, we don't need It.
Conclusion, enabling HT would allow additional ~10-12% in CB 2024, so it would win in most nT benches, but honestly, I can't say I really miss HT, I would rather see more E-cores.

 

[Det0x]

https://twitter.com/x/status/1849667110277959745

is this good compared to the 9950X? Has your 9950X reached 50K?
While it does lose HT, 285K is on a newer node and uses more power to achieve this.
@Det0x

I compared numbers and gave my opinion here

TLDR

Arrow Lakes seems stronger in MT than in ST, compared to Zen5 🤯
Also seems like 3dmark profiler and Cinebench R24 is the strongest for this cpu architecture when we compare the few LN2 uploaded to hwbot

Overclocking "Arrow Lake" - 285K, 265k, 245k...

😇 Let the games begin.....

www.overclock.net

But it should also be noted, sugi is not getting average samples..
Everything he shows is top 0.000001% 👍

 

[gdansk]

Extrapolating from earlier IPC comparison and comparing clock rates to the OC results posted above
A non-SMT Zen 5 would yield 1588 * (57/40) * 16 = 36,204
Arrow Lake would yield 1459 * (54/40) * 16 + 1702 * (58/40) * 8 = 51,257 very close to the observed

So apparently SMT has insane yield here. Over 40%...

Spoiler: Zen 5 vs Arrow Lake CB23 IPC

View attachment 110259

 

[DAPUNISHER]

I pretty much knew since I skipped through her video earlier. She was comparing the 265k while everyone else was testing the 285k so I thought i'd take a look. When considering cost the 265k doesn't seem nearly as bad as the 285k. Then again, I haven't looked into motherboard prices but have read they are up there.

Something went massively wrong with this launch. It was a terrible idea in the first place, to launch amidst all the turmoil, restructuring, and layoffs. "It hurts itself in its confusion" Did departing employees intentionally gaslight higher ups? I ask because Leo of KitGuru had a 7 on 1 teams call including Robert Hallock. He said they seemed confused and very serious.

 

[DrMrLordX]

"It hurts itself in its confusion"

Seems appropriate somehow.

 

[511]

How does It help them exactly, when their foundries will no longer make any client CPUs for them? Not sure how long they plan to continue producing Raptor Lake.

Raptor lake is eternal until a good Intel Node launches (18A) with a good design Intel 7/RPL despite being a power hog is a different beast something Intel is having difficulty surpassing 🤣

 

[511]

Then you're using your PC wrong!

...

ahem, sorry. Where was I?

@Hulk
And here we are with the first round of reviews where, unfortunately, it looks like some of those early GB6 leaks showing Arrow Lake to be putrid may have been accurate. Arrow Lake is all over the place, and we have no idea how much of that is Windows being stupid (except for the Phoronix review which still shows some signs of the Linux scheduler doing things in a less-than-optimal way). Plus MLID was right about instability on the 285k.

Needs updates and more testing, but for now Arrow Lake looks pretty bad unless for some reason it's really good in one application or game where it seems to perform well. Overall, hard pass, can't see why anyone would be picking Intel for DiY desktop right now, especially after all the Raptor Lake nonsense.

Wait till 14900K drops to 350$ and people will line up to buy it 🤣

 

[gdansk]

🤔 Despite all their differences it's actually amazing how close the 9950X and 285K ended up. Even down to the questionable slides about gaming. Intel and AMD are really in sync!

 

[Timorous]

IDK if this refers to Intel teams or you just don’t care 😝

I wish IDC still posted.

 

[511]

I wish IDC still posted.

Same IDC that is responsible for amazing P cores

 

[DAPUNISHER]

Raptor lake is eternal until a good Intel Node launches (18A) with a good design Intel 7/RPL despite being a power hog is a different beast something Intel is having difficulty surpassing 🤣

I did not have someone using Raptor Lake and eternal in the same sentence on my bingo card.

Most of the reviewers seem almost somber. I don't think I've ever witnessed this before.

 

[gdansk]

Most of the reviewers seem almost somber. I don't think I've ever witnessed this before.

It's the realization that this stagnation won't be fixed for 2 years.

 

[naukkis]

You're not gonna shave many cycles out of things not L3. Like 4c L1 is the norm.
They do need to make a better L3$, the current one is absolutely dogwater.

E-cores L1 latency is 3 cycles. Except all future high-ipc designs to have that 3-cycle L1 load latency. For Lion Cove removal of HT probably saved that one cycle from 5 to 4 - but yet they don't have backend to fully take advantage from that lowered load latency. They need to redo whole backend to single-thread optimized to really gain advantage from removing HT.

 

[Timorous]

Same IDC that is responsible for amazing P cores

.... No, a poster who used to post years ago who went by IDC (I dont care). He had excellent contributions.

 

[DavidC1]

All the hype of Skymont and it's only 18% more efficient than Gracemont. N3B vs Intel 7 and all they got was 18% improvement?

Arrowlake as a chip is quite a bit more efficient than Raptor Lake, but nobody cares do they? Why cause the peak performance is not much better, sometimes even worse. Lion Cove got 7% in the same case.

The peak performance in this case is easily over 30% better.

It's the realization that this stagnation won't be fixed for 2 years.

Maybe. What they do from here will really determine the fate of x86 and Intel.

It does explain though why Intel themselves were so quiet and didn't mention Arrowlake. Perhaps Pantherlake is going to be better at least.

 

[controlflow]

I wonder if the fact that the E core clusters are now interleaved between the P cores on the ring presents additional scheduling complexity. The topology of the compute tile is pretty different than anything else before it. There are definitely some workloads where threads are incorrectly ending up on E cores or bouncing between cores. Some of that is probably on Windows but it also seems like Intel's own thread director isn't quite working right for ARL.

 

[naukkis]

I wonder if the fact that the E core clusters are now interleaved between the P cores on the ring presents additional scheduling complexity. The topology of the compute tile is pretty different than anything else before it. There are definitely some workloads where threads are incorrectly ending up on E cores or bouncing between cores. Some of that is probably on Windows but it also seems like Intel's own thread director isn't quite working right for ARL.

Ring arrangement doesn't affect scheduling, it only balances ring load as e-core cluster bandwidth requirement is now more than P-cores isolating E-cores to one end of ring would lead to unbalances ring traffic.