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.





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.)

 

[Exist50]

We are talking about desktop not mobile. Exist50 basically claimed Raptor Lake-S is not good enough and they need MTL-S against Zen 4 which I and the reviews disagree with. Once again Intel is not in a hurry to bring MTL-S in 2023 because Zen 5 won't come before 2024. If anything AMD needs a price cut to compete with Intels offerings. There is lots of software and preparation work going on for mobile MTL right now and nothing for MTL-S, it's more than obvious Intel favours mobile MTL. Historically this is not a surprise. 14nm, 10nm....and possibly 20A Intel favours mobile on a new process.

For clarity, that statement of mine was meant to be inclusive of mobile, though they could stand to benefit from the node shrink's efficiency gains on desktop as well, even if they're still generally competitive.

If Intel were to release Meteor Lake today, it probably wouldn't budge the needle much in desktop, and would only maybe make them competitive in mobile vs Phoenix. Imo, they need to get Meteor Lake out of the way and move on to Arrow Lake ASAP. While I tend to think the Zen 5 hype train has gotten a little out of control, Zen 5 vs MTL would surely be a win for AMD, and the later MTL arrives, the later ARL is likely to.

 

[Geddagod]

@mikk Even if limiting to desktop you seem to ignore the existence of later V-Cache variants of Zen 4.

Even then, the 13900k seems to be ~10% faster than zen 4 in gaming. Unless Zen 4 V-cache is changed marginally from Zen 3 V-cache, it shouldn't be unjust to assume another ~15% gain from V-cache. Even if they magically make Zen 4 V-cache have the same thermals as Zen 4 vanilla, maybe that's another like ~5% gain (considering that's how much difference we saw with the 5800x vs 5800x3d in ST benchmarks and not gaming). So ye, I expect Zen 4 V-cache to be ~10% faster than RPL in gaming. Maybe Intel released a 13900ks to bring it to like ~7% slower in gaming on average, while winning in some games where cache isn't nearly as helpful, but either way yes, I expect RPL and Zen 4 V-cache to be competitive.

 

[DrMrLordX]

Interesting. You'd think it would be the other way around, with 20A being preferred for high performance. Volume-wise, this would imply pretty high volume on both.

It's easier to launch mobile parts on early iterations of a node. Think Intel 14nm.

Intel needs Meteor Lake to compete with Zen 4, not Zen 5.

There's still no indicator that Intel intends for Meteor Lake to be competitive anywhere except mobile/laptop. It has to compete against Phoenix Point, specifically.

 

[poke01]

There's still no indicator that Intel intends for Meteor Lake to be competitive anywhere except mobile/laptop. It has to compete against Phoenix Point, specifically.

mobile/laptop is the future and frankly where the money is. I think all future Intel designs will be mobile first or least have more focus on mobile than before.

 

[moinmoin]

mobile/laptop is the future and frankly where the money is. I think all future Intel designs will be mobile first or least have more focus on mobile than before.

Intel had a strong mobile focus before, to the point that e.g. Tiger Lake is mobile only with no desktop or server counterpart. The odd thing is that Alder Lake's hybrid approach should be perfectly suited to mobile, but instead it's being optimized for high frequency and area. Its optimization Raptor Lake boosts efficiency quite some, perfect for mobile usage, but is only available as a top end die on desktop, with the rest of the range being refreshed Alder Lake. A new Intel mobile chip with an actual mobile focus would be nice to see.

 

[DrMrLordX]

A new Intel mobile chip with an actual mobile focus would be nice to see.

That's probably Meteor Lake.

 

[RTX]

Tiger Lake is mobile only with no desktop

That was a fab capacity issue according to their engineers in the 11th gen AMA.

 

[Doug S]

There's no chance that N3 is cheaper than any of Intel's internal nodes.

Based on..?

TSMC has more volume than Intel to amortize their fixed expenses, and the variable costs are highly dependent on labor (advantage TSMC) and yield (Intel hasn't had a new node launch with good yields since 22nm while TSMC has had good starting and great sustaining yields for their last several nodes and N3/N3E is supposed to be even better than N5 and N7 were in initial yield)

Those factors may well be enough to give TSMC the advantage even when their profit margin is taken into account versus Intel's internal cost. Not saying it will just that it might, so your declaration of "no chance" doesn't seem reasonable to me.

 

[jpiniero]

That was a fab capacity issue according to their engineers in the 11th gen AMA.

By fab capacity the guy means yields... although I could see a Sales/Marketing guy thinking they are one and the same.

 

[mikk]

That was a fab capacity issue according to their engineers in the 11th gen AMA.

The same could apply to Intel 20A. Intel decided to give mobile parts priority over desktop for 20A. Luckily Intel isn't bound to some older inhous node for the chiplet generation and there is a nice side effect, they can launch the desktop parts before mobile. Desktop SKUs also don't have such a big mass production to store availability delay, it makes sense. And by using MTL-S (6+8?) for the lower to midrange models they can limit the required TSMC volume.

