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.

 

[gdansk]

Going by the latest leak from wccftech (i know)

That a 9% or 13% clockspeed deficitt depeding on if you compare aginst 14900K or KS, that you need to overcome with purely IPC

For Intel to reach +20% ST performance over 13900K they need a IPC increase of ~31%
For Intel to reach +20% ST performance over 13900KS they need a IPC increase of ~35%

There was another leak (now deleted) claiming 5.7GHz

 

[Wolverine2349]

Imagine a hypothetical CPU with only 4 Skymont cores in a budget school laptop. Would run nicely and last forever.

I guess the design of the CPU prevents that but it's an interesting thought experiment.

Or how about a hypothetical 12 core Skymont only desktop CPU. Best option for non hybrid gaming CPU for more than 8 cores in one tile/die/ring? Or does the design of it prevent such a thing even if Skymont IPC really does achieve Golden Cove levels and high 4.XGHz clocks?

Or would real world performance not be that great with the deign by itself and not supplemented by the Lion Cove cores?

Or other weaknesses (latency???) such as latency of Skymont where it would not perform even clock normalized 4.6GHz close across 100% workloads as Golden Cove?

Or would real world performance at 4.6GHz clock normalized by on par across almost all apps as a hypothetical 4.6GHz all 12 P core 12th Gen GLC with HT disabled?

 

[SiliconFly]

.

 

[Markfw]

They’re not having issues competing with TSMC leading edge processes. Look at the Sierra Forest benchmarks that uses only Intel silicon taking the lead for x86 efficiency in the 250W TDP range against Bergamo / Genoa on TSMC silicon.

NO, not according to your own link. I did the calculations on perf/watt. lower perf/watt for Xeon = not best efficiency

thats 2 x the reason the power is more than 250 watt, its total lower avg than 500

	model	perf	avg watt	perf/watt
bergamo	9754 2p	5609.56​	375.51​	14.93851​
Xeon	6780 2p	4233.57​	321.01​	13.18828​

 

[Wolverine2349]

Looks like Intel is on track, finally.


Like Intel, can't we have a "small" margin of error... like 10%? 🤔


I too remember seeing 5.7GHz & it looked legit.


I think Skymonts on desktops are a very different beast when compared to the ones in lunar lake. In desktop ARL-S, I think they're directly connected to the ring bus with full access to L3. This provides much better performance, lower latency, at the cost of a little bit of efficiency afaik.

A hypothetical 8+32 ARL-S should be powerful enough to swat top-end Zen5 X3D like a bug.

But would a hypothetical 12 core Skymont clocked 4.6GHz all core perform equally across all apps that don't use AVX512 as a hypothetical 12 p core Rapror Lake HT disabled all core 4.6GHz?

 

[H433x0n]

NO, not according to your own link. I did the calculations on perf/watt. lower perf/watt for Xeon = not best efficiency

	model	perf	avg watt	perf/watt
bergamo	9754 2p	5609.56​	375.51​	14.93851​
Xeon	6780 2p	4233.57​	321.01​	13.18828​

you’re selecting the less efficient model and going with a dual socket configuration. I specifically said 250W TDP.

 

[invisible_city]

If you are a CPU enthusiast you should.

I am pretty sure Zen 5 people wanted 20% not 16%. But you are saying 20% and 30% is the same thing.

That makes me doubt your claims.

Here, I will do it for you:
-13 geo mean/watt for Xeon 6780(4233/321)
-15.7 geo mean/watt for EPYC 9754(5905/375)

You claimed 28 for SRF and 21 for EPYC. Again, I doubt your claims, or do a much better job.

What people get up in arms about and what matters is unfortunately orthogonal. But that's an aside, let's focus.


-13 geo mean/watt for Xeon 6780(4233/321)

No LOL. 4233/150 = 28.22. Look at the mean not the IQR. Repeat for 9754

Also if we're going to be ants of the peda variety, it's 6780E; non-E reserved for GNR

edit: bah I was looking at 2P: 2855/(~150-170) vs 3944/(~250-270) = ~17 vs ~14 if we use the rough upper end of both, or using the most conservative numbers 2855/170 vs 3944/250, 17 vs 16

 

[invisible_city]

You don't seem to not understand how it work.

