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)	Lunar Lake	Panther Lake
Platform	Mobile H/U Only	Desktop Only	Desktop & Mobile H&HX	Mobile U Only	Mobile H
Process Node	Intel 4	Intel 20A	TSMC N3B	TSMC N3B	Intel 18A
Date	Q4 2023	Q1 2025 ?	Desktop-Q4-2024 H&HX-Q1-2025	Q4 2024	Q1 2026 ?
Full Die	6P + 8P	6P + 8E ?	8P + 16E	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.)

 

[H433x0n]

Hard to say whether it's RPL or RPL-R. Absolutely no context. We shouldn't be using this info until we have better clarity.

Assuming it's 1.05 ST perf over RPL-R I suppose it could be worse. That's the downside of following up a generation that pushed 6ghz clocks on a mature process. Let's hypothesize ARL-S tops out at ~5.0ghz boost clocks, that puts raw IPC at ~20% which is pretty believable.

That means Zen 5 would have to improve ST perf by +20% before it starts to take the lead (RPL-R will be +15% ST perf over Zen 4). It's definitely possible for Zen 5 to have the ST performance leadership this upcoming generation.

 

[SiliconFly]

15% nT perf ISO core count

Can you pls elaborate a bit.

 

[H433x0n]

Can you pls elaborate a bit.

He just means that total MT performance was increased by 15% with the same core counts gen over gen.

 

[SiliconFly]

Assuming it's 1.05 ST perf over RPL-R I suppose it could be worse. That's the downside of following up a generation that pushed 6ghz clocks on a mature process. Let's hypothesize ARL-S tops out at ~5.0ghz boost clocks, that puts raw IPC at ~20% which is pretty believable.

That means Zen 5 would have to improve ST perf by +20% before it starts to take the lead (RPL-R will be +15% ST perf over Zen 4). It's definitely possible for Zen 5 to have the ST performance leadership this upcoming generation.

Based on leaks, there's a distinct possibility that Zen 5 might edge out ARL-S in performance. But why would intel want to sacrifice performance for efficiency in nextgen desktops? It's mind-numbing. Makes no sense.

Also, MLID keeps claiming 40% increase in performance for ARL over previous gen (possibly MT perf). Is he blowing smoke as usual?

 

[H433x0n]

Based on leaks, there's a distinct possibility that Zen 5 might edge out ARL-S in performance. But why would intel want to sacrifice performance for efficiency in nextgen desktops? It's mind-numbing. Makes no sense.

According to the internet, that's winning™. That's what everybody says they wanted - efficiency.

Also, MLID keeps claiming 40% increase in performance for ARL over previous gen (possibly MT perf). Is he blowing smoke as usual?

He's basically just guessing with ARL. His logic (I assume) was that it was 2 node jumps and a brand new core arch. It ended up being incorrect because he over-estimated the effect of the node on performance, if anything the node is probably hurting overall performance and it would've been better off using Intel 3 (it'd be likely to clock better).

 

[Geddagod]

Much like who ate the last piece of pie, what's going on with this?

¯\_(ツ)_/¯ Recent rumors by Xino? IIRC on twitter said PTL as a whole was canned, and we are going to have to wait until Nova Lake for a new CPU arch. Whether that means Nova Lake is getting pushed up or not is doubtful. Doubt Royal Core shows up till 2026/2027.

For the same architecture, it does. For a better architecture, even more.

You can't multiply perf/watt numbers reported by Intel because where they are getting the perf/watt numbers on the curve is not likely to be the same on each node, first of all.
Also not all archs improve perf/watt much (cough SNC cough)

It's a 30-50% wider core than GLC. A 5% improvement is extremely disappointing.

Apple went wider too lol, look where it got them
Before I call it extremely disappointing, I would look at the perf/watt curve as well. Top end SC performance only matters for gamerzzzz. Looking forward to IPC leaks.

Hard to say whether it's RPL or RPL-R. Absolutely no context. We shouldn't be using this info until we have better clarity

The difference is like 5% between RPL and RPL-R lmao. And this is the second source to confirm a 5% improvement. And I'm pretty sure it's RPL-R.

Can you pls elaborate a bit.

8+16 vs 8+16. ARL (at least in 2024) isn't bringing higher core counts.

