Thursday, April 9th 2026

Intel "Nova Lake-S" Uses 2L-ILM Socket for Better CPU Cooler Contact

In the latest series of leaks about Intel's upcoming "Nova Lake" chips, we learn that the desktop "Nova Lake-S" designs will feature a new socket mounting and independent loading mechanism (ILM). According to exclusive information from VideoCardz, the NVL-S desktop designs on the LGA-1954 socket will include a new 2L-ILM socket mechanism to improve contact with CPU coolers. Given that Intel's upcoming NVL-S desktop processors will have up to 52 high-frequency, power-hungry cores, a better mounting mechanism is definitely needed. This 2L-ILM is a two-level independent loading mechanism, using one lever on each side of the CPU to secure it into the LGA-1954 socket. Applying pressure on both sides of the socket ensures improved flatness while the locking mechanism does its job.

Intel categorizes its desktop sockets in various forms. One of the most basic is the Default-ILM, along with the RL-ILM. As seen with "Arrow Lake," Intel's motherboards divide the socket locking mechanism into the ILM and RL-ILM designs, applied depending on the sector. For the lower-end sector, the Default-ILM socket is used, while the RL-ILM is applied to higher-end overclocking motherboards, providing better pressure and ensuring a flatter surface contact for a cooler. Noctua and Cooler Master already differentiate these two in their CPU coolers, but the mounting hardware is generally the same for both. The RL-ILM is simply a better design for overclocking due to the improved surface contact it provides.
Some readers may recall that Intel used its 2L-ILM on the older LGA 2011-3 HEDT socket back in the era of X99 Express motherboards, more than a decade ago in 2014. This HEDT platform featured 2011 LGA pins, which is close to the 1954 LGA pins we see with NVL-S. It seems that having close to 2,000 pins in the LGA socket requires two levers due to the pressure needed for the application of the CPU, as the surface area needs a stronger hold to ensure that every CPU contact pin on the package is electrically touching the pins on the socket it is mounted on.
Source: VideoCardz
Add your own comment

37 Comments on Intel "Nova Lake-S" Uses 2L-ILM Socket for Better CPU Cooler Contact

#1
john_
will have up to 52 high-frequency, power-hungry cores
Oh, come on. Mistakes like this shouldn't be on TPU. That 52 should be 16. The rest of the cores are E-cores, or even lower-powered cores.

I hope the mechanism has been properly designed so that if someone doesn't engage both levers simultaneously and correctly, it doesn't create uneven pressure on one side, risking bending the socket pins.
Posted on Reply
#2
AleksandarK
News Editor
john_Oh, come on. Mistakes like this shouldn't be on TPU. That 52 should be 16. The rest of the cores are E-cores, or even lower-powered cores.

I hope the mechanism has been properly designed so that if someone doesn't engage both levers simultaneously and correctly, it doesn't create uneven pressure on one side, risking bending the socket pins.
You tell me that having 52 cores is normal and not power hungry? Even if its e cores, they run at 3+ GHz, base TDP is pushed to 175, imagine what the boost will look like
Posted on Reply
#3
dgianstefani
TPU Proofreader
AleksandarKYou tell me that having 52 cores is normal and not power hungry? Even if its e cores, they run at 3+ GHz, base TDP is pushed to 175, imagine what the boost will look like
E cores on the ~$300 7 tier 270K Plus boost to 4.7 GHz, with a base of 3.2 GHz. These are as performant as Raptor Lake P cores at iso frequency, so I doubt that Nova Lake E cores will have lower clocks or performance. Regardless of core type, having 50+ is going to draw power under full load, I expect 350 W+ if the long duration power limits are removed and it's allowed to boost as much as it does for short duration/low core load. Of course, that will still likely be very efficient, considering the amount of work being done.
Posted on Reply
#4
Flying Fish
john_Oh, come on. Mistakes like this shouldn't be on TPU. That 52 should be 16. The rest of the cores are E-cores, or even lower-powered cores.
............but they're actually fully functional cores, not like AMD's fake bulldozer cores. so yes, it's 52 cores. because E cores are still fully functional and you can even have a fully working system with nothing but e-cores.
Posted on Reply
#5
Muser99
“This 2L-ILM is a two-level independent loading mechanism, using two levers on each side of the CPU…”

Surely one leaver on each side!
Posted on Reply
#6
john_
AleksandarKYou tell me that having 52 cores is normal and not power hungry? Even if its e cores, they run at 3+ GHz, base TDP is pushed to 175, imagine what the boost will look like
No, what I’m saying is that you are describing a CPU with 52 P-cores, which is incorrect.

