Fresh leaks are painting a clearer picture of Intel’s upcoming Nova Lake-S desktop CPUs, and the focus this time isn’t just raw core counts. New details touch nearly every enthusiast hot button: overclocking behavior, how Intel is reorganizing core clusters, what kind of AI acceleration to expect from the updated NPU, and why the eye-watering “700W” number needs context.
Overclocking: LP-E cores are off-limits, unlocked chips still target enthusiasts
One of the biggest clarifications involves the low-power efficiency (LP-E) cores on Nova Lake-S. These LP-E cores, reportedly based on Intel’s Arctic Wolf E-Core design, are said to be locked out of overclocking entirely. The LP-E cluster (a group of four cores on a low-power island) won’t respond to common clock-tuning methods like BCLK or ECLK adjustments. In other words, even if you’re an experienced tweaker, the LP-E cores are designed to stay within their intended power-efficient operating range.
That said, Intel’s unlocked “Core Ultra 400-K” Nova Lake desktop processors are still expected to offer enthusiast-friendly tuning for the main cores, including IA (core) overclocking, BCLK overclocking, and memory overclocking. The catch is that the full suite of OC features is expected to be concentrated on higher-end Z990 motherboards, while lower-tier chipsets may scale back some tuning options.
Boot flexibility: run without P-cores, and even disable entire compute dies
Another intriguing leak suggests Nova Lake desktop CPUs may allow extremely flexible core enablement. It should be possible to boot with P-cores disabled, using only E-cores, only LP-E cores, or a combination of E-cores and LP-E cores. There’s also mention of the ability to disable entire compute dies, a feature that could matter most for higher-end dual compute tile models. In practical terms, that could open up options for lower-power operation, troubleshooting, or even more targeted overclocking and thermal management by shutting off part of the silicon.
New core cluster layout: P-cores are clustered now, not just E-cores
Intel is also reportedly changing how workloads are organized at the physical core-cluster level. Previously, clustering was largely associated with E-cores and LP-E cores, but Nova Lake is said to cluster P-cores as well.
The reported arrangement groups two P-cores into a single cluster. Each P-core cluster is expected to include 4MB of L2 cache total (effectively 2MB per P-core). The important side effect: cores may only be disabled by cluster, not individually. Disable one P-core cluster and you lose two P-cores at once. Likewise, disabling an E-core or LP-E cluster would mean dropping four E/LP-E cores at a time. For power users, that changes how fine-grained core tuning and configuration might work compared with older designs.
NPU upgrade: a major leap in desktop AI performance
Perhaps the most headline-grabbing spec is the rumored NPU performance target. Leaks point to Nova Lake delivering up to 74 TOPS of NPU throughput, claimed to be around 5.6x faster than the current Arrow Lake lineup. If accurate, that would position Nova Lake as a serious “AI PC” desktop platform, with more headroom for on-device AI tasks like creative tools, real-time enhancements, transcription, and other NPU-accelerated workloads that are increasingly being built into modern apps.
Manufacturing is also rumored to be split across two leading-edge nodes: Intel 18A and TSMC N2. While it’s not confirmed which tile uses which process, the broader direction suggests Intel may mix process technologies across tiles and configurations, similar to the multi-tile strategies seen in other recent designs.
About that 700W number: it’s a “limits removed” scenario, not typical operation
The earlier figure claiming Nova Lake-S could hit over 700W understandably raised eyebrows. Updated context suggests that number reflects a dual compute tile CPU pushed with power limits removed, essentially a worst-case “unrestricted” scenario for a top configuration rumored to reach as many as 52 cores. It’s more like measuring what happens when the chip is allowed to stretch its legs with no guardrails, rather than what most users should expect in normal workloads.
More realistic guidance suggests Nova Lake-S could keep a familiar base power target in the 125W to 150W neighborhood (PL1), while turbo power limits (PL2) for the dual compute tile models may land somewhere in the 250W to 450W range depending on SKU and motherboard settings. The 700W+ region is being associated with extreme spikes and ceiling behavior when limits are removed or when short-duration protection thresholds come into play.
Nova Lake-S vs. Arrow Lake-S: what the leaked comparison implies
Based on the leak roundup, Nova Lake-S could represent a massive generational jump on paper, especially at the high end. The rumored top configuration goes up to 52 cores total (up to 16 P-cores, 32 E-cores, plus 4 LP-E cores), compared with Arrow Lake-S topping out at 24 cores (8 P-cores and 16 E-cores). Memory and platform I/O are also expected to move forward, including faster DDR5 support and more PCIe lanes, alongside a new desktop socket (LGA 1954).
Launch timing and the bigger battle in late 2026
Nova Lake-S and its new 900-series motherboard platform are currently expected later this year, setting the stage for a high-stakes desktop showdown in the second half of 2026. Intel’s strategy appears to be a blend of more cores, stronger AI acceleration, and a platform geared toward both enthusiasts and next-gen “AI PC” features—right as AMD prepares its own next wave of architectural and platform updates.
If these leaks hold, Nova Lake-S won’t just be about brute-force performance. It’s shaping up to be a more configurable, more AI-focused, and more aggressively scaled desktop platform than Intel’s current generation—provided you have the cooling, motherboard, and power delivery to match.
