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

Tigerick · Aug 22, 2022

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.)

igor_kavinski · Mar 1, 2024

Intel keeps forgetting that in the real world, people have to pay electricity bills.

H433x0n · Mar 1, 2024

igor_kavinski said:
Umm...that just looks like Intel's desktop-turned-mobile silicon dominating Hawk Point beyond its intended TDP.

But yeah, the lower TDP performance of Hawk Point is very nice.

It’s the same 6P+8E silicon as the 155H, just better binned. I wouldn’t really consider that desktop silicon.

igor_kavinski · Mar 1, 2024

H433x0n said:
It’s the same 6P+8E silicon as the 155H, just better binned. I wouldn’t really consider that desktop silicon.

I think it is this one: https://www.notebookcheck.net/Deskt...ture-up-to-Core-i9-CPUs.709900.0.html#9712846

At least the compute die of 185H is the performance part in that slide.

Hulk · Mar 1, 2024

Wow, there are some serious quality differences in Intel 4 silicon. 155H and 185H are the same dies, only differing in binning and the v/f curves are obviously quite different as the 185H is significantly more efficient. Assuming this data is legit.

adroc_thurston · Mar 1, 2024

Hulk said:
Wow, there are some serious quality differences in Intel 4 silicon.

Been like that since TGL-U FYI.
Maybe earlier even, just that I remember 1135G7 being tangibly worse than 1165G7.

H433x0n · Mar 1, 2024

Hitman928 said:
Thanks for find that. In their table, they say they are setting PL1 = PL2 for their power testing, but their power consumption graph with their score for repeated tests doesn't seem to agree:

This is the 185h over multiple runs:

View attachment 94690

Corresponding power consumption:

View attachment 94691

So with a base power of 70W and boost power of 95 W for roughly 30% of each run (looking at the later runs), the 185h scores ~18350.

Then, in the table and on their performance over power chart, they show the 185h scoring the same amount with PL1=PL2=65W.

View attachment 94692

There's no possible way this is accurate unless the PL2 isn't being enforced correctly and it is still boosting as before.

Edit: Same applies for the 8945hs in the video.

Why? You can limit power in armoury crate to plot these graphs.

poke01 · Mar 1, 2024

It would very interesting to see where Lunar Lake lands, considering both M3 and Lunar are N3B, we can easily compare geekbench scores. Let's hope they match Apple but its looking like its a going be under M3?

SiliconFly · Mar 2, 2024

adroc_thurston said:
at 90w.

Hilarious! ROTFLOL!!! 👏👏👏

Ultra 9 beats Ryzen 9 at just 45W!!! See the graph...

SiliconFly · Mar 2, 2024

poke01 said:
It would very interesting to see where Lunar Lake lands, considering both M3 and Lunar are N3B, we can easily compare geekbench scores. Let's hope they match Apple but its looking like its a going be under M3?

Lunar Lake, being a low-power SoC, should be a tad slower than MTL i think. And in no way it should be comparable to M3 by any standards. It shouldn't even be compared to ARL actually.

TESKATLIPOKA · Mar 2, 2024

SiliconFly said:
Hilarious! ROTFLOL!!! 👏👏👏

Ultra 9 beats Ryzen 9 at just 45W!!! See the graph...

I am looking at the graph and Ultra 9 performs the same at 45W as R9.

You should double-check the facts before you post some nonsense.

coercitiv · Mar 2, 2024

TESKATLIPOKA said:
I am looking at the graph and Ultra 9 performs the same at 45W as R9.

"ultra wide margin" is an euphemism for "the same", haven't you learned this in Post Truth class?

TESKATLIPOKA · Mar 2, 2024

coercitiv said:
"ultra wide margin" is an euphemism for "the same", haven't you learned this in Post Truth class?

Sorry, I didn't take that class, It was already full.

Hitman928 · Mar 2, 2024

H433x0n said:
Why? You can limit power in armoury crate to plot these graphs.

Are you asking me why it's not possible for a chip to score the same at 65 W as it does at ~80 W and then score higher at 80 W at the same time?

SiliconFly · Mar 2, 2024

TESKATLIPOKA said:
You should double-check the facts before you post some nonsense.

Not worse than your nonsense. Check it slightly above that.

Hitman928 · Mar 2, 2024

Here's a visual why I believe the Chinese reviewer numbers aren't accurate. Compare the Chinese power/perf chart to what JT got with PL2 higher than PL1. He doesn't report what the PL2 number is, but he didn't change it so PL2 > PL1. Based upon the earlier Zenbook reviews with the 155H and the numbers shown, I believe PL2 for JT's notebook should be around 35 W. JT does run long runs and takes the average, but PL2 still comes into play for a small part of the run, depending on how the test are repeated. Numbers are visual estimates based upon the chart for the Chinese reviewer.

JT's results are roughly in line with what you would expect when PL2 isn't modified to be equal to PL1 and looks agreeable with the Chinese reviewer's 155H results (which also agree with previous 155H results from other reviewers). The Chinese reviewer's 185H results, though, are well above JT's 185H results, even though JT didn't modify PL2 settings. Something that can't possibly be true. Unless you have a reasonable explanation? The only thing I could think of is that there is just a crazy high variability in silicon quality on Intel 4 leading to a massive difference in chip performance between samples. However, let's look at the Chinese reviewer's own numbers:

His own numbers disagree with each other. The above is not possible, unless again, you can come up with a logical explanation?