 

[Geddagod]

The same could apply to Intel 20A. Intel decided to give mobile parts priority over desktop for 20A. Luckily Intel isn't bound to some older inhous node for the chiplet generation and there is a nice side effect, they can launch the desktop parts before mobile. Desktop SKUs also don't have such a big mass production to store availability delay, it makes sense. And by using MTL-S (6+8?) for the lower to midrange models they can limit the required TSMC volume.

If Intel decided to give mobile priority over desktop, I would think they would want to use TSMC 3nm over Intel 20A. Maybe Intel 20A does have better perf characteristics, but at the time of planning, it would have made much more sense to use TSMC than their own internal nodes that have had issues lately with getting delayed, imo.

 

[mikk]

If Intel decided to give mobile priority over desktop, I would think they would want to use TSMC 3nm over Intel 20A. Maybe Intel 20A does have better perf characteristics, but at the time of planning, it would have made much more sense to use TSMC than their own internal nodes that have had issues lately with getting delayed, imo.

Much more sense because? When it comes to timelines Intel 20A makes more sense. They are likely launching MTL-P in Q3/Q4 2023 and roughly 1 year latey they can launch ARL-P on 20A. Also for Intel it makes more sense to go into production with smaller chips on a new node. The highend ARL-S compute tile is rumored to have 8+32 cores. ARL-P on the leaked roadmap roadmap only got 6+8 cores. For Intel it's less risky and easier to go into initial 20A production with 6+8 core tiles rather than 8+32.

 

[IntelUser2000]

Mobile is a much bigger market plus it's not socketed so you can't skip it. Meteorlake mobile if it's H2 2023,* then Arrowlake is H2 2024, period. If Intel does what some suggest and skip Meteorlake, then they have no mobile for nearly 1.5 years.

Since Meteorlake-S is nonexistent on the enthusiast side, you skip that and go straight to Arrowlake for the big impact. You don't want a big gap, so Arrowlake comes very early 2024, so yea you are 4-5 months behind, but you get a next generation core, which is worth it.

*Considering Raptorlake mobile isn't out yet, this date is highly optimistic, unless we're going to see that elusive 5W MTL-M, and rest in Q4 2023. But the weird thing is Lunar Lake is mobile only and it's on TSMC N3? What's up with that?

 

[Exist50]

Mobile is a much bigger market plus it's not socketed so you can't skip it. Meteorlake mobile if it's H2 2023,* then Arrowlake is H2 2024, period. If Intel does what some suggest and skip Meteorlake, then they have no mobile for nearly 1.5 years.

Since Meteorlake-S is nonexistent on the enthusiast side, you skip that and go straight to Arrowlake for the big impact. You don't want a big gap, so Arrowlake comes very early 2024, so yea you are 4-5 months behind, but you get a next generation core, which is worth it.

*Considering Raptorlake mobile isn't out yet, this date is highly optimistic, unless we're going to see that elusive 5W MTL-M, and rest in Q4 2023. But the weird thing is Lunar Lake is mobile only and it's on TSMC N3? What's up with that?

I think you're trying to rationalize a roadmap borne more out of necessity than any logical cadence. Where Intel is today, their best option is to release products as and when they're ready, roadmap cadence be damned.

Lunar Lake will definitely be the most interesting product, however. Very curious to see what kind of battery life numbers they can pull off.

 

[jpiniero]

I think you're trying to rationalize a roadmap borne more out of necessity than any logical cadence. Where Intel is today, their best option is to release products as and when they're ready, roadmap cadence be damned.

They're not going to do that. OEMs like a release cadence.

 

[eek2121]

Even then, the 13900k seems to be ~10% faster than zen 4 in gaming. Unless Zen 4 V-cache is changed marginally from Zen 3 V-cache, it shouldn't be unjust to assume another ~15% gain from V-cache. Even if they magically make Zen 4 V-cache have the same thermals as Zen 4 vanilla, maybe that's another like ~5% gain (considering that's how much difference we saw with the 5800x vs 5800x3d in ST benchmarks and not gaming). So ye, I expect Zen 4 V-cache to be ~10% faster than RPL in gaming. Maybe Intel released a 13900ks to bring it to like ~7% slower in gaming on average, while winning in some games where cache isn't nearly as helpful, but either way yes, I expect RPL and Zen 4 V-cache to be competitive.

The X3D parts will likely have BETTER thermals than the non X3D parts. I imagine AMD will be thinning out that heat spreader to make room for the stacked cache.

Based on..?

TSMC has more volume than Intel to amortize their fixed expenses, and the variable costs are highly dependent on labor (advantage TSMC) and yield (Intel hasn't had a new node launch with good yields since 22nm while TSMC has had good starting and great sustaining yields for their last several nodes and N3/N3E is supposed to be even better than N5 and N7 were in initial yield)

Those factors may well be enough to give TSMC the advantage even when their profit margin is taken into account versus Intel's internal cost. Not saying it will just that it might, so your declaration of "no chance" doesn't seem reasonable to me.

