Why tryriscv.org
The number of RISC-V devices has grown quickly over the past three years, but the information is fragmented: vendor sites describe “typical applications”, datasheets list “electrical characteristics”, distributors quote “pricing”, and the community shares “lessons learned the hard way”. For a team that has to choose parts, what is missing is not more documents but a table you can filter, compare, and trace back to its sources.
tryriscv.org turns public information into structured data so that a real requirement — “around 100 MHz, low power, 64 KB of memory, needs BLE” — narrows down in minutes to a handful of parts worth prototyping.
Where the data comes from
- Vendor product pages, datasheets, reference manuals, and selection guides;
- Official vendor repositories on GitHub, SDK documentation, and toolchain pages;
- Public alliance, community, and distributor material (used for cross-checking only).
Every record keeps sources and confidence fields. Parameters that are not published are left as null rather than guessed; µA/MHz figures derived from mA values state their conditions in the power notes.
Current coverage
- MCU: WCH, GigaDevice, Espressif, Bouffalo Lab, APT, CMiot, HPMicro, Telink, and more as data is added;
- IP: T-Head XuanTie, Nuclei, StarFive Dubhe, SiFive, Andes, and others;
- Solutions: IoT industry-education, low-power sensing, BLE, Wi-Fi, motor control, industrial control;
- AI compute: preview; hardware and toolchain information for NPU IP and AI SoCs.
Excluded non-RISC-V parts (correction)
Winner Micro W800 / W801 / W806 use the XT804 core (XuanTie E804, C-SKY architecture); their official SDKs ship the csky-elfabiv2 toolchain, so they are not RISC-V. They were previously listed as RV32IMAC by mistake and have been removed from the MCU catalog. Parts with 8051 or ARM cores are likewise excluded; report any other misclassified part via an issue.
Roadmap
- Complete power, package, and pricing data for Chinese RISC-V MCUs, with a quarterly review process;
- Add a maturity matrix for IP / AI-component / model / software-version matching;
- Add community measurements (power, boot time, RTOS ports, model inference) and reproducible benchmarks;
- Add reference schematics, BOMs, and lab suggestions to the solution packages.
Neutrality and corrections
This site does not accept paid vendor rankings. The default order combines data completeness, low power, and cost, and you can switch it freely. If you find a wrong parameter, a dead link, or a missing part, open an issue on GitHub with the part number in the title.
License
The code is released under the MIT license; the curated data is released under CC BY 4.0. When you republish it, credit the source and link back to this page.