Thursday, September 3rd 2026

Leak Confirms Intel's Long-Term Socket Plan with LGA-1954

Intel's "Nova Lake-S" desktop processors are drawing nearer, and today a leaked partner slide from China confirmed many of the plans we anticipated. Starting with the LGA-1954 socket, Intel's testing tools had already verified the pin count. This slide not only confirms the pin count but also introduces a small detail about it. In the bottom left corner, there is a disclaimer stating that Intel plans to keep the LGA-1954 socket for multiple generations. This allows enthusiasts to make a one-time investment and maintain the platform for many years. This is a welcome change, considering Intel has switched between three sockets in the last five years: from LGA-1200 to LGA-1700, and then to LGA-1851.

This long-term socket support suggests that Intel is finally adopting a strategy similar to AMD's, as seen with the introduction of AM4, which served for nearly a decade, and AM5, which is planned for support until 2029. With LGA-1954, Intel aims to achieve the same result, providing enthusiasts and cooler makers with a single socket solution for multiple generations. The slide below is machine-translated from Chinese.
The timeline for the "Nova Lake-S" rollout is also confirmed, with mass production set to begin in Q4 2026, and the first SKUs arriving in Q1 2027. The initial edition of "Nova Lake-S" will feature a 28-core, single-compute die SKU, while the flagship 52-core version remains undefined for now. Intel has only indicated that this powerful SKU will be available on desktops within 2027. Looking beyond 2027, "Razor Lake" and "Hammer Lake" are expected.
Sources: wxnod on X, via HardwareLuxx
Add your own comment

45 Comments on Leak Confirms Intel's Long-Term Socket Plan with LGA-1954

#26
Grimling
Actually this is a 24 core CPU (8 P-Core, 16 E-Core), because the 4 LP cores are not part of the Core-Chiplet, they sit in the SOC part of the CPU, they are very very slow and do not have the same access to memory bandwidth and cache as the other cores have. So i would not count them as actually useful cores.
I would not be surprised at all, if people find out deactivating them in the bios actually improves performance overall.
But we will see if my prediction is actually true in this regard.
Posted on Reply
#27
kazuviking
GrimlingI would not be surprised at all, if people find out deactivating them in the bios actually improves performance overall.
Memory and system latency will skyrocket imo. If windows will have proper scheduling then those 4lpe cores will have the same ipc as current arl e cores.
Posted on Reply
#28
gazwales26
Space Lynx28 cores on a single die, wow

AMD = best I can do is 8

or am I misunderstanding something?
Intel still only have 8 p cores the next gen AMD has 24 p cores intel still loosing
Posted on Reply
#29
enb141
AGlezBReally? Do you mind sharing the exact quote where I said or implied that?
Here you did:
I don't recall any recent AMD CPU architecture suporting more than one RAM gen, which makes sense because it adds complexity which in most cases isn't needed.
AGlezBUnlikely. DDR6 is expected to go full CAMM2 and that'll mean new sockets, new CPUs and zero backwards compatibility.
If DDR6 2 goes CAMM2 then this new socket will die when DDR6 appears. So at the end this socket will be limited to DDR5.
AGlezBThe spec is scheduled for 2028 and as with previous generations you will not see mainstream adoption until 2030 at the very least, possibly later given the expected change in form factor and higher initial pricing.
So is 4 years away, that means this new socket will live for 4 years.
cyberdamnWhy would you out Intel's plans like that? They're a poor indie company.
According to how DDR6 will be, which is using CAMM2 instead of slotted RAM, then this socket won't last "forever", this socket will die with DDR5.
Posted on Reply
#30
Sol_Badguy
GrimlingSo i would not count them as actually useful cores.
I feel that there is a chance they might handle background crap/OS bloat and leave 100% of the compute tile's performance for the apps.
Would be good if so, it would mean that professional review performance can be maintained even on an older OS install that has some "baggage".
gazwales26Intel still only have 8 p cores the next gen AMD has 24 p cores intel still loosing
Intel has 8P for the single tile and 8+8P for the dual tile.
AMD has 12 cores for single CCD and 12+12 cores for the dual CCD.
Apples to apples please.
Posted on Reply
#31
RinSenpai
GrimlingActually this is a 24 core CPU (8 P-Core, 16 E-Core), because the 4 LP cores are not part of the Core-Chiplet, they sit in the SOC part of the CPU, they are very very slow and do not have the same access to memory bandwidth and cache as the other cores have. So i would not count them as actually useful cores.
I would not be surprised at all, if people find out deactivating them in the bios actually improves performance overall.
But we will see if my prediction is actually true in this regard.
We already found out about that years ago. Disabling E-cores decreases performance. That's even the case on Windows 10 despite it not having the scheduler for that. It's even more visible if you run apps in background such as Discord and any web browser. Gaming benchmarks aren't representative of the average user, they're here to state the maximum performance you'll extract out of a piece of hardware in perfectly controlled conditions with no wasted ressources, yet we have some stubborn benchmarkers who insists on benchmarking Cities: Skylines 2 and saying it isn't playable with a 9800X3D+5090, when a 14700K, a 265K or just a 7900X are way above it for obvious reasons.
Posted on Reply
#32
kazuviking
RinSenpaiThat's even the case on Windows 10 despite it not having the scheduler for that.
Microsoft ported the scheduler to win 10 forever ago.
Posted on Reply
#33
Space Lynx
Astronaut
gazwales26Intel still only have 8 p cores the next gen AMD has 24 p cores intel still loosing
ya i mispoke earlier, i will just wait for gaming benchmarks from nova lake to see if its worth the upgrade or not. prob not, but if its like a 20% upgrade in fps i might do it, i think some of my games are still CPU limited like total war games, etc.
Posted on Reply
#34
AGlezB
enb141Here you did:
I don't recall any recent AMD CPU architecture suporting more than one RAM gen, which makes sense because it adds complexity which in most cases isn't needed.
Yet this is your claim:
enb141Because you said that AMD CPU work with either DDR5 or DDR6 not both.
You're lucky this isn't a legal forum or you'd be eaten alive. :laugh:

