THE COURSE / M3 DESIGN TO MASK / M3.2
RTL and verification
After this lesson you can:
- Say what RTL is, and how a testbench, coverage and formal proof each show a design works.
- Explain why random testing alone misses rare bugs, and what catches them.
- Say why verification takes about half of a chip project, and what synthesis turns RTL into.
Loading the lesson…
Sources
- IEEE, 1800-2023, "IEEE Standard for SystemVerilog: Unified Hardware Design, Specification, and Verification Language" (standard page, read 10 November 2026). Active, published 28 February 2024, superseding 1800-2017. A unified hardware design, specification and verification language; models hardware at the behavioural, register transfer level (RTL) and gate levels; supports test benches "using coverage, assertions, object-oriented programming, and constrained random verification."
- H. Foster (Siemens EDA), "2024 Wilson Research Group IC/ASIC functional verification trend report," white paper, February 2025 (read 10 November 2026). n = 597. "Only 14 percent of IC/ASIC projects achieved first silicon success," the lowest in twenty years. "Logic or functional flaws" remain the leading cause of respins; design errors the leading root cause; changing, incorrect or incomplete specifications a common concern. 75 percent of projects behind schedule (historically about 66). Design engineers spent 51 percent of their time on design and 49 percent on verification; verification engineers spend more time debugging than on anything else; in processors "a 5-to-1 ratio of verification to design engineers is not unusual." Formal property checking adoption growing at 5.8 percent a year; code coverage, functional coverage, assertions and constrained-random simulation matured between 2007 and 2014.
- H. Foster (Siemens EDA), "2026 Functional Verification Study: Executive Findings," Verification Academy, 8 September 2026 (read 10 November 2026). 604 participants, 320 on IC/ASIC projects, with "particularly strong participation from CPU, GPU, and AI accelerator-class projects." "First-silicon success continues to decline"; "verification consumes roughly half of project time"; "most projects are behind schedule." The page gives no IC/ASIC percentages, and none is cited from it.
- Intel, fourth-quarter 1994 earnings release, 17 January 1995, filed with the SEC as exhibit 99 (read 10 November 2026). The Pentium processor's floating-point divide flaw causes reduced precision "once in every nine billion random number pairs"; replacing the processors reduced fourth-quarter gross margin by $475 million, a $0.70 per share impact on 1994 earnings.
- YosysHQ, Yosys documentation, "Introduction" (read 10 November 2026). An open-source Verilog synthesis tool, "the GCC of hardware synthesis"; it takes a behavioural design description and produces an RTL, logical-gate or physical-gate description; an HDL synthesis tool "generates a netlist that implements the given circuit"; gate mapping with ABC and mapping to cell libraries.