"The dual Xeon 6780E configuration was at 85% the power consumption on average as AMD's flagship Bergamo processor, the EPYC 9754"

While the performance is only 73% as much as the EPYC 9754:

https://phoronix.com/benchmark/result/intel-xeon-6766e-xeon-6780e-sierra-forest-xeon-6-benchmarks/geometric-mean-of-all-test-results-result-composite-ix6x6sfx6b.svgz

Like how do you mess up this hard? Just look at the chart ffs.

Repeat after me:

mean performance/mean_power = ppw

4233/(2* ~150-170) vs 5905/(2 * ~250-270)

 

[Markfw]

Added single chip from the same source. and fixed typo 5609 to 5906. Xeon still lower perf/watt I believe I am using exactly what @invisible_city is suggesting

	model	perf	avg watt	perf/watt
bergamo	9754 2p	5905.56​	375.51​	15.72677​
Xeon	6780 2p	4233.57​	321.01​	13.18828​
bergamo	9754​	3944.38​	225.05​	17.52668​
Xeon	6780	2865.8​	169.1​	16.94737​

 

[H433x0n]

Added single chip from the same source. and fixed typo 5609 to 5906. Xeon still lower perf/watt I believe I am using exactly what @invisible_city is suggesting

	model	perf	avg watt	perf/watt
bergamo	9754 2p	5905.56​	375.51​	15.72677​
Xeon	6780 2p	4233.57​	321.01​	13.18828​
bergamo	9754​	3944.38​	225.05​	17.52668​
Xeon	6780	2865.8​	169.1​	16.94737​

You're conveniently ignoring the 6766E

 

[SiliconFly]

.

 

[Markfw]

You're conveniently ignoring the 6766E

	model	perf	avg watt	perf/watt
bergamo	9754 2p	5905.56​	375.51​	15.72677​
Xeon	6780 2p	4233.57​	321.01​	13.18828​
bergamo	9754	3944.38​	225.05​	17.52668​
Xeon	6780	2865.8​	169.1​	16.94737​
Xeon	6766e	2616.12​	142.1​	18.41042​

It is better on that one. I think the point is that your statement is a little exagaerated I did not check every CPU, One other could even beat that. Could be Genoa, Xeon or Bergamo.

I will not bother to spend the time further.

Edit: and that is one review, chosen by you. Others could be different. I will grant you this is much better than previous Xeons. But Turin is coming. And Intel Xeon Gen 6 I think.

 

[H433x0n]

	model	perf	avg watt	perf/watt
bergamo	9754 2p	5905.56​	375.51​	15.72677​
Xeon	6780 2p	4233.57​	321.01​	13.18828​
bergamo	9754	3944.38​	225.05​	17.52668​
Xeon	6780	2865.8​	169.1​	16.94737​
Xeon	6766e	2616.12​	142.1​	18.41042​

It is better on that one. I think the point is that your statement is a little exagaerated I did not check every CPU, One other could even beat that. Could be Genoa, Xeon or Bergamo.

How is it exaggerated if it is correct? Am I supposed to say it's 2nd best to not exaggerate.. wouldn't that be lying?

I personally think this is good news for both Intel & AMD. This is exactly what was needed to fend off further losses to ARM in the data center.

 

[del42sa]

They have provided ample comparisons for both mobile and desktop implementations:

• on mobile they're comparing with LP-E cores because the Skymont cluster is taking over their role as well
• for the desktop, they provide an IPC comparison with Raptor Cove, in a simulation where they both have access to the same fabric and memory performance

yes but there is a small diference , the slide where they compare it with Crestmont is Skymont LPE with 4MB L2 + 8MB SLC vs Crestmont LPE 2MB L2 with no L3/SLC cache because in Lunar Lake implementation E-core cluster are isolated from P-core cluster, so they can be switched off and save power.

While slide where they compare it to Raptor Cove is clearly not taken from Lunar Lake, but from comparison with Arrow Lake, where Skymont cores has acces to big L3 cache through P-Core cluster, which affect performance significantly.

from chipsandcheese : "Meteor Lake’s L3 cache may not be the biggest or fastest around, but even a mediocre L3 is miles better than nothing at all. Despite using the same core architecture, Meteor Lake’s E-Cores achieved over 30% better IPC than the LPE-Cores."

https://chipsandcheese.com/2024/05/20/comparing-crestmonts-no-l3-hurts/

 

[Markfw]

How is it exaggerated if it is correct? Am I supposed to say it's 2nd best to not exaggerate.. wouldn't that be lying?

