AMD has presented the second generation of Versal adaptive ‘systems on chip’, a new device that seeks to offer end-to-end acceleration in embedded systems with a unified approach to computing resources.
The semiconductor firm overcomes the need to create multi-chip solutions to address the various limitations presented by embedded systems driven by artificial intelligence.
Instead, brings together all computing resources on a single chip; an optimal combination of processors for the three phases of embedded systems acceleration, that is, preprocessing, AI inference and postprocessing.
Versal AI Edge Series Gen 2 presents, on the one hand, an FPGA programmable logic structure for real-time preprocessing to connect to a wide range of sensors and implement high-performance, low-latency data processing pipelines.
On the other, an array of vector processors in the form of next-generation AI engines for efficient AI inference. Added to this are CPU cores that provide the post-processing power necessary for complex decision making and control of safety-critical applications.
He has also presented Versal Prime Series Gen 2, In this case, aimed at the classic integrated systems, in which it offers end-to-end acceleration by combining programmable logic for sensor processing with high-performance integrated Arm CPUs.
With this design, AMD offers up to ten times more scalar computing than with the first generation, more efficient management and video signal processing that supports up to 8K in multi-channel workloads.