30/05/2026
I’m happy to share a meaningful milestone from our open hardware education project at EdgeChipLab.
Our custom pure-Verilog RV32I CPU core, EdgeChipLab_CPU_v1, has successfully passed the RISC-V RV32I Architectural Compatibility Tests (ACT v3.0).
Highlights:
• Pure RTL educational CPU core
• Cycle-by-cycle bit-true regression against Spike
• 48 architectural tests
• 47 passed + 1 waived (misaligned jump corner case with detailed public analysis)
• Public acknowledgement from the official RISC-V Arch-Test maintainer
GitHub Issue: https://github.com/riscv/riscv-arch-test/issues/1600
Compliance Repository: https://github.com/estlit/EdgeChipLab_RV32I_Compliance_Report
This milestone is especially meaningful because it directly connects to our broader educational roadmap:
RTL Design → RISC-V CPU → Architectural Verification → NPU Integration → FPGA-based Edge AI System
At EdgeChipLab, our goal is to make advanced semiconductor system design transparent and teachable through complete hands-on implementation.
The newly verified EdgeChipLab_CPU_v1 is planned to be integrated with our custom RTL-based AI NPU into a complete memory-mapped SoC platform on FPGA.
This next-stage integration is not simply connecting CPU and accelerator blocks—it requires a deep understanding of how an NPU processes data internally, how memory interfaces are organized, and how bit-true hardware verification is maintained across the entire system.
For students and engineers interested in following this development, I recommend studying the NPU architecture and verification flow presented in my Amazon book:
“AI NPU System Design with Python and Verilog”
by Roger Kim (Hyo Seob Kim)
Amazon: https://www.amazon.com/dp/B0GLQVJWMK
The book explains how a custom RTL-based neural processing unit can be designed from scratch, verified bit-by-bit using Python and Verilog, and implemented on FPGA.
This foundation becomes especially valuable as we move toward:
Verified RISC-V CPU + Custom NPU + Memory-Mapped SoC + FPGA Implementation
—from architecture fundamentals to a complete edge AI semiconductor system.
My hope is that students and engineers can learn not only individual modules, but the full engineering path of building, integrating, and verifying practical semiconductor systems from scratch.
Dear RISC-V Arch-Test Team, We are pleased to report that our custom pure RV32I RTL core, EdgeChipLab_CPU_v1, has successfully passed 100% of the Base Integer Architectural Certification Tests. We ...