Intel has released higher-resolution slides detailing their Skymont architecture to address the weaknesses in their prior Atom architectures, specifically with vector execution. Skymont aims to provide higher performance across various applications by adding more clusters for better decode capabilities. The unique “Nanocode” implementation increases decode throughput for complex instructions. The wider frontend and backend, along with improved memory subsystem and cache setup, enhance overall performance. Skymont also features increased vector and floating-point throughput, comparable to Arm’s Cortex X2. While Skymont doesn’t clock as high as Intel’s P-Core, it represents a significant step forward for Intel’s E-Cores. The architecture’s balance of performance per clock and clock speed makes it comparable to Arm’s flagship cores while focusing on lower power usage. Intel’s hybrid strategy combining E-Cores and P-Cores allows for customization to specific targets but lacks some ISA features like AVX-512 support.
https://chipsandcheese.com/2024/06/15/intel-details-skymont/