And you still haven't given any example of my supposed contradiction. Let me make it simple for you: just reply with 2 contradicting quotes from me. No comment is needed from you, just my quotes.
Posted on Reply
#35
enb141
AGlezBSame socket is almost a 100% guarantee the new CPU will work, but the mobo will require a BIOS update.

RAM support is part CPU (memory controller) part mobo (chipset, bios, slots, voltages, etc.) so if the CPU supports more than one RAM generation you might find mobos for each and in some cases combo mobos like ASTock G610M COMBO that support both, although not at the same time. Also, it's more an Intel thing; I don't recall any recent AMD CPU architecture suporting more than one RAM gen, which makes sense because it adds complexity which in most cases isn't needed.
Tek-CheckNo, it cannot happen on this platform. DDR6 is several years away, even on data centre CPUs. Don't expect it on desktop before 2030, especially not in this crazy DRAM market.
Posted on Reply
#36
Let3r
Sol_BadguyIntel has 8P for the single tile and 8+8P for the dual tile.
AMD has 12 cores for single CCD and 12+12 cores for the dual CCD.
Apples to apples please.
While Apple is clowning around with 3 P- cores per die.
Posted on Reply
#37
AGlezB
AGlezBSame socket is almost a 100% guarantee the new CPU will work, but the mobo will require a BIOS update.

RAM support is part CPU (memory controller) part mobo (chipset, bios, slots, voltages, etc.) so if the CPU supports more than one RAM generation you might find mobos for each and in some cases combo mobos like ASTock G610M COMBO that support both, although not at the same time. Also, it's more an Intel thing; I don't recall any recent AMD CPU architecture suporting more than one RAM gen, which makes sense because it adds complexity which in most cases isn't needed.
@enb141
I'm almost afraid to ask because this has gone longer than it should and my previous attempts at finding out what the suppossed contradiction is have obviously failed.
First I need to be sure you know what a contradiction is and then I need to you understand that the only way I have to know what you're refering to is for you to provide two or more separate quotes that contradict each other. Here is an example: [ICODE]a = b[/ICODE] and [ICODE]a ≠ b[/ICODE] are contradictory because they cannot both be true at the same time within the given context.
Please limit the quotes to the contradicting part. If you do it right the contradiction (or lack of it) should be clear to anyone reading without needing an explanation, which is what I expected when I asked before. Proving a contradiction is one of the easiest things to do both in logic and dialectic and something any half-braindead junior high student should be able to do with ease.
Posted on Reply
#38
enb141
AGlezB@enb141
I'm almost afraid to ask because this has gone longer than it should and my previous attempts at finding out what the suppossed contradiction is have obviously failed.
First I need to be sure you know what a contradiction is and then I need to you understand that the only way I have to know what you're refering to is for you to provide two or more separate quotes that contradict each other. Here is an example: [ICODE]a = b[/ICODE] and [ICODE]a ≠ b[/ICODE] are contradictory because they cannot both be true at the same time within the given context.
Please limit the quotes to the contradicting part. If you do it right the contradiction (or lack of it) should be clear to anyone reading without needing an explanation, which is what I expected when I asked before. Proving a contradiction is one of the easiest things to do both in logic and dialectic and something any half-braindead junior high student should be able to do with ease.
At the beginning of this "quotation" you said that some intel supported DDR4 and DD5 which is correct, then you mentioned that AMD CPUs never supported DDR4 and DDR5 (I don't know and I don't care because I don't use AMD), everything seems to be fine, but then this new socket LGA-1954 (which is supposed to be a long term socket) is locked to DDR5 (you mention later on correctly that DDR6 will be CAMM2).