But why would intel want to sacrifice performance for efficiency in nextgen desktops? It's mind-numbing. Makes no sense.

You say this as if they haven't for literally every generation before this.

Also, MLID keeps claiming 40% increase in performance for ARL over previous gen (possibly MT perf). Is he blowing smoke as usual?

I think so

According to the internet, that's winning™. That's what everybody says they wanted - efficiency.

If this was true, everyone wouldn't have loved Alder Lake / Golden Cove nearly as much as they did.

He's basically just guessing with ARL. His logic (I assume) was that it was 2 node jumps and a brand new core arch.

Ye I think so

It ended up being incorrect because he over-estimated the effect of the node on performance,

For ST perf, all the node does mostly is just allow the cores themselves to get bigger.

if anything the node is probably hurting overall performance and it would've been better off using Intel 3 (it'd be likely to clock better).

Both N3B and Intel 20A versions of ARL look to be bad. ST Clock limitations appear to be an architectural issue, not a node one.

 

[A///]

¯\_(ツ)_/¯ Recent rumors by Xino? IIRC on twitter said PTL as a whole was canned, and we are going to have to wait until Nova Lake for a new CPU arch. Whether that means Nova Lake is getting pushed up or not is doubtful. Doubt Royal Core shows up till 2026/2027.

PTL? Potato Lake?

What you're saying here is Royal Core does exist and wasn't a fabrication of someone's imagination which most have been arguing for because it really hasn't been brought up under good circumstances and only mentioned by questionable youtube personalities who look like human bobbleheads?

 

[H433x0n]

Before I call it extremely disappointing, I would look at the perf/watt curve as well. Top end SC performance only matters for gamerzzzz. Looking forward to IPC leaks.

It matters a lot (at least for me personally). There's still quite a few workloads where I'm waiting on a single thread.

You say this as if they haven't for literally every generation before this.

That's why it's so painful IMO. I would gladly trade ARL-S being 25% less efficient if it meant it gained an additional +15% ST & MT performance increase.

If this was true, everyone wouldn't have loved Alder Lake / Golden Cove nearly as much as they did.

I've got to disagree. The general consensus on the internet is the exact opposite. RPL-R will probably have +15% ST performance advantage over Zen 4 but according to a lot of people on this forum they'd rather have better efficiency in MT workloads and a smaller core.

Both N3B and Intel 20A versions of ARL look to be bad. ST Clock limitations appear to be an architectural issue, not a node one.

Big oof if true.

 

[Geddagod]

PTL? Potato Lake?

What you're saying here is Royal Core does exist and wasn't a fabrication of someone's imagination which most have been arguing for because it really hasn't been brought up under good circumstances and only mentioned by questionable youtube personalities who look like human bobbleheads?

PTL was Panther Lake. It was supposed to come after Arrow Lake.
I mean, everyone ranging from Exist50 to Raichu was talking abt Royal Core, so I do think it exists.

 

[A///]

PTL was Panther Lake. It was supposed to come after Arrow Lake.
I mean, everyone ranging from Exist50 to Raichu was talking abt Royal Core, so I do think it exists.

Who is Raichu again? Sadly or unsadly depending on whom you ask Exist is no longer with us.

Panther Lake yes right you mentioned it got canceled to me a while back. Was that a desktop platform with Lunar Lake being mobile?

 

[dullard]

Who is Raichu again?

A leaker on X (Twitter). He has some form of inside information considering he is already benchmarking Raptor Lake-R. https://browser.geekbench.com/v6/cpu/2605111

How much inside information he has is unknown.

 

[Glo.]

If Intel's hyper-focused on efficiency, then ARL mobile's power-efficiency will be off the charts due to 20A. Will definitely have a huge impact in mobile.

ARL-S (desktop) can also have a huge PPW improvement. But without much performance improvement how that translates into sales is a big if.

Don't hype the product, or yourself for it. It will be what it will be.

I wouldn't go that far with efficiency of ARL, as an architecture.

 

[A///]

A leaker on X (Twitter). He has some form of inside information considering he is already benchmarking Raptor Lake-R. https://browser.geekbench.com/v6/cpu/2605111

How much inside information he has is unknown.

