Intel senior vice president Cho Min-sung announced the upcoming Razor Lake-AX platform at the Semiconductors Powering Humanoids summit in South Korea. The new architecture is designed to deliver 700 TOPS of AI compute performance on die.
This performance level represents a 3.88-fold increase compared to the current Panther Lake mobile chips, which offer 180 TOPS. The boost in compute power is primarily achieved through a large integrated GPU and a substantial CPU complex.
The Razor Lake-AX platform will feature two versions of the integrated GPU, utilizing 16 or 32 cores based on the Xe3 graphics architecture. The flagship 32-core variant is expected to be nearly three times as powerful as the Arc B390 found in the PTL lineup.
The GPU die for the new platform is estimated to measure approximately 162.84 mm². Intel also intends to reintroduce on-package memory, specifically supporting LPDDR5X or LPDDR6 standards, despite previous indications that this design approach would end with Lunar Lake.