These E-cores aren't just P-cores running at lower frequencies. They are cores designed without certain features to make them more energy-efficient. Even if they were full fat P-cores running at lower frequencies, most people know how power consumption skyrockets as frequencies increase.
In any case, I would expect the text to read something like this:
'NVL-S desktop processors will feature up to 52 cores, with up to 16 of them being high-frequency, power-hungry P-cores.

With these changes, the argument is not weakened; rather, the description of the NVL-S CPUs is more accurate and avoids creating false impressions.
Flying Fish............but they're actually fully functional cores, not like AMD's fake bulldozer cores. so yes, it's 52 cores. because E cores are still fully functional and you can even have a fully working system with nothing but e-cores.
So you are saying that because AMD fans say this and that, TPU should somehow bent the description of hybrid Intel CPUs to make them look better? You sure this is a good argument?

The only ones insisting on that example are Intel fans. Most AMD fans didn't consider those CPUs as having 'up to 8 cores.' That’s why Bulldozer CPUs were losing even to Thuban in some benchmarks. It was a bad design.
That being said, making excuses for multi-billion dollar companies that sell configurations like—to use an extreme example—1 P-core and 51 E-cores as a '52-core CPU', is not something I want to do. I guess you wouldn't have a problem with that.
Posted on Reply
#7
Flying Fish
john_That being said, making excuses for multi-billion dollar companies that sell configurations like—to use an extreme example—1 P-core and 51 E-cores as a '52-core CPU', is not something I want to do. I guess you wouldn't have a problem with that.
I indeed wouldn't have a problem with that. Because it's technically correct. The best kind of correct.
Posted on Reply
#8
john_
dgianstefaniI expect 350 W+ if the long duration power limits are removed and it's allowed to boost as much as it does for short duration/low core load. Of course, that will still likely be very efficient, considering the amount of work being done.
700W for the top dual compute die model with those 52 cores (24 per compute die, plus 4 on the I/O). So, 350W per compute die.
Flying FishI indeed wouldn't have a problem with that. Because it's technically correct. The best kind of correct.
Good for you. Me, I would had a problem if they where selling me a box of "14 chocolate eggs" and I was getting this


I will agree that all those chocolate eggs will probably have the same sweetness and they are indeed FULL chocolate eggs.
Posted on Reply
#9
dj-electric
I was always a fan of the two lever system. good to have it back.
Posted on Reply
#11
Arctucas
Meh, it is coming off the board for direct die waterblock anyway...
Posted on Reply
#12
RejZoR
I don't understand why they even use these modern garbage sockets other than cost cutting with inferior products. The fact that both parts of socket frame aren't connected is just idiotic and it all looks and feels so flimsy, no wonder it's all bending and having terrible contact. It's good business for Thermalright and ThermalGrizzly selling contact frames that could be entirely avoided if factory socket frames weren't just plain shit.
Posted on Reply
#13
Sol_Badguy
- (almost) HEDT number of pins
- HEDT ILM
- HEDT-worthy number of cores
- supposedly HEDT performance
- HEDT TDP
- hopefully Celeron pricing? 🤞

Sounds about right. 🤪
Posted on Reply
#14
TheinsanegamerN
AleksandarKYou tell me that having 52 cores is normal and not power hungry? Even if its e cores, they run at 3+ GHz, base TDP is pushed to 175, imagine what the boost will look like
dgianstefaniE cores on the ~$300 7 tier 270K Plus boost to 4.7 GHz, with a base of 3.2 GHz. These are as performant as Raptor Lake P cores at iso frequency, so I doubt that Nova Lake E cores will have lower clocks or performance. Regardless of core type, having 50+ is going to draw power under full load, I expect 350 W+ if the long duration power limits are removed and it's allowed to boost as much as it does for short duration/low core load. Of course, that will still likely be very efficient, considering the amount of work being done.
A total package power draw does not make cores power hungry.

