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.)

 

[jpiniero]

That's my point. If a CPU with less cores and lower frequencies has worse battery life is that the comparison is not well made.

125H is probably where all the crap bins go. So it's not unrealistic.

 

[cebri1]

12000 pts at 40W is not a very good result.
6P+8E i7-1370P manages 13,159 at PL1: 55W, PL2: 30W
4P+8E i7-1360P manages 11,072 at PL1: 43W, PL2: 28W

1370P has a boost power limit of 64W, not 55.

Intel® Core™ i7-1370P Processor (24M Cache, up to 5.20 GHz) - Product Specifications | Intel

Intel® Core™ i7-1370P Processor (24M Cache, up to 5.20 GHz) quick reference with specifications, features, and technologies.

www.intel.com

12000 / 40 = 300p/watt
13159 / 64 = 205p/watt

That would make the MTL chip 46% more efficient.

 

[Hitman928]

From NBC tests it look like the 7840HS is set at about 35W, the 7840U score about 12000@30W.

Also notice the evolution from R15 to R23, no need to improve your CPU IPC, "improve" the bench instead, that work much better, same from GB5 to GB6...

Edit : The difference between 125 and 155 in battery life is not significant, in the first test it s 3%, so within error margins.

The 7840HS should score near 14000 at 35W.

Spoiler

 

[TESKATLIPOKA]

1370P has a boost power limit of 64W, not 55.

Intel® Core™ i7-1370P Processor (24M Cache, up to 5.20 GHz) - Product Specifications | Intel

Intel® Core™ i7-1370P Processor (24M Cache, up to 5.20 GHz) quick reference with specifications, features, and technologies.

www.intel.com

12000 / 40 = 300p/watt
13159 / 64 = 205p/watt

That would make the MTL chip 46% more efficient.

OEMs can set It as they want, 64w is the maximum. That laptop had It at 55W.

And there is no 46% better efficiency.
11072/43=257.5
I got 17% higher efficiency vs i7-1360P and this model has less P-cores.
And there is still the question, If It can finish that one CB run until this TDP is active.

 

[Abwx]

The 7840HS should score near 14000 at 35W.

Spoiler

View attachment 90063

Quite possible, the HP Elite Book with a 7840U score about 12k at 25W, this 14k score at 35W is adequate with the power/frequency scaling of PHX.

That being said those MTL numbers do not look good for the 6P + 8E, in principle they only need 6P + 4E to match a 7840, here there s double the E cores count at work for about the same perf and not even a better efficency despite a theorically much improved process, or possibly that they dont use TSMC for the CPU tile.

 

[Hitman928]

Quite possible, the HP Elite Book with a 7840U score about 12k at 25W, this 14k score at 35W is adequate with the power/frequency scaling of PHX.

That being said those MTL numbers do not look good for the 6P + 8E, in principle they only need 6P + 4E to match a 7840, here there s double the E cores count at work for about the same perf and not even a better efficency despite a theorically much improved process, or possibly that they dont use TSMC for the CPU tile.

CPU tile should be Intel 4 process.

 

[dullard]

I got 17% higher efficiency vs i7-1360P and this model has less P-cores.

25% better than the 1370P. 17% better than the 1360P. That is right in line with Intel's claim of 20% perf/watt gain for Intel 4.

Anyone expecting more is not being honest with their expectations.

 

[Hitman928]

25% better than the 1370P. 17% better than the 1360P. That is right in line with Intel's claim of 20% perf/watt gain for Intel 4.

Anyone expecting more is not being honest with their expectations.

How do you calculate 25% better than the 1370p?

 

[dullard]

How do you calculate 25% better than the 1370p?

155H: supposedly 12000 points in 40 W. 12000 points / 40 W = 300.0 points per Watt.
1370P: 13159 points / 55 W = 239.3 points per Watt
1360P: 11072points / 43 W = 257.5 points per Watt
I'm just using the numbers TESKATLIPOKA posted here: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...akes-discussion-threads.2606448/post-41120221

300 / 239.3 = 1.25 or 25% better in performance per Watt.

That is pretty much in line with Intel's claim:

 

[Hitman928]

155H: supposedly 12000 points in 40 W. 12000 points / 40 W = 300.0 points per Watt.
1370P: 13159 points / 55 W = 239.3 points per Watt
1360P: 11072points / 43 W = 257.5 points per Watt
I'm just using the numbers TESKATLIPOKA posted here: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...akes-discussion-threads.2606448/post-41120221

300 / 239.3 = 1.25 or 25% better in performance per Watt.

Gotcha, sorry, I was confusing 1370p with the 13705h I posted above.

