Cassia Proposes ‘Better Math’ for AI Efficiency
By Sally Ward-Foxton, EETimes | October 1, 2025
Startup Cassia wants to ‘do math better’ by approximating mathematical functions and implementing these approximations in smaller, more power efficient hardware IP for AI acceleration.
The idea began as a spare time project for Cassia founder and CEO James Tandon after he came across scientific papers from another group that discussed the accuracy loss associated with approximating certain math functions rather than quantizing (reducing numerical precision, which reduces prediction accuracy).
To read the full article, click here
Related Semiconductor IP
- Dataflow AI Processor IP
- AI Processor Accelerator
- Powerful AI processor
- AI DSA Processor - 9-Stage Pipeline, Dual-issue
- High-performance AI dataflow processor with scalable vector compute capabilities
Related News
- CAST Introduces 400 Gbps UDP/IP Hardware Stack IP Core for High-Performance ASIC Designs
- Hardware Root of Trust Essential for AI Chip Integrity
- Rambus Introduces PCIe® 7.0 Switch IP with Time Division Multiplexing for Scalable AI and Data Center Infrastructure
- Titan IC Unveils Enhancements to RXP Hardware Search Acceleration Engine at RSA Conference
Latest News
- RISC-V Market Leadership Helped Andes Technology Drive Cumulative Shipments of AndesCore-Powered™ SoCs Beyond the 20 Billion Mark
- BrainChip Unveils Communication Reference Platform, Fueling Signal Intelligence at the Edge
- Qualinx and EUSPA announce the completion of full Galileo OSNMA support on QLX3Gx Series ultra-low power GNSS receiver.
- Ceva Launches Microsoft-Certified Spatial Audio Software for PC Gaming Headsets
- Nuclei and Lauterbach Successfully Complete World's First RISC-V Data Trace Solution