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.)

 

[Goop_reformed]

Lol so MTL excels in light work load. Now I've heard it all. Embarrassing if true.

 

[mikk]

Seems like on the CPU side MTL is underwhelming. The GPU is better but also worse than expected if the 3000 points score in timespy is representative. This won't be enough to match Phoenix I think. We should wait for real tests of course, who knows if it's running on newest firmware and driver.

 

[adroc_thurston]

Lol so MTL excels in light work load. Now I've heard it all. Embarrassing if true.

So true, this post was fact checked by real Intel patriots.

 

[trivik12]

Anyway MTL is getting "released" on 12/14. I am sure few OEM like MSI or Acer will have some laptops out and I wont be surprised if some of the big sites/influencers get early dibs next week or at least a reference platform from Intel.

 

[DrMrLordX]

The results shown are significantly lower than the lowest of notebookcheck's results.

Probably better to cross-reference the Meteor Lake results against known Phoenix scores, then, instead of relying on the same leaker to provide useful data about a known product.

We'll be needing new benchmarks.

Why? To make Meteor Lake look better than it really is?

Meteor Lake is expected to jump through all the same hoops as other mobile hardware.

 

[H433x0n]

Tech internet can be unbearable. It may or may not be underwhelming- is it impossible to avoid jumping to conclusions until then?

I can go cherry-pick data right now to make it appear that the 13700H is more efficient than Phoenix by selecting 2 different laptops with varying battery capacity and different power limits and posting benchmark scores with zero context.

Is it too much to ask to wait for a comparison of the same laptop (e.g. XPS 13) equipped with RPL & MTL to collect data with external equipment to measure power consumption? Or is everybody just to anxious to reinforce their prior bias and root for their team?

 

[jpiniero]

Anyway MTL is getting "released" on 12/14. I am sure few OEM like MSI or Acer will have some laptops out and I wont be surprised if some of the big sites/influencers get early dibs next week or at least a reference platform from Intel.

It sure looks like retailers have product. I think it will be available for sale shortly after the announcement. How much product though...

 

[controlflow]

View attachment 90077
So true, this post was fact checked by real Intel patriots.

Do all these machines have the same battery size?

 

[tamz_msc]

View attachment 90077
So true, this post was fact checked by real Intel patriots.

I think we can safely discount this leak. My TGL-H35 laptop on full charge can deliver 6 hours of video STREAMING from YouTube at 4K 60 FPS, on default performance modes and 50% brightness.

And this does less than that with LOCAL video playback? And look at the difference between standby and actual use.

 

[adroc_thurston]

and 50% brightness

Ughhhh well yeah.
That's the point, they didn't.

 

[tamz_msc]

Ughhhh well yeah.
That's the point, they didn't.

The difference between average total power consumption at 50% brightness and 100% brightness is like 1.5-2 watts. At 50% brightness I played back my 4K HEVC copy of Lawrence of Arabia. The average total power consumption was around 8 watts, with the iGPU consuming 0.5 watts on average, as reported by HWinfo. Do the same at 100% brightness and battery life will drop by 30 minutes at most.

 

[cebri1]

Before freaking out people should stop drawing conclusion from one data point. Example: https://browser.geekbench.com/v6/cpu/compare/3912284?baseline=3884691

 

[mikk]

There is a 6400 vs 7467 memory speed comparison from Lenovo and Core Ultra 100H: https://videocardz.com/newz/lenovo-...gher-gaming-performance-compared-to-6400-mt-s

Attention to the timespy scores, roughly 4000 points which is much higher versus the chinese leak from yesterday (3339 overall/3077 GPU). We don't know the exact SKU and if it's a GPU or overall score though (edit: they say gfx, should be the GPU score then). Or maybe it uses much more power there.

 

[Philste]

Raichu leaked 4000 Timepsy Score with 7500MT memory few weeks ago. Yesterdays Video is fake. 13500H has nearly 40% higher GB5 Singlethread than Ultra 5 125H. Phoenix has not even 5h Battery at Idle. It can't get more fake than this.

 

[Hitman928]

Tech internet can be unbearable. It may or may not be underwhelming- is it impossible to avoid jumping to conclusions until then?