Edit: I believe the 13705h is the better comparison as it is RPL versus the 1370p appears to be ADL. Intel ARK says the 1370p is RPL but notebookcheck says it is ADL and if you look at the supported PCIE version, socket number, and performance, it's clearly ADL.

Edit2: I take that back, I think, maybe. It's on a new socket but I think that's just to enable pcie5 and has a much higher starting PL level though not sure how to explain it's much better efficiency than the 1370p if they are both RPL with the same cores and threads.

Edit3: If they are both RPL, then the 1370p result @TESKATLIPOKA pointed to is from running at it's base PL1 power level of 30W. That would match the 13705h performance and efficiency.

 

[trivik12]

https://twitter.com/x/status/1733141912498934068

Not sure if this is good?

 

[Hitman928]

https://twitter.com/x/status/1733141912498934068

Not sure if this is good?

Posted on the last page but if true, not really good. Slower than RPL-H and not any more efficient. If the bilibili score of ~15k is legit at the same 40W, then that would be a huge upgrade in efficiency and would pass even the 7840hs in that regard. However, the bilibili leak gives no power numbers so who knows how much it was drawing to get that ~15k, if that score is even real to begin with, lol.

 

[Abwx]

Posted on the last page but if true, not really good. Slower than RPL-H and not any more efficient. If the bilibili score of ~15k is legit at the same 40W, then that would be a huge upgrade in efficiency and would pass even the 7840hs in that regard. However, the bilibili leak gives no power numbers so who knows how much it was drawing to get that ~15k, if that score is even real to begin with, lol.

In Bilibili what is not logical is that MTL go from slower than RPL in CB R15 to little slower in R20 and then quite faster in R23, as if the reviewer started the test with the latter and the chip went throttling when he got to the latest test.

Anyway from 12k to 15k there s inherently 60-70% more power required, that would be the difference between 40W and the usual 64W PL1 used by Intel for some chips, HXL stated that it s 12k@40W.

 

[Saylick]

OEMs can set It as they want, 64w is the maximum. That laptop had It at 55W.

And there is no 46% better efficiency.
11072/43=257.5
I got 17% higher efficiency vs i7-1360P and this model has less P-cores.
And there is still the question, If It can finish that one CB run until this TDP is active.

Intel could still, in theory, claim 50% better efficiency over Raptor Lake but at idle or something where the compute tile is shut off. At load, I agree that it should only be about 20% better perf/W since it's the same core with a process improvement.

 

[Hitman928]

In Bllibili what is not logical is that MTL go from slower than RPL in CB R15 to little slower in R20 and then quite faster in R23, as if the reviewer started the test with the latter and the chip went throttling when he got to the latest test.

Anyway from 12k to 15k there s inherently 60-70% more power required, that would be the difference between 40W and the usual 64W PL1 used by Intel for some chips, HXL stated that it s 12k@40W.

I really think, if it's legit at all, the Billibili tests were sometimes run on battery power, that's the only way I can explain some of their results.

~15k at 64W would make sense and correlate to the other leak of ~12k at 40W. Both results would basically match up with RPL-H performance and efficiency.

 

[TESKATLIPOKA]

155H: supposedly 12000 points in 40 W. 12000 points / 40 W = 300.0 points per Watt.
1370P: 13159 points / 55 W = 239.3 points per Watt
1360P: 11072points / 43 W = 257.5 points per Watt
I'm just using the numbers TESKATLIPOKA posted here: http://www.portvapes.co.uk/?id=Latest-exam-1Z0-876-Dumps&exid=thread...akes-discussion-threads.2606448/post-41120221

300 / 239.3 = 1.25 or 25% better in performance per Watt.

That is pretty much in line with Intel's claim:
View attachment 90066

You are comparing 55W vs 40W and then claiming 25% efficiency? That's wrong.
The chart from @Hitman928 shows 13705H at 40W scoring ~12579 pts, which is better than this MTL.
Without knowing PL1, PL2 or the length of CB's run this is just an inaccurate comparison.

 

[Hitman928]

If HXL's leak is accurate and the Bllibili is accurate at 64W, then I'll have to give credit to @adroc_thurston in calling the lack of improvement in MTL for a while now despite some leakers (e.g. Raichu) claiming huge efficiency improvements. At this point the launch event will either be entertaining in that the leaks will be proven horribly inaccurate, or entertaining in how Intel tries to spin it's latest part being no improvement over it's last gen parts but still more expensive. Either way I expect a lot of focus on the graphics tile and AI, AI, AI!

 

[Abwx]

I really think, if it's legit at all, the Billibili tests were sometimes run on battery power, that's the only way I can explain some of their results.

