Intel Nova Lake bLLC limited to K-series: Core Ultra 5 435K with 144 MB cache
Leak: Intel Nova Lake bLLC limited to K-series, debuting with Core Ultra 5 435K (144 MB). Up to 60% performance uplift plus NPU and LPE cores and AI features.
Leak: Intel Nova Lake bLLC limited to K-series, debuting with Core Ultra 5 435K (144 MB). Up to 60% performance uplift plus NPU and LPE cores and AI features.
© RusPhotoBank
Fresh details have surfaced about Intel’s upcoming Nova Lake processors. Leaker Jaykihn clarified that the last-level bLLC cache will be supported only on K-series chips, despite earlier chatter suggesting a wider rollout. The move points to a deliberate, enthusiast-first introduction rather than a blanket feature across the lineup.
Earlier reports indicated bLLC would land in a Core Ultra 5 model; the insider now specifies it as the Core Ultra 5 435K, a preliminary configuration with 8 performance cores and 12 efficiency cores. This chip is expected to carry 144 MB of bLLC, positioned as an analog to AMD’s 3D V-Cache.
According to the source, the implementation may overlap with server-class Clearwater Forest solutions that use the Foveros Direct 3D interface. That kind of stack enables tighter integration and could make overclocking more straightforward, since the compute tile may interact directly with the heat spreader. If the design mirrors those server techniques, it would neatly align with the K-series focus on tuning.
The insider also claims Intel is not developing versions with two stacked bLLC modules. Meanwhile, rumors suggest AMD is preparing a Ryzen 9 9950X3D2 with expanded 3D V-Cache on both CCDs, for a total of 192 MB. A measured start from Intel is easy to understand, as Nova Lake will be the company’s first consumer platform to use this type of cache.
Nova Lake is expected to deliver up to a 60% uplift in multithreaded performance versus the current Arrow Lake generation. The platform will also be Intel’s first desktop lineup with an integrated neural processing unit and power-efficient LPE cores, similar to those seen in Panther Lake—signaling a broader push to blend cache-heavy designs with built-in AI acceleration.