So according to you in this quote, I qualify as a half-braindead junior high student (or less) so I have to be very detailed explaining why in my opinion you contradicted yourself, so here is.

The contradiction is that intel could make CPUs that support more than one DDR, just intel not AMD, since this is an intel CPU, and this supposed to be a long term socket, then this should support DDR5 and DDR6, but then DDR6 will use CAMM2, so DDR5 is too different from DDR5 that will require a new socket.

So according to your words, intel could make a CPU with DDR5 and DDR6, with only a BIOS update for the motherboard or in the rare case, a motherboard with both "slots" for DDR5 and DDR6, but since DDR6 will be so different, this time won't happen.

Hopefully this time other half-braindead junior high students like me will understand this.
Posted on Reply
#39
AGlezB
enb141then you mentioned that AMD CPUs never supported DDR4 and DDR5 (I don't know and I don't care because I don't use AMD),
No. I said "I don't recall any recent AMD CPU architecture suporting more than one RAM gen".
enb141(you mention later on correctly that DDR6 will be CAMM2).
No. I said "DDR6 is expected to go full CAMM2". I was more informal on that one so let me rephrase more formally: "I (and others) expect CAMM2 will be the main form factor for DDR6 on desktop PCs". There is no reason DDR6 has to be limited to CAMM2.
enb141The contradiction is that intel could make CPUs that support more than one DDR, just intel not AMD, since this is an intel CPU, and this supposed to be a long term socket, then this should support DDR5 and DDR6, but then DDR6 will use CAMM2, so DDR5 is too different from DDR5 that will require a new socket.

So according to your words, intel could make a CPU with DDR5 and DDR6, with only a BIOS update for the motherboard or in the rare case, a motherboard with both "slots" for DDR5 and DDR6, but since DDR6 will be so different, this time won't happen.
🤦
I even gave you the link, told you the easiest way to prove it and gave you a clear example. I don't remember the last time someone failed so miserably at making an argument.
Let me try to level the playing field a bit. Here what I actually said:
AGlezBSame socket is almost a 100% guarantee the new CPU will work, but the mobo will require a BIOS update.

