Intel’s Nova Lake (NVL) CPU lineup is set to be one of the company’s next big offerings in the desktop CPU segment, and given its market position in the client CPU segment, it is indeed necessary for Intel to nail this release. Nova Lake has been discussed for several years now, even by former CEO Pat Gelsinger, so we do know the company is making significant efforts with the lineup.
When we talk about discussions around Nova Lake, it appears that Intel plans to completely revamp the architectural design by changing tile configurations in a way that the company has never done before. While we’ll go more in-depth about this ahead, Intel has apparently realized that without finding a way to compete with AMD’s 3D V-Cache implementation, the company cannot consolidate its position in the CPU market, which is why with Nova Lake, we could expect the trend of ‘disappointments’ to break, which was previously created by the launch of Arrow Lake.
Intel’s Nova Lake CPU Lineup: An Extensive Look at Architectural Upgrades, “bLLCs?” & Tile Configuration
One of the biggest changes with Nova Lake this time is how Intel plans to revamp architectural configurations, not just by integrating advanced P/E-core microarchitectures but also by adopting newer implementations.
With Nova Lake, one of the earlier leaks we saw with the disclosure of possible SKUs within the desktop series, and in that particular instance, we saw that Intel plans to ramp up the onboard core count to 52 cores, a massive upgrade from Arrow Lake, whose top-tier SKU features 24 cores. We got some information about eight different models, which are as follows:
Core Ultra 9 –Â 16 P-Cores + 32 E-Cores + 4 LP-E Cores (150W)
Core Ultra 7 –Â 14 P-Cores + 24 E-Cores + 4 LP-E Cores (150W)
Core Ultra 5 –Â 8 P-Cores + 16 E-Cores + 4 LP-E Cores (125W)
Core Ultra 5 –Â 8 P-Cores + 12 E-Cores + 4 LP-E Cores (125W)
Core Ultra 5 –Â 6 P-Cores + 8 E-Cores + 4 LP-E Cores (125W)
Core Ultra 3 –Â 4 P-Cores + 8 E-Cores + 4 LP-E Cores (65W)
Core Ultra 3 –Â 4 P-Cores + 4 E-Cores + 4 LP-E Cores (65W)
Based on this analysis alone, Nova Lake is expected to be a significant uplift over Arrow Lake in on-paper specifications, with around 2.16x uplift in core/thread configurations, along with additional low-power island cores and higher TDP ratings. In terms of microarchitectures, Nova Lake will benefit from the power of Coyote Cove P-Cores and Arctic Wolf E-Cores, but one of the more interesting aspects of NVL is that Intel is introducing a dual-compute-tile option. With this layout, Intel is essentially moving LP-E cores to a separate SoC tile, along with 2x compute tiles each featuring 8 P-Cores and 16 E-Cores.
Nova Lake-S vs Arrow Lake-S
FamilyNova Lake-SArrow Lake-S
Core Count (Max)5224
Thread Count (Max)5224
Max P-Cores168
Max E-Cores3216
Max LP-E Cores40
Max Cache (L2+L3)160-320 MB76 MB
Max bLLC Cache144-288 MBN/A
DDR5 (1DPC 1R)8000 MT/s7200-6400 MT/s
PCIe 5.0 Lanes (Max)3624
PCIe 4.0 Lanes (Max)164
Socket SupportLGA 1954LGA 1851
Max TDP (PL1)125-175W125W
Max Power~700W (Dual)
~350W (Single)~400W
Launch2H 20261H 2026
It’s also important to note that Intel has new plans for overclocking with microarchitectures, since with Nova Lake, the LP-E core island won’t be affected by BCLK or ECLK adjustments. This is a crucial detail to note, since it leads to a discussion about power limits, which we’ll discuss ahead as well.
You might have guessed it right; Intel is indeed pivoting to AMD’s CCD route, but the more intriguing element of this move is that Intel plans to step into the “cache game” as well, and their implementation is called bLLC (Big Last-Level Cache). It’s an alternative to what AMD does with 3D V-Cache, and Intel is anticipated to introduce four different SKUs featuring bLLC, and based on a prior leak, here is how the expected configurations will pan out:
2x 8+16 (48 Cores + 4 LPE Cores) + 288 MB bLLC
2x 8+12 (40 Cores + 4 LPE Cores) + 288 MB bLLC
8+16 (24 Cores + 4 LPE Cores) + 144 MB bLLC
8+12 (20 Cores + 4 LPE Cores) + 144 MB bLLC
Image Credits: Wccftech (AI-generated)
The top-end 52-core and 44-core models will feature 144MB of bLLC per compute tile, giving Intel an edge over AMD. Team Red’s 3D V-Cache is currently mounted on just one of the compute tiles, leading to asymmetrical access and creating scheduling complexities for the OS when it wants to assign hefty workloads to cores with more cache. Intel’s bLLC on both compute tiles will allow symmetrical cache layout. With 144 MB of bLLC per tile, Team Blue is making scheduling easier, bridging the gap created by extreme latency penalties or the need for specific software whitelists.