Lastly, and this is a bit more rough but is additional data that should be good enough for a sanity check, if you take Notebookcheck's review of the model with the 185H, you can get a rough idea of what it should perform like at restricted power levels. I say rough idea because, again, they keep PL2 at stock and there is a large gap between PL1 and PL2 for their model. However, they also provide test data which shows how much performance it loses in CB15 over time as PL2 runs out and becomes a smaller part of each run. If you take the ratio of peak to sustained performance and apply it to their CB23 numbers (which are single run tests) at each PL1 setting, you get this:

So both NBC and JT don't show this large increase in performance for the 185H over the 155H which make sense given that they are the same chips outside of some binning and a higher default TDP. The thing that doesn't make sense is for the 185H to have a near generational leap in perf/efficiency just by better binned silicon. So unless we get additional data that supports the Chinese reviewer's numbers, I don't think they are accurate and I am much more inclined to believe the results that JT and NBC showed in their reviews which has the 185H roughly equal to the 155H at the same power.

tl;dr

JT review shows 185H roughly equal to 155H in performance/efficiency
NBC review shows 185H roughly equal to 155H in performance/efficiency
Chinese reviewer shows 185H significantly higher in performance and efficiency but their own numbers contradict themselves when comparing their charted perf/power to their measured power numbers.
Every data point seems to be in agreement with each other except the Chinese reviewer's 185H results, so I do not believe they can be trusted.

Edit: fixed some wrong data in the initial charts.

TESKATLIPOKA · Mar 2, 2024

SiliconFly said:
Not worse than your nonsense. Check it slightly above that.

What do you mean? What should I look for? Be more specific.

Weren't you talking about this?

Because you were replying to @adroc_thurston and he was replying "at 90W" to your comment about "wide margin" based on this graph from @H433x0n .

Hitman928 · Mar 2, 2024

Further evidence from JT that 185H ~ 155H. He recently reviewed another Zenbook 14 with the 155H. So same laptop model, one with the 155H and one with the 185H.

So JT tested the same model with both the 155H and 185H and found that the 185H had a very modest increase in single core performance and exactly the same multi-core performance at the same TDP.

H433x0n · Mar 2, 2024

Hitman928 said:
Thanks for find that. In their table, they say they are setting PL1 = PL2 for their power testing, but their power consumption graph with their score for repeated tests doesn't seem to agree:

This is the 185h over multiple runs:

View attachment 94690

Corresponding power consumption:

View attachment 94691

So with a base power of 70W and boost power of 95 W for roughly 30% of each run (looking at the later runs), the 185h scores ~18350.

Then, in the table and on their performance over power chart, they show the 185h scoring the same amount with PL1=PL2=65W.

View attachment 94692

There's no possible way this is accurate unless the PL2 isn't being enforced correctly and it is still boosting as before.

Edit: Same applies for the 8945hs in the video.

I’m pretty sure that graph you’re referencing is the default behavior when PL1/PL2 is set to auto. I don’t see any accompanying data that says it’s what they’re using for their 65W results - that’d be completely nonsensical.

Does anybody here understand mandarin and can translate?

Hitman928 · Mar 2, 2024

H433x0n said:
I’m pretty sure that graph you’re referencing is the default behavior when PL1/PL2 is set to auto. I don’t see any accompanying data that says it’s what they’re using for their 65W results - that’d be completely nonsensical.

Does anybody here understand mandarin and can translate?

I posted the table from the review where they say PL1=PL2 with scores at 65W and 45W and the scores from that table match up perfectly with the scores in their plot. If the scores are for default PL2 behavior, then the table is inaccurate and the plot over power is worthless because we have no idea how much power is actually being used at each step.

Edit: Looking at his r15 numbers, they are very much inflated as well, especially the 45W one. I am very confident at this point that the 185H perf/w numbers in his review are worthless.

H433x0n · Mar 2, 2024

Hitman928 said:
I posted the table from the review where they say PL1=PL2 with scores at 65W and 45W and the scores from that table match up perfectly with the scores in their plot. If the scores are for default PL2 behavior, then the table is inaccurate and the plot over power is worthless because we have no idea how much power is actually being used at each step.

Edit: Looking at his r15 numbers, they are very much inflated as well, especially the 45W one. I am very confident at this point that the 185H perf/w numbers in his review are worthless.

So the official Canon for Anandtech is that the reviewer either faked the results or was too incompetent to set power limits in Armoury crate?

Assuming that those 2 charts are correlated, they're not even that far off. That graph you posted appears to be the default PL1/PL2 which is 70W/95W. When running default power limits it got a score of ~18300 on the 15th pass (when running 70W for majority of the run and thoroughly heatsoaked). In his other chart, he claims that when a hard 65W cap is set, that also achieves a score of ~18300. I don't see that as being implausible considering the diminishing returns at that part of the power curve.

