Calligo Technologies Unveils TUNGA, World's First Posit-Based RISC-V CPU

Hyderabad-based Calligo Technologies has announced TUNGA, claimed to be the world's first eight-core RISC-V CPU implementing the Posit number format for AI and scientific computing workloads. Backed by the Digital Lighthouse Initiative, TUNGA delivers higher accuracy than IEEE 754 floating point at equivalent precision bits, targeting autonomous systems, climate modelling, and AI inference tasks where floating-point errors accumulate. The chip is targeted at research institutions and industrial AI deployments.

Semicon Hunt -> product-launch -> Calligo Technologies

2026-09-01

Featured image
Feature icon

TUNGA: India Produces the World's First Posit-Enabled Processor

Calligo Technologies, a Hyderabad-based semiconductor design company supported by the Digital Lighthouse Initiative, has announced the tape-out and functional validation of TUNGA, an eight-core RISC-V processor that it claims is the first in the world to implement the Posit number system as a native arithmetic format alongside conventional IEEE 754 floating point. TUNGA represents a genuine architectural innovation rather than an incremental improvement on existing designs, and its announcement places India alongside the most advanced research-driven processor projects globally. The chip targets AI inference, scientific simulation, and signal processing workloads where the accumulation of floating-point rounding errors degrades result quality or model accuracy over extended computation sequences.


What Posit Arithmetic Delivers Over IEEE 754

The Posit number format, developed by John Gustafson and introduced in 2017, is a variable-precision floating-point representation that allocates more bits to high-value numeric ranges and fewer to extremes, unlike the fixed-exponent-field structure of IEEE 754. The practical result is that Posit arithmetic produces more accurate results for the same number of bits in a wide range of numerical computations, including neural network inference, where accumulated quantisation error is a persistent challenge in deploying int8 and fp16 models without accuracy loss relative to full fp32 baselines. TUNGA's native Posit arithmetic units allow software to invoke Posit operations without emulation overhead, making Posit a practical tool for production AI workloads rather than a research curiosity confined to software simulation.

TUNGA Architecture and Eight-Core Design

TUNGA implements the RISC-V RV64GC instruction set across eight cores arranged in a shared-memory symmetric multiprocessing topology, with RISC-V custom instruction extensions that invoke Posit arithmetic operations natively. Each core includes both IEEE 754 and Posit floating-point execution units, allowing software to mix the two number formats within a single application as the precision requirements of different computational stages vary. The chip is manufactured on a 28nm process node, which is commercially accessible and cost-effective for a first-generation research and early adoption product. The design includes a tile-based cache hierarchy and coherent memory interconnect suitable for multi-threaded numerical workloads.

Target Applications and Go-to-Market Strategy

Calligo is positioning TUNGA for three primary market segments: academic and national laboratory research institutions running computational physics, climate, and genomics simulations that require high numerical accuracy; industrial AI applications including autonomous vehicle sensor processing and industrial defect detection where model accuracy is safety-critical; and defence signal processing applications where ITAR restrictions on foreign processors create demand for domestically designed alternatives. Calligo is working with DLI and the Defence Research and Development Organisation to evaluate TUNGA for specific defence programme requirements. Commercial development boards and an FPGA-based emulation platform for early software development are available to select partners ahead of the full production chip release.

View Calligo Technologies Company Profile