I personally think this is good news for both Intel & AMD. This is exactly what was needed to fend off further losses to ARM in the data center.

First, look at my edit. Second, to evaluate ALL CPUs of both and then says they are king is what is needed before you claim 1st place.

But I do agree it is looking better for both.

 

[invisible_city]

Added single chip from the same source. and fixed typo 5609 to 5906. Xeon still lower perf/watt I believe I am using exactly what @invisible_city is suggesting

	model	perf	avg watt	perf/watt
bergamo	9754 2p	5905.56​	375.51​	15.72677​
Xeon	6780 2p	4233.57​	321.01​	13.18828​
bergamo	9754​	3944.38​	225.05​	17.52668​
Xeon	6780	2865.8​	169.1​	16.94737​

Ah, maybe I missed where they stated the avg watt in number, was eyeballing, thanks for summarizing.

 

[gdansk]

For it to be an effective comparison of efficiency it should be a chart across multiple core power limits. And it doesn't look like we have that from anyone 3rd party. I bet we would if Geekerwan cared about hyperscaler cores.

But in any case it's pretty good for Intel 3 to be competitive with TSMC N5.

 

[Hitman928]

You're conveniently ignoring the 6766E

EPYC 9734 would be the comparison point for the Xeon 6766E. Unfortunately it’s not in the review but should be a bit more efficient than the 9734 in their testing.

 

[H433x0n]

EPYC 9734 would be the comparison point for the Xeon 6766E. Unfortunately it’s not in the review but should be a bit more efficient than the 9734 in their testing.

Why would it be more efficient? If anything that SKU would be *less* efficient than the 9754.. It's got a larger power budget per core.

 

[Hitman928]

Why would it be more efficient? If anything that SKU would be *less* efficient than the 9754.. It's got a larger power budget per core.

Lower TDP, lower frequency, less idling cores. The majority of the Phoronix tests don’t max out the cores anyway so having less cores won’t hurt very much but the lower TDP and less idle draw will help efficiency.

 

[invisible_city]

"The dual Xeon 6780E configuration was at 85% the power consumption on average as AMD's flagship Bergamo processor, the EPYC 9754. Or in single socket configurations, the Xeon 6780E was around 66% the power consumption of one EPYC 9754."

edit: Recalculating using the actual average numbers, which I see now, thanks @Markfw

(4233/5905)/(291/375) = .91 for 2P using the average power consumption
(2855/3944)/(169/225) = .96 for 1P using the average power consumption

(4233/5905)/(.85) = .85 for 2P using the stated power consumption
(2855/3944)/(.66) = 1.09 for 1P using the stated power consumption


Not sure where .66 power consumption number comes from for 1P, or .85 for 2P, but whatever.

So it's treading water relative to Bergamo, but much better than Emerald Rapids. Nice step, let's see CWF.

 

[HoveringStyle]

It's higher than 21% in Int, but lower than 71% in FP. You got the calculations wrong.

I didn't, check the graphs at 8:35 and 9:40:

Skymont est. (MTL LPE + 38% INT +68% FP) = 1.90 INT 2.86 FP
Ultra 5 125H E 5.66 INT 6.44 FP @ 3.6GHz = 1.57 INT 1.79 FP
i9 14900K E 5.64 INT 6.01 FP @ 3.6GHz = 1.57 INT 1.67 FP
i5 13500H E 5.11 INT 5.79 FP @ 3.5GHz = 1.46 INT 1.65 FP
Ultra 5 125H LPE 3.44 INT 4.26 FP @ 2.5GHz = 1.38 INT 1.70 FP

 

[invisible_city]

“
The Xeon 6780E had an average of 1.17 Watts per core and a peak of 2.47 Watts per core. Meanwhile the AMD EPYC 9754 had a 1.75 Watts per core and a peak of 3.1 Watts per core. The Xeon Platinum 8592+ Emerald Rapids CPU meanwhile equated to an average of 4 Watts per core and a peak of 6.3 Watts per core.
“

We’re getting some oddities with the reporting, maybe uncore contributing. 6780E is per core half the power use of bergamo (.57 for David )

 

[poke01]

Can we move all the Xeon and stuff to the other Intel thread.

This should be Client CPUs only.

 

[invisible_city]

Back on track: https://wccftech.com/intel-core-ult...con-spotted-clocks-close-5-ghz-rvp-ai-devkit/

~5ghz lunar lake clock