Monday, April 21st 2025
Intel "Nova Lake-S" CPUs to use LGA 1954 Socket, 24 x 25 mm Chipset
Based on recent shipping manifests that point to custom test hardware for its next "Nova Lake-S" processors, Intel seems to be gearing up for a big desktop platform update. According to documents obtained by NBD.ltd, Intel isn't moving full motherboards yet. Instead, it's sending out mechanical interposers and re-balling jigs for an 888-ball BGA chip. Those tools, likely meant for the upcoming 900-series PCH, aren't finished products but specialized fixtures used to check voltage regulation during assembly. Right now, desktop builders are tied to the LGA-1700 socket, but Nova Lake-S is expected to adopt LGA 1954, which offers 1,954 active lands (and possibly more than 2,000 total pads when you count debug pins). That boost in pin count could allow Intel to expand power delivery and I/O without compromise. Still, anyone holding out for Nova Lake-S will have to wait. Intel plans to roll out "Arrow Lake Refresh" later this year, with Nova Lake-S not arriving until 2026.
Under the "NVL-S" codename, Nova Lake is said to combine two clusters of eight high-performance "Coyote Cove" P-cores with 16 "Arctic Wolf" E-cores, plus four ultra-low-power LPE cores in a separate SoC tile. That layout would create a 52-core hybrid chip, though Intel engineers are reportedly still fine-tuning the exact mix before tape-out. The new south-bridge package measures about 25 × 24 mm (around 600 mm²), slightly smaller than the 650 mm² footprint used by today's 800-series chipsets. Shipping lists reveal jig sizes from roughly 38 × 28 × 6.97 mm up to 50 × 50 × 6.32 mm, which tells us multiple fixture types are being used in test labs around the world. Additionally, moving from LGA-1851 to LGA-1954 could hint at a longer-lived socket, but Intel's history suggests most desktop platforms span just two generations. Whether Nova Lake-S or any future "Razer Lake" follow-up will fully exploit this new interface remains to be seen. We are watching closely as more concrete specs emerge ahead of that 2026 launch.
Sources:
via VideoCardz, Tom's Hardware
Under the "NVL-S" codename, Nova Lake is said to combine two clusters of eight high-performance "Coyote Cove" P-cores with 16 "Arctic Wolf" E-cores, plus four ultra-low-power LPE cores in a separate SoC tile. That layout would create a 52-core hybrid chip, though Intel engineers are reportedly still fine-tuning the exact mix before tape-out. The new south-bridge package measures about 25 × 24 mm (around 600 mm²), slightly smaller than the 650 mm² footprint used by today's 800-series chipsets. Shipping lists reveal jig sizes from roughly 38 × 28 × 6.97 mm up to 50 × 50 × 6.32 mm, which tells us multiple fixture types are being used in test labs around the world. Additionally, moving from LGA-1851 to LGA-1954 could hint at a longer-lived socket, but Intel's history suggests most desktop platforms span just two generations. Whether Nova Lake-S or any future "Razer Lake" follow-up will fully exploit this new interface remains to be seen. We are watching closely as more concrete specs emerge ahead of that 2026 launch.

51 Comments on Intel "Nova Lake-S" CPUs to use LGA 1954 Socket, 24 x 25 mm Chipset
next intel desktop cpu is on a new socket?
but those on LGA1700 was probably expecting an upgrade on existing socket...
Apart from Arrow Lake the socket is utilized for Meteor Lake-PS which is an embedded/IoT variant.
is it possible to move tiles around and retain the same socket?
Pick one:
1 - A socket for Arrow and Nova.
2 - A socket for Nova and what comes after it (with the previous being for
Meteor andArrow).The former is a pretty weak option IMO, the latter gets the benefit of a doubt for obvious reasons. Too bad for actual Arrow owners tho..
Doesn't matter anyway, that's not the reason why Intel is doing it. The reason is Meteor, as someone pointed out.
Smaller pins will be harder to repair, but a larger socket makes it harder to layout the Motherboard.
But yeah, anyone who does do a bit of a dive should quickly conclude that atm....there is no reason to go for Intel.
But who knows what the future holds.
Either way, with only 104 more pins it's hardly a big change for repairs or layouts.
LGA 1851 is kinda irrelevant to someone who owns a Raptor Lake, if it was an earlier platform (Z590/490) it will be a huge jump, but for 12th, 13th and 14th gen guys, nahhh uhhh!!
the uplift will not be so dramatic that you will want to upgrade, current cores are under utilized as is and a single generation upgrade hasn't made sense in years.