RAM support is part CPU (memory controller) part mobo (chipset, bios, slots, voltages, etc.) so if the CPU supports more than one RAM generation you might find mobos for each and in some cases combo mobos like ASTock G610M COMBO that support both, although not at the same time. Also, it's more an Intel thing; I don't recall any recent AMD CPU architecture suporting more than one RAM gen, which makes sense because it adds complexity which in most cases isn't needed.
  1. Same socket is almost a 100% guarantee the new CPU will work, but the mobo will require a BIOS update.
    • "Almost" is because there is a chance, no matter how unlikely, that there will exist a combination of CPU and mobo that aren't compatible because they don't support a common DDR gen or for some other reason. Do I also need to explain the BIOS update part?
  2. RAM support is part CPU (memory controller) part mobo (chipset, bios, slots, voltages, etc.).
    • Is common knowledge that you require matching CPU+mobo+RAM.
  3. if the CPU supports more than one RAM generation you might find mobos for each and in some cases combo mobos.
    • "If" because I don't offer any guarantee about what a future CPU will or will not not support.
    • "Might" because the is no guarantee that (a) they'll exist and (b) you will get your hands on one.
  4. that support both, although not at the same time.
    • You cannot use both gens at the same time. RTFM.
  5. it's more an Intel thing; I don't recall any recent AMD CPU architecture suporting more than one RAM gen.
    • Both are claims about existing architectures. No mention whatsoever about future ones.
  6. which makes sense because it adds complexity which in most cases isn't needed.
    • That is my reasoning about why there isn't "any recent AMD CPU architecture suporting more than one RAM gen" and complexity equals cost so given the propotion of architectures supporting one gen vs. multiple I can confidently say my reasoning is correct.
enb141Hopefully this time other half-braindead junior high students like me will understand this.
Not like you. You placed yourself below the level of a half-braindead junior high student. And I wasn't joking when I said you'd be eaten alive in a legal forum. My face would be on fire if I had facepalmed every time I read "you said" or "you mentioned" followed by a twisted interpretation of what I actually said. Normally I appreciate when someone points to a mistake I made because I gives me a chance to correct and learn from it but this is the most effort I've ever put into trying to get someone to point one out only to be reminded of how an average is calculated.

In any case DM me if you want to give it another try because we've already subjected the rest to enough torture. Just don't be surprised if ignore anything but a sound argument.
Posted on Reply
#40
enb141
AGlezBNo. I said "I don't recall any recent AMD CPU architecture suporting more than one RAM gen".


No. I said "DDR6 is expected to go full CAMM2". I was more informal on that one so let me rephrase more formally: "I (and others) expect CAMM2 will be the main form factor for DDR6 on desktop PCs". There is no reason DDR6 has to be limited to CAMM2.


🤦
I even gave you the link, told you the easiest way to prove it and gave you a clear example. I don't remember the last time someone failed so miserably at making an argument.
Let me try to level the playing field a bit. Here what I actually said:
  1. Same socket is almost a 100% guarantee the new CPU will work, but the mobo will require a BIOS update.
    • "Almost" is because there is a chance, no matter how unlikely, that there will exist a combination of CPU and mobo that aren't compatible because they don't support a common DDR gen or for some other reason. Do I also need to explain the BIOS update part?
  2. RAM support is part CPU (memory controller) part mobo (chipset, bios, slots, voltages, etc.).
    • Is common knowledge that you require matching CPU+mobo+RAM.
  3. if the CPU supports more than one RAM generation you might find mobos for each and in some cases combo mobos.
    • "If" because I don't offer any guarantee about what a future CPU will or will not not support.
    • "Might" because the is no guarantee that (a) they'll exist and (b) you will get your hands on one.
  4. that support both, although not at the same time.
    • You cannot use both gens at the same time. RTFM.
  5. it's more an Intel thing; I don't recall any recent AMD CPU architecture suporting more than one RAM gen.
    • Both are claims about existing architectures. No mention whatsoever about future ones.
  6. which makes sense because it adds complexity which in most cases isn't needed.
    • That is my reasoning about why there isn't "any recent AMD CPU architecture suporting more than one RAM gen" and complexity equals cost so given the propotion of architectures supporting one gen vs. multiple I can confidently say my reasoning is correct.
Not like you. You placed yourself below the level of a half-braindead junior high student. And I wasn't joking when I said you'd be eaten alive in a legal forum. My face would be on fire if I had facepalmed every time I read "you said" or "you mentioned" followed by a twisted interpretation of what I actually said. Normally I appreciate when someone points to a mistake I made because I gives me a chance to correct and learn from it but this is the most effort I've ever put into trying to get someone to point one out only to be reminded of how an average is calculated.

In any case DM me if you want to give it another try because we've already subjected the rest to enough torture. Just don't be surprised if ignore anything but a sound argument.
Since I'm less than a half-braindead junior high student (by far) and I love being facepalmed (a lot):

The real issue here is, as I said, that when the Intel CPUs support DDR6, you will need a new motherboard, but you said and keep claiming that you will only need a BIOS update. Why because you said that the CPU controls what Memory can used (some intel CPUs supported DDR4 and DDR5), but as I keep saying, this socket LGA 1954 will work only for DDR5 CPUs, when DDR6 is being released, I doubt DDR6 CPUs will support both DDR5 and DDR6.

