October 17, 2023 at 10:48AM
MemComputing is a company that aims to break the von Neumann bottleneck in computing by combining processing and data in memory. This could potentially solve complex mathematical problems, such as prime factorization, that are currently difficult for classical computers. The company’s software emulation has shown promising results in reducing the time required for factoring large numbers. Their goal is to develop an Application Specific Integrated Circuit (ASIC) that can solve a 2048-bit factorization problem in minutes. If successful, this could have implications for current encryption methods.
The meeting notes discuss MemComputing, a company and computing philosophy that aims to break the ‘von Neumann bottleneck’ by combining processing and data in memory. This bottleneck refers to the latency introduced by having separate storage and processing and the need for communication between the two.
The computational complexity of certain mathematical problems increases exponentially, resulting in a long processing time for classical computers. This is particularly true for prime factorization, which is essential for RSA-based encryption. While current methods struggle to factorize large numbers within a reasonable timeframe, quantum computers have the potential to solve this problem more efficiently.
MemComputing’s approach, using combined memory/processing, aims to solve difficult problems more quickly. Simulation shows that the complexity/time ratio increases polynomially instead of exponentially. In a study, MemComputing used software emulation to test problems up to 300 bits and found that the necessary time for factorization followed a 2nd-degree polynomial in the number of bits.
The company aims to extend the effective range beyond 300 bits and develop an Application Specific Integrated Circuit (ASIC) to realize the capability. The researchers estimate that an ASIC could potentially solve a 2048-bit factorization problem in tens of minutes, but this is still theoretical and not yet demonstrable.
If MemComputing’s approach proves practical, it could have implications for encryption and potentially challenge the need for quantum computers. However, this is still a theory based on existing research and requires further development and testing.
Overall, MemComputing’s aim is to provide an alternative solution to the limitations of classical computing in solving complex mathematical problems, such as prime factorization, within a reasonable timeframe.