AM4 was an anomaly because Zen 1 and 2 sucked
Please tell me we aren't headed for anutha ++++++++++++++++++++++++++++++++++++++++++++++ conglomerated clusterflook fiasco......:roll:
It will probably still be quite a long time before we'll see mass DDR6 memory adoption on desktop systems though, so 256-bit DDR5 could be a more cost-effective alternative. There are already "AI-oriented" systems that come with 256-bit bus width anyway, e.g. those equipped with the latest AMD Ryzen Al Max+ CPUs: www.amd.com/en/products/processors/laptop/ryzen/ai-300-series/amd-ryzen-ai-max-plus-395.html
LPDDR memory is most often soldered though, so for high(er) bandwidth while offering flexibility at the same time such systems should probably use LPCAMM modules, with which it's easier to achieve higher speeds than DIMM modules (current standard on desktop systems).
Reviews can only do so much, same with hobbyists like us. Word of mouth is what reaches people before all of those things.. and NVIDIA weaponized that fact. Besides the fact NVIDIA has more often than not arguably had better products GPU wise, this is the main thing that keeps them ahead IMO, besides RT.. CUDA.. yadada. Point is, word of mouth is good. AMD should try to get some.That's debatable, I think they still have a edge in workstation / certain software scenarios but it isn't nearly as big of a edge (if at all) compared to what NVIDIA has over AMD in gpu's. AMD cpus are still mighty good in workstation tasks, and at least for Zen 4 and 5, were much more efficient than their current gen Intel competitors. Ran with less wattage too for similar performance, which is the main thing that stopped me from buying Intel anymore. I lasted up until 8th gen, and then 9th gen came out. I saw the 9900K and saw the writing on the wall and bailed ASAP.
15th gen, Ultra 200, whatever you wanna call them, was a step in the right direction but I don't think they're gonna try and beat AMD at the efficiency crown. Until the CPU's drop though I wont know for sure.. im not that knowledgeable about CPU's personally.
Forward-compatibility with sockets are much harder to achieve than backwards compatibility, at the very least without major compromises. Usually you'll get faster memory speeds, PCIe lanes and USB, and it's very hard to validate older chipsets and boards for future standards, which means at the very least a CPU upgrade will be very limited in usefulness. So what is the common scenario where a CPU upgrade makes sense? So do people buy an overpriced motherboard and a cheap CPU expecting to slot in a better one in two years? Why would you buy a expensive CPU when you don't get to use the new IO features of it on an old motherboard?
As we've seen with AM4, the quality of compatibility can be very mixed.
Backwards compatibility is much more useful; if your board goes bad and the CPU is still good enough, you have a larger chance of finding a good replacement.
Longevity? Nah, not an issue, the hardware lasts for good. Were it not for Microsoft bullshitting its way with Windows 11, there'd be people who wouldn't even look at replacing their old hardware. Heck, even today a Sandy-to-Kaby i7 suffices for many in their everyday computing. Whoever got them at launch and is still using them got the deal of a lifetime.
When talking about longevity, I am not talking about how long a platform will be usable, but for how long it will be supported. Now, 10 years ago when people had to chose Bulldozer or Core, paying Intel every 2 years for a new motherboard, was something that some people couldn't avoid. But today it looks like a bad idea, especially considering that the Intel platform is not cheaper and the latest Intel CPUs are not exactly desirable.
In any case it makes someone wonder if Intel is changing sockets on purpose, to sell chipsets like in the past, or if Intel engineers keep changing their minds about how the next CPU should be and they have to keep changing sockets to accommodate those changes.
The Z890 mATX or Z890M aren't that much more premium compared to high-end AM5 mATX boards. Entry-level CPUs are also quite similar, so I was considering this option.
I'm currently using an 8-core Ryzen 7000 series (AM5), and I doubt it will be the last CPU in this board. I'm hoping to upgrade to a 2-CCD or Zen/ZenC hybrid CPU in the future.
Sandy Bridge has been too slow for heavy web browsing etc. for 1-2 years now, Haswell is probably getting there soon, but anything from the Skylake family will remain useful for years to come. Keep in mind that the comparable Zen 3 is still selling like hot cakes, and Zen 3 will become obsolete before Skylake, as Skylake is more responsive. (While Zen 3 is better for batch workloads, more cores does not age better than faster cores for user-interactive applications.)
Raptor Lake has very inconsistent performance though, due to too aggressive boosting and throttling, and it is very noticeable and annoying. So despite all the misinformation regarding Arrow Lake, it will probably have much more longevity in terms of how long buyers keep their system before "needing" an upgrade.Why are you arguing against things I never said? :rolleyes:
Try reading my post again…