DavidC1 · Mar 2, 2024

Hulk said:
Wow, there are some serious quality differences in Intel 4 silicon. 155H and 185H are the same dies, only differing in binning and the v/f curves are obviously quite different as the 185H is significantly more efficient. Assuming this data is legit.

If it's due to a quality difference it is very possible there are big differences even between different 155H or 185H devices. It wouldn't be far fetched considering Intel 4 is a quick and dirty thing to bring Intel 3 online.

In that case(ESPECIALLY) don't be surprised if same is repeated with 20A.

Abwx · Mar 2, 2024

H433x0n said:
In his other chart, he claims that when a hard 65W cap is set, that also achieves a score of ~18300. I don't see that as being implausible considering the diminishing returns at that part of the power curve.

It score 19000 pts at 90W, that s far fetched to assume that it can score 18300 pts at 65W, and that s assuming that the CPU is actually at 90W because the laptop use 155W in CB R13 and end at 145W, this amount to a 140 - 150W delta when accounting its 5W iddle power, that s more reminescent of 100-110W CPU power, close to the 115W official max limit.

Also according to Computerbase it s not possible for the sample they got to precisely set up the power limit in Windows since the chip still had a boost behaviour of its own.

https://www.notebookcheck.net/Honor-MagicBook-Pro-16-2024-A-multimedia-laptop-with-a-165-Hz-IPS-panel-and-RTX-4060.808536.0.html

DavidC1 · Mar 2, 2024

Geddagod said:
uh, he was talking about conroe vs the first iphone cpu

Late reply but Conroe seemed impressive only against the faltering AMD of that time.

iPhone's release and the subsequent breathtaking advancement was a harbinger for what was to come. Otellini was hailed for Conroe/Merom and then in just a few years were scrambling to get into mobile. And scrambling is an understatement. In light of ARM's advancements Intel's "stellar" products don't seem so.

Remember when Silvermont came out and was kicking ass of AMD's Jaguar and in two months later A6 came out and made Silvermont look like a has-been? That was their most efficient CPU and the platform itself delivered great battery life improvement that hasn't been seen since Pentium M in 2003 - 10 years prior. From Intel's standpoint, a computing device that could deliver 8 hour battery life with just 20WHr battery was remarkable.

Both Nehalem's HT and Sandy Bridge's uop caches are going to go. HT due to many cores, security issues, and difficulty of validation increasing risk, and uop caches being there pretty much as a "free" high pipeline, high clocked CPU.

The Cortex X3 of last year already beats Golden Cove/Zen 4 cores. And that's ARM, not Apple nor Nuvia.

Hulk · Mar 2, 2024

DavidC1 said:
Late reply but Conroe seemed impressive only against the faltering AMD of that time.

AMD was doing pretty well from a technological point-of-view up until Conroe, which of course is why Intel made a 180 from Netburst to Core. Conroe was monumental, nearly doubling the IPC of the last iteration of the P4. In fact, it put them so far ahead of AMD that they coasted on it for years and that is part of the reason they are struggling today. Don't underestimate Conroe because it didn't have any competition. It didn't have any competition because it was that good.

Hitman928 · Mar 2, 2024

H433x0n said:
So the official Canon for Anandtech is that the reviewer either faked the results or was too incompetent to set power limits in Armoury crate?

Assuming that those 2 charts are correlated, they're not even that far off. That graph you posted appears to be the default PL1/PL2 which is 70W/95W. When running default power limits it got a score of ~18300 on the 15th pass (when running 70W for majority of the run and thoroughly heatsoaked). In his other chart, he claims that when a hard 65W cap is set, that also achieves a score of ~18300. I don't see that as being implausible considering the diminishing returns at that part of the power curve.

Using 3rd party tools on a laptop for power/frequency/voltage manipulation is never a guarantee. Just because it works on desktop or on a different laptop, doesn't mean it will work correctly on every model, I thought this was common knowledge here? As far as I know, he doesn't even show how he was setting PL1 and PL2 numbers. So yes, most likely the setting of PL2=PL1 was not being enforced correctly and the reviewer didn't do a sanity check to make sure it was working as presented. The numbers bear this out as I've shown.

I'm not sure how you can see the chart I posted and think it looks reasonable. If the chip was heat soaked at 70/95 W, how did it continue to increase in performance up to 95/95 W? It is literally impossible to get the results they did and for the chart to be with PL2=PL1. I will gladly eat crow on this if you can show me any other example of a processor that both plateaus in performance with increasing power, and increases in performance with increasing power simultaneously. If you can find that mythical CPU, I will take back my whole argument.

Edit: he does show he is using the Armoury Crate software to control the LEDs on the laptop, so presumably he is using it to set PL1 and PL2 as well, but again, that doesn't mean it's actually working as expected.

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

Senior member

Attachments

Lifer

Senior member

Lifer

Diamond Member

Platinum Member

Senior member

Senior member

Golden Member

Golden Member

Platinum Member

Diamond Member

Platinum Member

Diamond Member

Golden Member

Diamond Member

Platinum Member

Diamond Member

Senior member

Diamond Member

Senior member

Senior member

Lifer

Senior member

Diamond Member

Diamond Member