Describing all 52 cores as "power hungry" shows a remarkable lack of knowledge about CPUs, and frankly should be inexcusable for a news editor of a tech forum..... There are 16 P cores, the rest are e or ultra e cores, and the efficiency of those individual cores is quite good. An efficient CPu core does not become "power hungry" because you put a dozen of them together. The PACKAGE might be "power hungry", but that is a subjective opinion, and that does not make the statement "52 power hungry cores" correct.
Posted on Reply
#15
Flying Fish
john_Good for you. Me, I would had a problem if they where selling me a box of "14 chocolate eggs" and I was getting this


I will agree that all those chocolate eggs will probably have the same sweetness and they are indeed FULL chocolate eggs.
As a lover of cadbury mini eggs that is a terrible example for you to use because lil eggs are fantastic :D
TheinsanegamerNA total package power draw does not make cores power hungry.

Describing all 52 cores as "power hungry" shows a remarkable lack of knowledge about CPUs, and frankly should be inexcusable for a news editor of a tech forum..... There are 16 P cores, the rest are e or ultra e cores, and the efficiency of those individual cores is quite good. An efficient CPu core does not become "power hungry" because you put a dozen of them together. The PACKAGE might be "power hungry", but that is a subjective opinion, and that does not make the statement "52 power hungry cores" correct.
It depends on how you interpret it. I read it as 52 cores as a whole being power hungry and indeed 52 cores would be quite the gobbler of power
Posted on Reply
#16
rattlehead99
john_Oh, come on. Mistakes like this shouldn't be on TPU. That 52 should be 16. The rest of the cores are E-cores, or even lower-powered cores.

I hope the mechanism has been properly designed so that if someone doesn't engage both levers simultaneously and correctly, it doesn't create uneven pressure on one side, risking bending the socket pins.
So the 32x 4.7Ghz E-Cores that rival the IPC of ArrowLake P-Cores, are not power hungry? oook.
Posted on Reply
#17
phanbuey
rattlehead99So the 32x 4.7Ghz E-Cores that rival the IPC of ArrowLake P-Cores, are not power hungry? oook.
this

Those E cores are absurdly strong even on arrowlake - they're small but the density and size/count of cores puts out tons of heat.
Posted on Reply
#18
john_
Flying FishIt depends on how you interpret it. I read it as 52 cores as a whole being power hungry and indeed 52 cores would be quite the gobbler of power
Well, at least 4 cores, those low power E-cores on I/O die, are low frequency (2.5GHz I believe) and definitely not 'power hungry' cores. So, again the number 52 is wrong.
rattlehead99So the 32x 4.7Ghz E-Cores that rival the IPC of ArrowLake P-Cores, are not power hungry? oook.
phanbueyThose E cores are absurdly strong even on arrowlake - they're small but the density and size/count of cores puts out tons of heat.
Replying to both.
IPC is not a measure of power consumption. And as @TheinsanegamerN said above
TheinsanegamerNAn efficient CPu core does not become "power hungry" because you put a dozen of them together. The PACKAGE might be "power hungry"
Now, let's go AI. I asked Google's AI about Arrow Lake cores. We can guess more or less about Nova Lake.
So.
Power consumption under (stress test) per core
P-core (Lion Cove) 5.7 GHz ~25W+
P-core (Lion Cove) 4.6 GHz ~10W
E-core (Skymont) 4.6 GHz ~3-5W
and an estimation about the cores on Meteor/Lunar Lake I/O die
LP-E cores 0.5W - 0.8W
Posted on Reply
#19
phanbuey
john_Well, at least 4 cores, those low power E-cores on I/O die, are low frequency (2.5GHz I believe) and definitely not 'power hungry' cores. So, again the number 52 is wrong.



