- Sagence brings simulated memory operations to redefine AI reasoning
- Reduce power consumption by 10 times and reduce cost by 20 times
- Integration with PyTorch and TensorFlow is also provided
Sagence AI introduces an advanced analog memory computing architecture designed to solve power, cost and scalability issues in: AI reasoning.
The architecture uses a simulation-based approach to improve energy efficiency and cost-effectiveness while delivering performance comparable to existing high-end GPU and CPU systems.
This bold move positions Sagence AI as a potential disruptor in a market dominated by: NVIDIA.
efficiency and performance
The Sagence architecture has advantages when processing large language models such as Llama2-70B. When normalized to 666,000 tokens per second, Sagence’s technology achieves 10x lower power consumption, 20x lower cost, and 20x smaller rack space than leading GPU-based solutions.
This design prioritizes inference needs over training needs, reflecting a shift in the focus of artificial intelligence computing within data centers. With its efficiency and affordability, Sagence provides solutions to the growing challenge of ensuring return on investment (ROI) as AI applications scale to large-scale deployments.
At the core of Sagence’s innovation is analog memory computing technology, which combines storage and computing within memory cells. This approach simplifies chip design, reduces cost and improves power efficiency by eliminating the need for separate storage and predetermined multiply-accumulate circuits.
Sagence also uses deep sub-threshold computing (an industry first) across multiple tiers of storage cells to achieve the efficiency gains needed for scalable AI inference.
Traditional CPU- and GPU-based systems rely on complex dynamic scheduling, which increases hardware requirements, inefficiencies, and power consumption. Sagence’s static scheduling architecture simplifies these processes and mirrors biological neural networks.
The system is also designed to integrate with existing artificial intelligence development frameworks such as PyTorch, ONNX and TensorFlow. Once a trained neural network is imported, Sagence’s architecture does not require further GPU-based processing, simplifying deployment and reducing costs.
“Fundamental advancements in artificial intelligence inference hardware are critical to the future of artificial intelligence. The use of large language models (LLM) and generative artificial intelligence has driven the need for rapid and large-scale changes in computing cores, which require the highest performance with the lowest power consumption. An unprecedented combination of performance and economics that aligns costs with value created.” Founder of Sagence AI.
“Today’s traditional computing devices capable of ultra-high-performance artificial intelligence inference are too costly, economically unfeasible, and consume too much energy to be environmentally sustainable. Our mission is to disrupt these capabilities in an environmentally responsible way and financial constraints.
through IEEE Spectrum