~15k at 64W would make sense and correlate to the other leak of ~12k at 40W. Both results would basically match up with RPL-H performance and efficiency.

That s also in line with MLID who displayed a curve starting at 13k@50W and reaching 15k at about 65W, so seems that his leak was genuine after all...

 

[Saylick]

If HXL's leak is accurate and the Bllibili is accurate at 64W, then I'll have to give credit to @adroc_thurston in calling the lack of improvement in MTL for a while now despite some leakers (e.g. Raichu) claiming huge efficiency improvements. At this point the launch event will either be entertaining in that the leaks will be proven horribly inaccurate, or entertaining in how Intel tries to spin it's latest part being no improvement over it's last gen parts but still more expensive. Either way I expect a lot of focus on the graphics tile and AI, AI, AI!

Heh, we'll see what Intel's NPU is capable of, TOPS-wise. If it's only 13 TOPS then AMD was wise to juice up the clocks on PHX -> HWK so they could maintain a marketing advantage there. I agree that the remaining advantages for MTL over HWK would be the iGPU and the dual encoders/decoders, but whether or not that matters as much as having better perf/W is yet to be seen. I personally don't see much benefit to having dual encoders/decoders for a mobile product like this and would much rather have better perf/W on the CPU side, but I'm sure the calculus could be different for someone else.

 

[adroc_thurston]

then I'll have to give credit to @adroc_thurston in calling the lack of improvement in MTL for a while now despite some leakers (e.g. Raichu) claiming huge efficiency improvements

Yeah it's dog everyone knew it's dog ever since Intel shipped MTL RVPs.
It was a lot more buggy than it is now, too (even if it costs a C0 stepping on a laptop part of all things).
If you want to talk about maybe good Intel part, focus on LNL-M.

or entertaining in how Intel tries to spin it's latest part being no improvement over it's last gen parts but still more expensive

I mean they'll just focus on the iGP Timespy numbers and that's it.

Either way I expect a lot of focus on the graphics tile and AI, AI, AI!

They can't really talk AI since it doesn't pass the MS cutoff point.
iVPU is too tiny and GPU has no XMX systolics to boost the TOPS number to reach the cutoff point.

 

[H433x0n]

If HXL's leak is accurate and the Bllibili is accurate at 64W, then I'll have to give credit to @adroc_thurston in calling the lack of improvement in MTL for a while now despite some leakers (e.g. Raichu) claiming huge efficiency improvements. At this point the launch event will either be entertaining in that the leaks will be proven horribly inaccurate, or entertaining in how Intel tries to spin it's latest part being no improvement over it's last gen parts but still more expensive. Either way I expect a lot of focus on the graphics tile and AI, AI, AI!

These laptops being compared have different battery sizes and chassis. We need an apples/apples test with both of these chips in the same chassis like what JarrodTech did with 13700H vs 7840HS.

The only improvement we should expect is what is shown below.

 

[desrever]

View attachment 90068

The chips need to run at 4.5ghz+, this chart is useless?

 

[H433x0n]

The chips need to run at 4.5ghz+, this chart is useless?

There’s Crestmont cores with an fmax <4.5ghz and Redwood Cove won’t be running >4.5ghz in the 28-45W range.

This is also an ARM core in the chart so it’s not exactly apples/apples but it gives you an idea of what to expect.

 

[TESKATLIPOKA]

These laptops being compared have different battery sizes and chassis. We need an apples/apples test with both of these chips in the same chassis like what JarrodTech did with 13700H vs 7840HS.

The only improvement we should expect is what is shown below.

View attachment 90068

If this chart was comparable to reality, then MTL at 3.3GHz would consume ~40% of Raptor. So If Raptor needed 45W MTL would need ~16W at 3.3GHz.

@H433x0n yeah, It's simplified, but I only wanted to show tha the difference is big.

 

[Hitman928]

Yeah it's dog everyone knew it's dog ever since Intel shipped MTL RVPs.
It was a lot more buggy than it is now, too (even if it costs a C0 stepping on a laptop part of all things).
If you want to talk about maybe good Intel part, focus on LNL-M.

I mean they'll just focus on the iGP Timespy numbers and that's it.

They can't really talk AI since it doesn't pass the MS cutoff point.
iVPU is too tiny and GPU has no XMX systolics to boost the TOPS number to reach the cutoff point.

Yeah, I don't expect MTL to be much of anything in AI performance, but from a marketing perspective, something is better than nothing so comparing to RPL, it will be used as a big plus. With AI fever right now too, Intel won't pass up the opportunity to tell the world they have it as well. Marketing will find whatever apps and examples they can to showcase how their AI powered chip can enhance the user experience but will shy away from any real spec or performance comparisons.