Intel’s Nova Lake-S: New Socket Design, Extreme Power Ratings & Motherboard Configurations
Given the phenomenal upgrades with Nova Lake, one of the major changes is the transition to an entirely new socket design: LGA 1954. This change will bring overhead for motherboard manufacturers as well, in terms of shifting to a wholly new socket, and, more importantly, thermal/power constraints with NVL are a massive question. If you think the “dual compute” tile option sounds interesting, well, a recent leak suggested that the power consumption of such models could go up to a whopping 700W (~2x higher than Arrow Lake), with thermal ratings stated to reach up to 100 degrees, given the performance such SKUs will pack.
Intel’s Nova Lake “Preliminary” Power Limits (PL1/PL2/PL3):
Nova Lake-S (Dual Compute Tile) –Â 150W / 496W / 854W
Arrow Lake-S (Core Ultra 9 285K) –Â 125W / 250W / 425W
Raptor Lake-S (Core i9-14900KS) –Â 150W / 253W / 350W
Now, these power limits sound ‘absurd’ from a consumer perspective, but there’s one reason Intel has brought such ratings. The dual compute tile option won’t be targeted directly at gamers, but rather at HEDT (High-End Desktop) users, who are more focused on professional workloads such as 3D rendering, video editing, and complex simulations. This is one of the reasons most of the 900-series Nova Lake motherboards won’t support the full power of the 52-core model.
If you are a bit overwhelmed about “power” with Nova Lake, well, don’t be. The 700W delivery figure is just a number obtained by removing all power limits on the dual-compute tile NVL-S configuration. However, in reality, Nova Lake is expected to provide gamers with greater power-dynamics adjustment, including the ability to boot only E-Cores and even switch off a compute tile for light workloads.
We did talk a bit about motherboard designs earlier with Nova Lake. Fortunately, this time around, we also have an extensive leak around Intel’s “900-series”, which is positioned entirely towards NVL. The 900-series chipset family will include five PCH SKUs: Z990, Z970, W980, Q970, and B960, each designed to cater to multiple customer segments, from client to workstation. Here’s an extensive summary of each chipset and its respective configurations:
Intel 900-Series Chipsets (Source: Jaykihn):
Chipset NameZ990W980Q970Z970B960
Total PCIe Lanes4848443434
CPU USB4/TB4 Ports22211
DMI Gen5 Lanes44422
PCH PCIe 5.0 Lanes1212800
PCH PCIe 4.0 Lanes1212121414
SATA 3.0 Lanes88844
USB 3.2 20G Ports55422
USB3.2 10G Ports1010844
USB3.2 5G Ports10101066
IA OCYesNoNoYesNo
BCLK OCYesNoNoNoNo
Memory OCYesYesNoYesYes
ECC SupportNoYesNoNoNo
Displays Supported44444
Apart from the talk around core elements, a significant upgrade on Nova Lake comes with the integration of the sixth-generation (NPU6), and according to initial discussions, the power is expected to be rated at 74 TOPS or 5.6x faster than the existing Arrow Lake lineup. Intel already has very decent NPU capabilities and support in its Panther Lake/Lunar Lake SoCs, so this update would make Nova Lake the most premium AI chip for desktops.
Intel’s Nova Lake-S CPU Lineup: The Xe3P iGPU Architecture & Expected Performance
Intel’s Nova Lake desktop CPUs are set to feature the company’s Xe3P architecture, a more advanced version than the one in Panther Lake. One of our very first interactions with Xe3P was with Intel’s Crescent Island ‘inference-focused’ solution, which was announced at the ITT. While we are currently unaware of the architectural upgrades over Xe3P, it appears the uplifts won’t be significant, given that the generation is similar to Panther Lake.
In a previous leak, it was reported that Nova Lake’s iGPU will deliver up to 25% higher performance than Xe3. Given the massive performance uplift from Xe2 to Xe3, it is clear that Nova Lake’s iGPU will be capable. It is also claimed that, with Nova Lake-S, Intel could bring a hybrid graphical architecture, with Xe3P dedicated to graphical workloads and the next-gen Xe4 architecture for display output and video processing.
The above information clearly shows us that Intel is taking a massive gamble with Nova Lake, through microarchitectures, tile layout, and graphical upgrades. It would be interesting to see how this gamble actually plays out.
Intel’s Nova Lake-S CPU Lineup: The Expected Release Timeline
The Nova Lake launch timeline is rumored to be in H2 2026, which means the official unveiling should happen by next quarter if things continue at the current pace. However, we are a bit skeptical about what Intel plans for NVL at the moment, given that the demand for its server CPU offerings is forcing the company to put its client business on the back burner for now. More specifically, Intel’s Nova Lake series is claimed to feature a combination of its internal 18A and TSMC’s N2 nodes for its compute tiles, which could delay the launch due to supply constraints.
If the H2 2026 timeline remains consistent, we could expect an unveiling at this year’s Computex. If you see the Arrow Lake-Refresh series instead of NVL, well, you should eventually realize that Nova Lake could be pushed to 2027 as well.