Custom Mass Multiplication Circuits to revolutionize AI Computation
In a ground-breaking leap forward for artificial intelligence and computational efficiency, the United States Patent and Trademark Office (USPTO) has officially approved Gigantor Technologies‘ Patent Number 11,748,061, detailing the innovative Custom Mass Multiplication Circuits.
This invention marks a monumental shift in AI processing by utilizing mass-scale calculations in the order of millions, billions, and even trillions to achieve unparalleled speeds through intricate layered directed graphs.
Traditional AI processing often faces limitations due to finite computational resources, hindering its ability to perform complex calculations at high speeds. Gigantor’s newly patented Mass Multiplier technology bypasses these constraints by implementing a unique non-Von Neumann approach.
By integrating massive arrays of calculations into a unified process, the Custom Mass Multiplication Circuits harness the power of collective computation, enabling calculations that dwarf even the most advanced GPUs and TPUs in the market.
CTO Mark Mathews, said: “The core bottleneck for CNNs is the sheer volume of multiplications required to compute each node of an abstract mathematical graph. While GPUs and TPUs simulate the operation of an abstract mathematical model with instruction streams, GigaMACS™ implements the model directly in hardware, which yields the optimal combination of maximum performance at minimum power.”
The newly patented technology’s potential applications are far-reaching, encompassing high-speed image recognition, robotics, augmented reality, and autonomous systems. GigaMACS™ pipeline architecture provides ultra-high performance with near-zero latency, processes HD and 4K frames at 240 FPS or higher, and enables Unlimited Object Recognition with our Synthetic Scaler technology.
The capability to harness vast computational power without compromising performance is ushering in a new era of AI advancements.