I can go cherry-pick data right now to make it appear that the 13700H is more efficient than Phoenix by selecting 2 different laptops with varying battery capacity and different power limits and posting benchmark scores with zero context.

Is it too much to ask to wait for a comparison of the same laptop (e.g. XPS 13) equipped with RPL & MTL to collect data with external equipment to measure power consumption? Or is everybody just to anxious to reinforce their prior bias and root for their team?

​

Before freaking out people should stop drawing conclusion from one data point. Example: https://browser.geekbench.com/v6/cpu/compare/3912284?baseline=3884691

Yes, as multiple people (including myself) have stated many times, we obviously need to wait for the actual release and independent reviewers to put out numbers before we really know what MTL can do. That shouldn't stop us from discussing though when leaks come out. The reason I, and probably others, jump into the discussion more when things like the bilibili leak happen is because (at least in theory) the comparison tests are being done by the same person/group so you assume, or at least hope, that there is some kind of methodology that makes the testing standardized and somewhat fair. The HXL leak is very interesting because they give (supposedly) actual power numbers to go with the score. These are far more interesting than finding random GB numbers to compare where there is pretty much a guarantee of no control between examples and no power numbers or other info to go along with the scores. I actually already pointed out a lot of weirdness with the bilibili leak and said it seems hard to truly trust the scores given that weirdness, but if it's true, here's my take.

For instance, there are multiple scores from the same RPL-H laptop model in your link with significantly higher scores (e.g. https://browser.geekbench.com/v6/cpu/compare/3884691?baseline=3452340). So which number do we take? The one where RPL-H beats the MTL sample in what you think should be a similar laptop model, or the ones where it is lower? What is the power difference of the MTL CPU to the RPL-H model for these scores? Having a power number to go with HXL's score provides an actual comparison point for discussion. If the leaks are wrong and actual performance comes out much better, I'll say again, it will be very entertaining to watch because the leakers will have to spin it to explain how they weren't actually wrong. Same thing happens when we get leaks that seem at all legit for any product from any of the big 3, I'm just here for the ride, lol.

 

[mikk]

Raichu leaked 4000 Timepsy Score with 7500MT memory few weeks ago. Yesterdays Video is fake. 13500H has nearly 40% higher GB5 Singlethread than Ultra 5 125H. Phoenix has not even 5h Battery at Idle. It can't get more fake than this.

Maybe it is with APO enabled, I wonder. But even then there is a too big gap between 3077 to 4000 points. Apparently it was enabled in earlier MTL drivers. On Arc A380 there is roughly 18% between APO on and off in timespy graphics. I don't think the video from yesterday is fake, there can be various reasons or mistakes for a too low score or battery life on a pre launch user test. Too bad he didn't run Geekbench, we have much more data points for this.

 

[cebri1]

Another round of leaks

https://x.com/9550pro/status/1733486866614329655?s=20

185H at 45W TDP scores 767 points on CPU-Z. That would make it 11% faster than the 155H leaked yesterday, which due to clock difference (185H has a 6% clock advantage) seems unlikely.

edit: for comparison 13700H reaches 733 points and it has a 45W TDP as well. However, PL2 is at 115W. Still unknown for the 185H as far as I'm aware.

CPU-Z Benchmark for Intel Core i7-13700H (1T) - CPU-Z VALIDATOR

Best CPU performance - 64-bit - May 2024

valid.x86.fr

 

[csbin]

Another round of leaks

https://x.com/9550pro/status/1733486866614329655?s=20

185H at 45W scores 767 points on CPU-Z. That would make it 11% faster than the 155H leaked yesterday, which due to clock difference (185H has a 6% clock advantage) seems unlikely.

45W is TDP

 

[cebri1]

45W is TDP

yes edited and added 13700H comparison. Question is to how much does the 185H boost during the benchmark. 13700H can reach 115W.

