ZEN SEMICONDUCTOR
Compute & Cores

Configurable RISC-V Core

A configurable RISC-V core that mines the custom instructions your code actually needs — and generates only the logic to run them. A clean, lean alternative to the heavyweight "custom cores" from Cadence Tensilica and MIPS ARC, without the baggage they carry.

LLVM instruction miningOverhead-free RTLCustomer-owned
Datasheet · coming soonWatch demo
How it works

From your code to owned silicon.

Instruction-mining flow Your sourceC / C++ / kernels LLVM mininghot-primitive analysis Mined extensionscustom ISA ops Generated corefetch · decode · execonly logic used OwnedRTL
Only the logic your mined instructions require is generated — no dead features, and you own the result.
Why it's different

No needless overhead.

Incumbent configurable cores are bloated by design. Their generators bolt every optional feature onto a fixed, general-purpose template — so the RTL you ship is full of logic you never use. In a typical offering, adding even a single unused instruction — a NOP — can drag in on the order of 100,000 lines of supporting RTL.

The difference

We generate only what your design uses.

Our core is synthesized from the instructions your code mines — nothing more. The result is dramatically smaller, cleaner RTL: faster to verify, easier to close timing, cheaper in area and energy, and far simpler to own and maintain. Less silicon doing exactly your job.

Optimize for your target

One core, your priorities.

Minimum energy is only one dial. Tell us what matters most and we tune the mined extensions and microarchitecture to hit it.

Target

Maximum Performance

Widen and specialize for the hottest kernels — the most throughput per clock for your workload.

Target

Minimum Energy

Fuse operations and strip waste so each result costs the fewest joules — ideal for always-on and edge.

Target

Minimum Area

Smallest footprint that still runs your code well — the most function per square millimeter.

How it compares

The honest custom core.

Configurable processors aren't new — Cadence Tensilica and MIPS ARC built the category. What's new is the method and the discipline: we mine extensions automatically from your source with an LLVM-based pipeline, generate only the logic those extensions require, optimize to your chosen target, and deliver a core you fully own.

Specifications

Base ISARISC-V, with mined custom extensions
Generated RTLOnly the logic your instructions use — no dead features
Optimization targetsMaximum performance · Minimum energy · Minimum area
Instruction discoveryLLVM-based analysis of your code's hot primitives
DeliverablesCustom core + GCC/Clang toolchain + instruction-set simulator
Category peersCadence Tensilica, MIPS ARC
DeliverySynthesizable RTL — customer-owned
Back to Products