Oh yes him or her, right right. I saw the KF benches last night and was impressed. I want to see real world software performance differences but that will come out sooner or later. This is a go for an october launch, right?

 

[dullard]

Oh yes him or her, right right. I saw the KF benches last night and was impressed. I want to see real world software performance differences but that will come out sooner or later. This is a go for an october launch, right?

Officially Intel has announced Sept 19/20 for "Next Generation Systems and Platforms" which later is explained as "delivering on future processor technology such as Meteor Lake and Intel ® Xeon®" Raptor Lake-R isn't mentioned but everyone assumes it will be soft-launched as well during the event. https://www.intel.com/content/dam/w...s/2023-07/intel-innovation-yourguide-2023.pdf

Rumors are for the actual launch to be Oct 17 with the advertising embargo lifting Oct 16. That would be when we get 3rd party benchmarks with production BIOSes.

 

[A///]

Officially Intel has announced Sept 19/20 for "Next Generation Systems and Platforms" which later is explained as "such as Meteor Lake and Intel ® Xeon®" Raptor Lake-R isn't mentioned but everyone assumes it will be soft-launched as well during the event. https://www.intel.com/content/dam/w...s/2023-07/intel-innovation-yourguide-2023.pdf

Rumors are for the actual launch to be Oct 17 with the advertising embargo lifting Oct 16.

That's about when Intel released Raptor Lake last year. PC Canada has some of the 14th gen refreshes up on display according to hasan's site or videocards. The pricing seems to be 4% or abouts over the 13th gen pricing at launch based on inflation. Much like last year I think these prices will fall considerably by November or December. The 13900K was sold out for a long time or going for much higher prices on some websites but it's a good deal now if you want to stick with intel. The KF bench raichy did was with DDR5 7000 or 7200 RAM which is doable with 13th gen and a decent board, but it has spurred the rumors of an improved IMC for 14th gen. I saw some unconfirmed reports last week from a leaker worse than mlid who claimed low 8000 stable was possible with the i9s. Sounds like hooey but eh who knows.

 

[H433x0n]

I ran GB6 to compare to the leaked numbers by Raichu and this is what I got. Looks like a +4% ST perf uplift. My system is not overclocked and is running DDR5-6400 64Gb. I think my score being higher than the average 13900K could be that I've got dual rank memory and I disable hyperthreading.



Just incase anybody wanted context on the 14900K numbers.

 

[dullard]

I ran GB6 to compare to the leaked numbers by Raichu and this is what I got. Looks like a +4% ST perf uplift. My system is not overclocked and is running DDR5-6400 64Gb. I think my score being higher than the average 13900K could be that I've got dual rank memory and I disable hyperthreading.

Just incase anybody wanted context on the 14900K numbers.

Thanks. Pretty similar to the 13900KF comparison I posted yesterday and basically in line with the 3.4% turbo frequency increase. There may be a percent change here or there if BIOSes are updated between now and the actual launch. With the exception of the i7 getting rumored additional E cores for multithreading performance increases, it is a pretty MEH product line. But, I'll always take a performance increase.

 

[A///]

Rumoured? I thought it was confirmed the i7 is was getting a bump in e cores?

 

[S'renne]

Mobile is where the battle will be the biggest and toughest between AMD and Intel in upcoming years.

Well I thought about this whilst including LNC being used in Lunar Lake that had a development sample leak that ran at 8W, and it's not like AMD hasn't done something similar on their end too which is Ryzen 6000. Having the full Raptor Lake performance(or slightly better in this case) that fits the mobile thermal/power limits would still be really good.

 

[JoeRambo]

Why do you say that? Do you have any recent leaks?
I thought ARL was supposed to be the first itineration of the new core design?

Are leaks really needed? Does anyone here believes that Intel will take Redwood Cove ( in Meteor Lake, pretty much a known perf quantity at this point due to leaks ) and completely turn it around for one generation in Arrow Lake on same socket for massive performance increase. And then throw it away to introduce Lion Cove in Generation 16?
Not really, right?
Intel already has people committing stuff for Lion Cove in perfmon, but somehow this 30% IPC increase magic mushroom architecture is skipped, even if those advances would require ton of PMU updates in 2024 for all these new ports and capabilities.
Yeah right.

