Unique Architecture Boosts Deep Learning Speeds

By Mat Dirjish

At CES 2020, PQ Labs launched QuantaFlow AI, a unique architecture that includes a classical RISC-V processor, QuantaFlow Generator, and a QuantaFlow Evolution Space. Reportedly, it’s the first such architecture with the potential to change the future of artificial intelligence (AI) and deep learning paradigms with the ability to boost deep learning speeds by 10 to 15 times.

QuantaFlow simulates a virtual transformation per evolution space for qf-bit registers. The single-core RISC-V processor provides logical control and result-observation retrieval.

The QuantaFlow Generator converts input data from low dimensional space to high dimensional space and then starts continuous transformations per evolution. The process is of minimum granularity, parallel in nature, and asynchronous.

By the end of the process, information needs to be extracted from the evolution space by the Bit Observer unit. In addition, Hot-Patching can be used to change the evolution path of qf-bits dynamically. When more significant deformations for the evolution space are needed, the RISC-V processor will issue a warm reboot to the evolution space.

All these operations can be executed extremely fast. With the help of these dynamic operations, QuantaFlow is capable of running numerous neural network models like ResNet-50 (2015), MobileNet (2017), EfficientNet (2019), etc., without speed degradation or hitting the memory wall.


According to PQ Labs, QuantaFlow can achieve a 10x speedup in ResNet-50 (batch=1, accuracy=93%, INT8) compared to Nvidia V100 in the same network configuration. For more information, checkout the PQ Labs’ QuantaFlow page.

And to stay ahead of the tsunami of sensor product and sensor application innovations, sign up for Sensors Daily Newsletters. It’s free, fast, and incredibly easy. For a media kit and unique and effective sponsorship opportunities, contact Michael Mitchell via email.