Intel leaker reveals Moon Lake chip as a likely replacement for N-series SoCs — four lower-power-efficiency cores, no performance cores, and a 6W TDP

Moon lake
(Image credit: Getty Images)
12

After Intel let loose its Panther Lake chips earlier in this year, followed by Wildcat Lake in April, the firm has been pretty quiet about upcoming consumer CPUs. Known leaker Jaykihn just let a tiny cat out of the bag, though: the base specs for Moon Lake. The diminutive chip is expected to have four low-power-efficiency (LP-E) cores, and zero performance cores. These are rumors, so have your salt shaker at hand.

Given the all-efficiency layout, it's a fair guess that the cores will be Darkmont LP-E silicon. The leaker additionally notes that the chip layout is likely to be the same as Wildcat Lake, with the efficiency cores in their own island. According to Jaykihn's table, the expected TDPs range from 6 W to 15 W, within the ballpark of phone and tablet SoCs on the low side.

Latest Videos FromTom's Hardware
TOPICS
Bruno Ferreira
Contributor

Bruno Ferreira is a contributing writer for Tom's Hardware. He has decades of experience with PC hardware and assorted sundries, alongside a career as a developer. He's obsessed with detail and has a tendency to ramble on the topics he loves. When not doing that, he's usually playing games, or at live music shows and festivals.

  • usertests
    If it goes down to $100-150 like the ol' N100 mini PCs, I'm in.
    Reply
  • bit_user
    My N97 is good enough for my present needs, but the improvements between Gracemont and Darkmont have been so massive that it'd be hard to resist upgrading!

    My N97 industrial board cost about $220, in early 2024 (IIRC).
    Reply
  • usertests
    bit_user said:
    My N97 is good enough for my present needs, but the improvements between Gracemont and Darkmont have been so massive that it'd be hard to resist upgrading!

    My N97 industrial board cost about $220, in early 2024 (IIRC).
    Moon Lake appears to be positioned as the legitimate successor to Alder Lake-N, targeting a 6W TDP, or higher if allowed.

    But I'm still concerned that forgoing the P-cores is a mistake for many people, because it's a lot of single-threaded performance to lose.

    There's also the matter of the Xe3 cores, but we don't know if Moon Lake will have 1 or 2 maximum yet.

    Minisforum had a barebones Core 3 304 mini PC for $263 early bird pricing, $329 regular. I would certainly be comparing Moon Lake offerings to that.
    Reply
  • bit_user
    usertests said:
    But I'm still concerned that forgoing the P-cores is a mistake for many people, because it's a lot of single-threaded performance to lose.
    I'm not. If you look at an earlier CPU like Arrow Lake (via the 270K+), Skymont is about 77% as fast as Lion Cove on integer performance and 73% on floating point. That's single-thread performance, not IPC.

    So, Intel basically halved the difference between their P-cores and E-cores since Alder Lake. Under the more stringent frequency restrictions of a low-power laptop SoC, I think the difference would be even smaller. That said, I'm not aware of the same test being done on Panther Lake or Wildcat Lake.

    I guess what I'd be more concerned about is that these are LPE cores, rather than regular E-cores.

    usertests said:
    Minisforum had a barebones Core 3 304 mini PC for $263 early bird pricing, $329 regular. I would certainly be comparing Moon Lake offerings to that.
    One thing I need to see is the option to enable in-band ECC support, like Alder Lake-N had. Very few systems exposed the option, but enough eventually did that you could find one which will probably met your needs.
    Reply
  • thestryker
    bit_user said:
    I guess what I'd be more concerned about is that these are LPE cores, rather than regular E-cores.
    The only loss here over the predecessors is the 6MB L3, but it does have a more performant L2 with twice the capacity at 4MB. It'll be interesting to see if that's enough for keeping these cores fed.

    It does sound like these are going to be their own die so Intel must be trying to maximize die per wafer. I would think that the regular Darkmont design would be rather scaleable. Of course who's to say that the non-client market won't be served by something based on the CWF Compute Tile.
    bit_user said:
    the improvements between Gracemont and Darkmont have been so massive that it'd be hard to resist upgrading!
    This thinking is exactly why I replaced my N5105 with a N305: Tremont to Gracemont was pretty big!
    Reply
  • usertests
    thestryker said:
    It does sound like these are going to be their own die so Intel must be trying to maximize die per wafer.
    I wouldn't make that assumption just yet, since this could clearly be made by disabling Wildcat Lake. But it's pretty interesting if it is.
    Reply
  • thestryker
    usertests said:
    I wouldn't make that assumption just yet, since this could clearly be made by disabling Wildcat Lake. But it's pretty interesting if it is.
    Jaykihn said WCL-R is on 18A-P and MNL is 18A.