I don't have any info, but common sense says -> Arrow Lake is using same Redwood Cove on new process, likely increasing L2 to 3MB and maybe L3 to some more per core. That's it. And clocking better than 5Ghz incarnation in Meteor Lake.

https://twitter.com/x/status/1689646409329569792

That's right, Your 30% IPC core does not even have it's own CPUID branch for PMU "selection". Guess what current Intel's core does not have one as well? Hint, the one that increased L2 to 2MB and has 100% identical PMU.

 

[SiliconFly]

Why do you say that? Do you have any recent leaks?
I thought ARL was supposed to be the first itineration of the new core design? It doesn't make sense that they could not get hyperthreading working (if rumor is correct) if it was the same basic core as ALD/RL.

It's very much possible for ARL to have the 1st iteration of LNC (maybe even without hyper-threading or the rumored rentable cores). There's a small chance it has RWC+ (instead on LNC). But in that case, Intel dropping hyper-threading feels mind-numbing. Doesn't add up.

There were some old rumors that suggested that Intel was actually working on both RWC+ & LNC at the same time for ARL, and later dropped RWC+ in favor of LNC (due to some progress or hitting some milestone or something like that). But these unsubstantiated rumors and maybe totally false as the source doesn't have much credibility.

At this point, not much is known.

 

[ondma]

Are leaks really needed? Does anyone here believes that Intel will take Redwood Cove ( in Meteor Lake, pretty much a known perf quantity at this point due to leaks ) and completely turn it around for one generation in Arrow Lake on same socket for massive performance increase. And then throw it away to introduce Lion Cove in Generation 16?
Not really, right?
Intel already has people committing stuff for Lion Cove in perfmon, but somehow this 30% IPC increase magic mushroom architecture is skipped, even if those advances would require ton of PMU updates in 2024 for all these new ports and capabilities.
Yeah right.

I don't have any info, but common sense says -> Arrow Lake is using same Redwood Cove on new process, likely increasing L2 to 3MB and maybe L3 to some more per core. That's it. And clocking better than 5Ghz incarnation in Meteor Lake.

https://twitter.com/x/status/1689646409329569792

That's right, Your 30% IPC core does not even have it's own CPUID branch for PMU "selection". Guess what current Intel's core does not have one as well? Hint, the one that increased L2 to 2MB and has 100% identical PMU.

Sorry, but I dont understand what you are trying to say. I was making no comment about a "magical 30%" performance increase. All I was saying is that it makes no sense for hyperthreading to be dropped if ARL is simply another iteration of Redwood Cove. And the fact that Lunar Lake will have a new architecture doesnt rule out that ARL will too.

 

[mikk]

ARL-S gets new extensions over MTL and partially ARL mobile, means new architecture. It's not Redwood Cove.

 

[AMDK11]

Moving from 3-way to 4-way x86 decoding represents a 33% increase. Intel had 4-way decoding from Conroe (Core 2 - 2006) to SunnyCove (CypressCove - 2021), and that's about 15 years of development and a very large increase in IPC. Going from 4-way to 6-way x86 decoding in GoldenCove represents a 50% increase. Why can't we assume that a lot of IPC can be achieved with 6-way decoding?

Could GoldenCove be so close to perfectly exploiting the limits of 6-way x86 decoding?Even Zen 3 and Zen 4 with 4-way x86 decoding have largely comparable capabilities to GoldenCove.According to rumors, Zen 5 will also gain x86 6-way decoding.Isn't it possible that RedwoodCove, with the already known improvements, can gain even about 15% higher IPC?

 

[Geddagod]

Moving from 3-way to 4-way x86 decoding represents a 33% increase.Intel had 4-way decoding from Conroe (Core 2 - 2006) to SunnyCove (CypressCove 2021), and that's about 15 years of development and a very large increase in IPC.Going from 4-way to 6-way x86 decoding in GoldenCove represents a 50% increase.Why can't we assume that a lot of IPC can be achieved with 6-way decoding?

Idk what this is in response too, but no one is assuming high IPC can't be achieved with 6 way decoding. Look at how long a 4 way decode was milked. RWC not getting much higher IPC vs GLC isn't an assumption based on both having 6 way decoders, but the lack of many architectural changes down the rest of the pipeline as well, leaks from other leakers, and leaked GB6 scores too.