Industry’s first ACAP from Xilinx

Xilinx developed the world’s first adaptive compute acceleration platform (ACAP) combining the power of software and hardware programmability. It is now available to early access customers. All the rest will have access to the solution in the second half of 2019. The designed combination will definitely change the existing computer landscape by providing a platform suitable for any application and any developer: from hardware engineers to data scientists. We’ll try to explain why.

Meet Versal XCVC 1902 — the platform formed to succeed traditional GPUs, CPUs, and FPGAs. It is built using TSMC’s 7-nm FinFET process technology and integrates the following parts: 

  • Scalar processing engines: Arm Cortex-A72application processor, Arm Cortex-R5 real-time processor, and platform management controller for fast 
  • Intelligent engines: DSP engines with high-precision floating-point and low latency providing granular control for customized datapaths. Also, there are high throughput, low latency, and power-efficient AI engines which are ideal for AI interference and advanced signal processing.
  • Adaptable hardware engines: re-architected foundational HW fabric for greater computer density which enables custom memory hierarchy and “on-the-fly” or 8 times faster dynamic reconfiguration.
  • Network-on-Chip (NoC): The truly innovative solution connecting the Versal's key interfaces, engines, and controllers and high bandwidth and low latency to dynamically customizable accelerated computing solutions.

You may see the block diagram of ACAP in the picture below.

 

Source: Xilinx

 

No doubt the solution has a revolutionary impact on the whole industry. Stay tuned to follow the new applications based on the platform.