Inserting a middleman into your production process would only be cheaper if said middleman can do it cheap enough that they can carve out healthy margins for themselves and still save you money. TSMC is rumored to be charging almost $20k/wafer for N3. I will be shocked if Intel ports their core design over. They really have no need to unless Intel 4 is delayed.

Intel 4 and TSMC N3 are very close performance wise. (unless Intel waters down the specs to hit deadlines)

They have also indicated capacity won’t be an issue for the EUV machines.

 

[coercitiv]

I imagine AMD will be thinning out that heat spreader to make room for the stacked cache.

Last time they thinned out the CPU die so that the bonded part was mechanically identical.

 

[igor_kavinski]

Question - Intel Q3 Results

https://d1io3yog0oux5.cloudfront.net/_2b116bffa4ca8890f09df4510f3c972f/intel/db/887/8873/earnings_release/Q3+22_EarningsRelease.pdf They actually technically made a profit (albeit an small operating loss) but CCG revenue was down 17% and Datecenter was down 27%. Says $3B in "Cost Reductions"...

forums.anandtech.com

Assuming the final step is done on 31st December, can we hope to see 14h gen production silicon in stores by June-2023?

 

[trivik12]

Production stepping is out, then its all about ramp up. Question is if its just MTL-M or if it will have MTL-U or P.

 

[igor_kavinski]

I'm thinking it's mobile because they really need something potent to fight Dragon Range and Phoenix.

 

[Exist50]

They're not going to do that. OEMs like a release cadence.

They like a release cadence, but they also like competitive products. Besides, rumors are that they may partially interleave Meteor Lake and Arrow Lake. I definitely don't see them holding off on MTL just because RPL mobile is late, however.

Intel 4 and TSMC N3 are very close performance wise. (unless Intel waters down the specs to hit deadlines)

Intel 4 and N3 are basically a full node apart. There's a reason Intel appears to be using N3 over even Intel 3.

 

[IntelUser2000]

I definitely don't see them holding off on MTL just because RPL mobile is late, however.

You know they have to right? Cause Intel might want it one way, but what their partners want is another.

You cannot introduce a new laptop gen 6 months after the previous gen. Either one will be skipped or we'll have overlaps for two generations(first with Meteor/Raptor and second with Arrow/Meteor).

Aside from density, since Intel 7 is pretty close to TSMC N7, why would N3 be that much ahead? Performance and performance watt it should be much closer, if not parity.

Assuming the final step is done on 31st December, can we hope to see 14h gen production silicon in stores by June-2023?

Again, since Raptorlake mobile is not out yet, whatever is going to be introduced in January(?) of next year, we're not going to see a Meteorlake version in June of that same year.

 

[jpiniero]

They like a release cadence, but they also like competitive products. Besides, rumors are that they may partially interleave Meteor Lake and Arrow Lake. I definitely don't see them holding off on MTL just because RPL mobile is late, however.

There's no indication that Raptor Lake mobile is going to be late. If you're saying "Well, it's just Alder Lake rebranded"... well it is what it is.

And even Alder Lake rebranded would be OK. AMD is mostly going to be mostly Barcelo anyway because of DDR5. If you think about it, if Meteor/Arrow only support DDR5 holding back those isn't going to be a problem.

 

[Exist50]

You cannot introduce a new laptop gen 6 months after the previous gen. Either one will be skipped or we'll have overlaps for two generations(first with Meteor/Raptor and second with Arrow/Meteor).

The OEMs can simply split their product lines between the two. They don't upgrade everything at once, nor do they do so immediately at launch. So something like mainstream business laptops getting Raptor Lake updates in the first half of the year, and more consumer-oriented devices getting Meteor Lake in the latter half. And on desktop, Rocket Lake -> Alder Lake showed they're willing to refresh in <1 year.

Aside from density, since Intel 7 is pretty close to TSMC N7, why would N3 be that much ahead? Performance and performance watt it should be much closer, if not parity.

N3 is two full node jumps from N5, while Intel 3 is only one full node jump from Intel 7. That's where the divergence comes from. In reality, I expect Intel 4/3 will be competing with N5/N4, and 20A/18A with N3E/N3P/etc. That's not to say that they'll be 100% equivalent, but it's probably the closest comparison.

And again, clearly Intel feels similarly, with Arrow Lake being N3 and 20A. We'll be able to get a really good comparison out of that.

 

[Exist50]

There's no indication that Raptor Lake mobile is going to be late.

I'm not sure if it's late compared to their roadmap (not that they've published one in years...), but typically companies want to launch new consumer products in time for holiday shopping. Unless Intel's been shipping them for a while, that's not gonna happen.