Revolutionary Compute-In-Memory APU Delivers GPU-Level AI Performance with Minimal Energy Usage

GSI Technology has unveiled a revolutionary advancement in artificial intelligence processing with its Compute-In-Memory (CIM) technology. Recent research from Cornell University reveals that GSI’s Gemini-I Associative Processing Unit (APU) can achieve GPU-level performance for large-scale AI applications while significantly reducing energy consumption.

Key Findings from Cornell University Study

The study confirms numerous advantages of GSI’s CIM architecture:

• GPU-Class Performance: The Gemini-I APU demonstrated throughput comparable to NVIDIA’s A6000 GPU during retrieval-augmented generation (RAG) tasks.
• Energy Efficiency: The APU consumes over 98% less energy compared to GPUs when processing large datasets, highlighting its sustainability.
• Speed Advantages: The innovative design of the APU allows for retrieval tasks to be completed up to 80% faster than traditional CPUs.

Industry Impact and Future Prospects

Lee-Lean Shu, the Chairman and CEO of GSI Technology, expressed confidence in the findings, stating that CIM technology could disrupt the $100 billion AI inference market. The APU’s performance and efficiency make it a viable option for various sectors, including:

• Edge AI applications in robotics, drones, and IoT devices.
• Defense and aerospace industries that require high performance within strict energy constraints.

The research paper, titled “Characterizing and Optimizing Realistic Workloads on a Commercial Compute-in-SRAM Device,” is published in ACM and was presented at the Micro ’25 conference. This analysis represents one of the first thorough evaluations of commercial CIM devices under realistic performance conditions.

Advancements in GSI Technology APUs

GSI Technology is continuously innovating, recently releasing the second generation of its APU silicon, Gemini-II. This new silicon promises approximately ten times faster throughput and enhanced energy efficiency for memory-intensive AI workloads. Looking forward, the company anticipates further advancements with the upcoming Plato APU, which aims to offer even greater compute capabilities at reduced power for edge applications.

Conclusion

GSI Technology’s cutting-edge APU technology signifies a transformative step in AI processing. With an emphasis on energy efficiency and high performance, GSI is poised to capitalize on a growing demand for sustainable computing solutions across various industries.