 

[Hitman928]

Another round of leaks

https://x.com/9550pro/status/1733486866614329655?s=20

185H at 45W TDP scores 767 points on CPU-Z. That would make it 11% faster than the 155H leaked yesterday, which due to clock difference (185H has a 6% clock advantage) seems unlikely.

edit: for comparison 13700H reaches 733 points and it has a 45W TDP as well. However, PL2 is at 115W. Still unknown for the 185H as far as I'm aware.

CPU-Z Benchmark for Intel Core i7-13700H (1T) - CPU-Z VALIDATOR

Best CPU performance - 64-bit - May 2024

valid.x86.fr

I think we can all at least agree that the job they did with the thermal paste is absolutely atrocious.

PL2 won't factor into the single core score at all, not if both are set to 45W PL1. Actual power use is still unknown, obviously, but neither should be using more than 45 for a single core test.

 

[mikk]

45W is TDP

This was known, Ultra 9 is 45W and Ultra 5 is 28W on default, although the real power usage can differ, OEMs have lots of freedom. 1.228V for 5.1 Ghz is high voltage, Intel 4 is clearly on the limit. At this speeds it hardly will be more efficient I have to assume.

 

[Hitman928]

Another round of leaks

https://x.com/9550pro/status/1733486866614329655?s=20

185H at 45W TDP scores 767 points on CPU-Z. That would make it 11% faster than the 155H leaked yesterday, which due to clock difference (185H has a 6% clock advantage) seems unlikely.

edit: for comparison 13700H reaches 733 points and it has a 45W TDP as well. However, PL2 is at 115W. Still unknown for the 185H as far as I'm aware.

CPU-Z Benchmark for Intel Core i7-13700H (1T) - CPU-Z VALIDATOR

Best CPU performance - 64-bit - May 2024

valid.x86.fr

Checked scores, it's hard to find because not many reviewers use cpuz. Guru3d has the 13900k getting an 891 score. So 891/5.8 = 153.6 pts/GHz. The 185H there gets a score of 767 with a max boost of 5.1 GHz which gives 150.4 pts/GHz or roughly equal perf/clock.

Edit: 13700h score would give 146.6 pts/GHz so MTL sits right inbetween the two in perf/clock. I think the 13700h link is an average of scores though so it probably gets dragged down a little by CPUs not fully boosting properly. Either way, these scores should roughly equal perf/clock for MTL with RPL which is not unexpected.

 

[eek2121]

Checked scores, it's hard to find because not many reviewers use cpuz. Guru3d has the 13900k getting an 891 score. So 891/5.8 = 153.6 pts/GHz. The 185H there gets a score of 767 with a max boost of 5.1 GHz which gives 150.4 pts/GHz or roughly equal perf/clock.

Edit: 13700h score would give 146.6 pts/GHz so MTL sits right inbetween the two in perf/clock. I think the 13700h link is an average of scores though so it probably gets dragged down a little by CPUs not fully boosting properly. Either way, these scores should roughly equal perf/clock for MTL with RPL which is not unexpected.

MTL was never going to be faster by any significant stretch.

IPC gains for both core types are low-mid single digit.

MTL should be at least 20% more efficient, that is, the same workload should require 20% less energy than an equivalent Raptor Lake chip.

You also typically want to compare similar tiers. Comparing a 155 to a 7840hs is not a good comparison due to binning.

Any non-process related efficiency gains would add to the 20% number.

Wait for better leaks or official benchmarks.

 

[Hitman928]

MTL was never going to be faster by any significant stretch.

IPC gains for both core types are low-mid single digit.

MTL should be at least 20% more efficient, that is, the same workload should require 20% less energy than an equivalent Raptor Lake chip.

You also typically want to compare similar tiers. Comparing a 155 to a 7840hs is not a good comparison due to binning.

Any non-process related efficiency gains would add to the 20% number.

Wait for better leaks or official benchmarks.

A 7840hs is AMD's closest competitor to the 155h.

 

[H433x0n]

A 7840hs is AMD's closest competitor to the 155h.

No, it’s not. The closest competition is Core 7 165H or 185H.

Core 7 155H - 4.8ghz, 28W
Core 7 165H - 5.0ghz, 28-45W?
Core 9 185H - 5.1ghz, 45W
Ryzen 7840HS - 5.1ghz, 35-54W