Hopefully this slap in my face goes in my other cheek.
Posted on Reply
#41
AGlezB
enb141Since I'm less than a half-braindead junior high student (by far) and I love being facepalmed (a lot):

The real issue here is, as I said, that when the Intel CPUs support DDR6, you will need a new motherboard, but you said and keep claiming that you will only need a BIOS update. Why because you said that the CPU controls what Memory can used (some intel CPUs supported DDR4 and DDR5), but as I keep saying, this socket LGA 1954 will work only for DDR5 CPUs, when DDR6 is being released, I doubt DDR6 CPUs will support both DDR5 and DDR6.

Hopefully this slap in my face goes in my other cheek.
If you're going to argue at least get the facts right. This is what you said:
enb141Same socket ≠ will work with all CPUs with that socket (correct me if I'm wrong)
To which I responded:
AGlezBSame socket is almost a 100% guarantee the new CPU will work, but the mobo will require a BIOS update.
Then you continued with this:
enb141that socket could change from DDR 5 to DDR 6, so I doubt DDR 6 Processors will work with old motherboards with DDR 5 RAM
And here is what I said:
AGlezBRAM support is part CPU (memory controller) part mobo (chipset, bios, slots, voltages, etc.) so if the CPU supports more than one RAM generation you might find mobos for each and in some cases combo mobos like ASTock G610M COMBO that support both, although not at the same time. Also, it's more an Intel thing; I don't recall any recent AMD CPU architecture suporting more than one RAM gen, which makes sense because it adds complexity which in most cases isn't needed.
As you can see I'm not the one claiming LGA-1954, and I quote, "could change from DDR 5 to DDR6" despite both the leaked presentation and the wiki page saying DDR5 and making no mention of DDR6.
Also, at the time I didn't think necessary to spoonfeed you the response so I just wrote in paragraphs instead of splitting it like I'm doing here.

In any case I'm tired of this so here's a contradiction for (from) you:
enb141that socket could change from DDR 5 to DDR 6
enb141this socket LGA 1954 will work only for DDR5 CPUs
Simple and easy. :laugh:
Posted on Reply
#42
enb141
AGlezBIf you're going to argue at least get the facts right. This is what you said:

To which I responded:

Then you continued with this:

And here is what I said:

As you can see I'm not the one claiming LGA-1954, and I quote, "could change from DDR 5 to DDR6" despite both the leaked presentation and the wiki page saying DDR5 and making no mention of DDR6.
Also, at the time I didn't think necessary to spoonfeed you the response so I just wrote in paragraphs instead of splitting it like I'm doing here.

In any case I'm tired of this so here's a contradiction for (from) you:


Simple and easy. :laugh:
Simpler to any visitor that reads this:

My claim: LGA-1954 will only work on DDR5 CPUs, DDR6 will require a new motherboard.

Your claim: Could be or could not be, could work or could not work.
Posted on Reply
#43
Tyler-98-W68
If you are old enough to remember. Asrock had an AM2 MB that was native DDR1. But had a proprietary slot, and when released the card (that was designed for the proprietary slot had DDR2 support and newer AM2+ cpus. I had a board but never used the upgrade board.
Posted on Reply
#44
enb141
Tyler-98-W68If you are old enough to remember. Asrock had an AM2 MB that was native DDR1. But had a proprietary slot, and when released the card (that was designed for the proprietary slot had DDR2 support and newer AM2+ cpus. I had a board but never used the upgrade board.
Apple did something similar, they had a proprietary RAM slot (overpriced as every RAM / Storage Apple sells of course).

I think DDR6 will be so different that will need a new motherboard, something like RDRAM, currently you can see laptops with CAMM or slotted DDR5, so when DDR6 gets released, the most likely "slot" will be CAMM2, so that will kill older motherboards, forcing users to get brand new ones.
Posted on Reply
#45
konataVT
Space Lynx28 cores on a single die, wow

AMD = best I can do is 8

or am I misunderstanding something?
yeah to my knowledge AMD just cant hang with intel on the monolithic design. as you scale chiplets, you scale the latency and interconnect needed to unify the dies, leading to wrose gaming performance (see EPYC and threadripper gaming) so intel may be in the running for a serious overwhelm in core count on monolithic designs. also, 18A means power is down down and density is way up. keep an eye out...
Posted on Reply
Add your own comment
Nov 12th, 2026 06:32 PST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts