GPU Memory Hierarchy: Registers, Shared Memory, L1/L2, and HBM3e/HBM4
Deconstructing the GPU nanosecond memory ladder: from single-cycle register files and banked SRAM shared memory to 8 TB/s 3D stacked HBM.
Read Post →Featured Projects
View All Projects →Latest Blog Posts
Read All Posts →The Tensor Core Revolution: From FP32 Matrix Multiplication to FP4/FP8 Acceleration
How systolic matrix engines redefined AI hardware: GEMM dataflow, Volta to Blackwell generational evolution, FP8/FP4 microscaling, and roofline dynamics.
Read Post →Gemini 4 Hydrogen: 2.5x Faster Agentic Execution, 1M Unified Context, and the Antigravity Orchestrator
Inside Gemini 4 Hydrogen: outscoring Argon (51 vs 50) at $1.50/M, 57% on Terminal-Bench 4.1, 1M long decodes, and dynamic Orchestrator handoffs.
Read Post →