Cerebras Systems announced the CS-4 server to accelerate AI chatbot queries and compete with Nvidia. The system integrates three massive, dinner-plate-sized WSE-3 Turbo chips. This hardware utilizes the Nexus architecture with pluggable modules for simplified data center installation.
The CS-4 delivers twice the speed of the previous CS-3 model. It provides up to 30 times the performance of competing GPU-based solutions for specific AI tasks. TSMC manufactures the WSE-3 Turbo chips using a 5-nanometer process.
Cerebras is targeting the AI inference market to generate responses from trained models. The system is scheduled for release in the third quarter.