    They could of course be recycling WCL, but I'm not totally sure why he'd make that distinction otherwise.
    Reply
  • usertests
    bit_user said:
    I'm not. If you look at an earlier CPU like Arrow Lake (via the 270K+), Skymont is about 77% as fast as Lion Cove on integer performance and 73% on floating point. That's single-thread performance, not IPC.

    ...

    I guess what I'd be more concerned about is that these are LPE cores, rather than regular E-cores.
    Cache changes could make a difference, so it could still be in the 20-25% ballpark.

    That's a big chunk of ST performance to lose forever. ST performance is omnipresent, while MT differences are going to be less felt since the CPU is not operating at 100% all the time.

    Memory/storage is disproportionately impacting low-end pricing right now. It comes down to pricing if you don't care about running at 6W, and that could be a tough call. If the difference is $200 Moon Lake vs. $350 Wildcat 1P + 4LPE, both barebones, that gap will shrink after you add the memory/storage.

    Graphics is another consideration, but we don't know about it yet. You also have to step up to a more expensive WCL SKU if you want 2 Xe3 cores.

    I'm just going to have some fun at the end here. If this is a different tile/die, there's nothing stopping them from deleting the NPU and adding 4 Xe3 cores. Suddenly, Moon Lake is better at gaming than Wildcat Lake and the conversation would change. But I think people are expecting this will have 1 Xe3 core. Point is, I'm willing to give Moon Lake a chance if there's an unexpected boon, or the price is right.
    Reply
  • bit_user
    usertests said:
    That's a big chunk of ST performance to lose forever. ST performance is omnipresent, while MT differences are going to be less felt since the CPU is not operating at 100% all the time.
    What's this "forever" business? We're talking about CPUs, so of course you're buying a fixed incarnation! New firmware updates will bring vulnerability mitigations that further reduce performance, but nobody here is under the impression any core is going to get faster, over time!

    And yes, when you go with low-cost and low-power, we're talking about reduced performance. If it's a P-core, the clock speeds are going to be heavily trimmed back. If it's an E-core, clock speeds take a smaller haircut. Either way, it's a compromised implementation. The goal is simply to find the best balance between cost, power, and performance.

    The Intel Core 3 304 has 1x P + 4x LPE and a boost power of 35W. Its P-core has a turbo speed of 4.3 GHz. The E-cores on Panther Lake can boost as high as 3.7 GHz. With a 6W PL1 , Moon Lake will surely have a PL2 of more like 15W, which would cut down that P-core clock speed advantage even further.

    usertests said:
    It comes down to pricing if you don't care about running at 6W,
    I don't need a 6W limit, or else a N100 would've been a better choice for me than N97. Still, I care a lot about efficiency, in an always-on machine. What I don't like about P-cores is how quickly power utilization climbs when they become active.
    Reply
  • usertests
    bit_user said:
    What's this "forever" business? We're talking about CPUs, so of course you're buying a fixed incarnation!
    My poorly stated point is that by adopting Moon Lake, you are locking yourself into lower single-threaded performance for as long as you use the thing.

    I believe single-threaded performance is of high importance to users of budget devices. It affects every aspect of use. The user is more likely to perceive 20% ST loss than 50% MT loss.

    I believe that hybrid/heterogeneous is the most beneficial path for users of power-constrained devices, because ST will never cease to be relevant. It's why we also see smartphone/tablet SoCs with 1-2 "prime" cores and 6-7 slower cores.

    bit_user said:
    And yes, when you go with low-cost and low-power, we're talking about reduced performance.
    Wildcat Lake was at the bottom of new Intel products, and ideally would have been offered at rock bottom prices. Now there's something beneath the bottom.

    Kind of interesting that Wildcat Lake will be sandwiched between Moon Lake and Wildcat Lake Refresh with an additional 2 P-cores. I don't think the latter is strictly necessary. Just get to 1-2 P-cores and you're good.

    I'll have more to say once we actually have benchmarks and reviews to look at.

    bit_user said:
    Still, I care a lot about efficiency, in an always-on machine. What I don't like about P-cores is how quickly power utilization climbs when they become active.
    Maybe operating systems should be handling it better. Obviously, it depends on what you're using the system for, but usually you could tolerate the system being in a lower power state if you're AFK. Not applicable if you're using it as a server.

    Wildcat Lake officially has what Intel is now calling a "Minimum Assured Power" (formerly cTDP Down) of 10 Watts. This thing can be tweaked for decent efficiency, and juiced for extra performance when desired.
    Reply