Replying to both.
IPC is not a measure of power consumption. And as @TheinsanegamerN said above



Now, let's go AI. I asked Google's AI about Arrow Lake cores. We can guess more or less about Nova Lake.
So.
Power consumption under (stress test) per core
P-core (Lion Cove) 5.7 GHz ~25W+
P-core (Lion Cove) 4.6 GHz ~10W
E-core (Skymont) 4.6 GHz ~3-5W
and an estimation about the cores on Meteor/Lunar Lake I/O die
LP-E cores 0.5W - 0.8W
Right but there's 4 of them in the same space as a single P-core - so 12W-20W per cluster vs 10-25W per P core - they're not all that different.

I've overclocked the 265K's e cores and was shocked at how much heat they could put out.
Posted on Reply
#20
john_
phanbueyRight but there's 4 of them in the same space as a single P-core - so 12W-20W per cluster vs 10-25W per P core - they're not all that different.

I've overclocked the 265K's e cores and was shocked at how much heat they could put out.
OK then, 16 power hungry cores and 8 power hungry clusters. Not 52 power hungry cores.
Posted on Reply
#21
Dr. Dro
john_Oh, come on. Mistakes like this shouldn't be on TPU. That 52 should be 16. The rest of the cores are E-cores, or even lower-powered cores.

I hope the mechanism has been properly designed so that if someone doesn't engage both levers simultaneously and correctly, it doesn't create uneven pressure on one side, risking bending the socket pins.
Remember, Arctic Wolf has about the same performance as, if not faster than Zen 4. That's a lot of punch on E-cores. It's also said that Nova Lake will come with instruction set parity on all cores, meaning they should all support AVX10.2 and APX. If all of these early rumors are true, this CPU will be a mean one.

Let's suppose each P-core uses 10 W and each E-core chugs 3, you're looking at the typical 250 W TDP they're usually rated for, and that's excluding the 4 LPE that are situated on the IO tile (that's just for the 48 cores in the CPU tile). I think more realistic these CPUs can and will use 350-400 W under AVX load and high frequencies, just like RPL can do 320-350 W today.
Posted on Reply
#22
Sol_Badguy
john_OK then, 16 power hungry cores and 8 power hungry clusters. Not 52 power hungry cores.
The Battle of Semantics Hill - the place where Sgt. @john_ fought valiantly defending the proper use of terms against overwhelming odds.
For his bravery he was decorated with the Semiconductor Cross.
Dr. Drothis CPU will be a mean one
1000 W tier 1 PSUs hiding in fear at the sight of the fire-breathing leviathan.
The reign of terror has begun!
Posted on Reply
#23
Dr. Dro
Sol_BadguyThe Battle of Semantics Hill - the place where Sgt. @john_ fought valiantly defending the proper use of terms against overwhelming odds.
For his bravery he was decorated with the Semiconductor Cross.

1000 W tier 1 PSUs hiding in fear at the sight of the fire-breathing leviathan.
The reign of terror has begun!
Haha yeah. Added to the 600W+ fire breathing behemoths from the leather jacket man.

Good thing I have a 1300 W unit :pimp:
Posted on Reply
#24
john_
Sol_BadguyThe Battle of Semantics Hill - the place where Sgt. @john_ fought valiantly defending the proper use of terms against overwhelming odds.
For his bravery he was decorated with the Semiconductor Cross.
Am I wrong? Is it ridiculous or extreme behavior in a tech site people to be able to differentiate between E-cores and P-cores?

@Dr. Dro
You probably haven't read the rest of the posts before posting. Everything is there for you to read.

P.S. Get ready to upgrade to 2.4KW PSU if you are considering DLSS5.
Posted on Reply
#25
Sol_Badguy
john_Am I wrong? Is it ridiculous or extreme behavior in a tech site people to be able to differentiate between E-cores and P-cores?
No, I agree with you. Post #15 also provides a very good argument.
But I thought of making a little joke because of your persistence.
Posted on Reply
Add your own comment
Nov 12th, 2026 04:42 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts