One notable aspect of the new generation is the inclusion of more hard logic than before. This is especially true for image and video encode and decode to handle data that’s commonly processed using such a platform. The family can support ultra-high-definition (UHD) video streaming and recording applications including multichannel, 8K video streams in real-time.
There’s also support for high-speed, 100-Gb Ethernet needed to move data in and out of the chip. It handles PCIe Gen 5 that enables support for the latest Compute Express Link 2.0 (CXL), including CXL-attached memory.
The hard-core compute infrastructure leverages the latest Arm Cortex-A78AE application processors (x8) and Cortex-R52 real-time processors (x10). These give a performance boost by a factor of ten compared to the previous generation. It also turns the chips into significant compute platforms even without the additional acceleration units and programmable logic.
Cores can be configured for lockstep support, which is often required for safety-critical applications such as automotive electronic control units (ECUs). The Arm Mali-G78AE GPU provides additional computation and display capabilities.
Security is addressed from the start with this incarnation of Versal. Its secure boot employs post-quantum cryptography. This means using newer encryption algorithms that are resistant to attacks by algorithms running on quantum computers. The chips can be certified for automotive security standards like ISO 21434.
Introducing AMD Versal Gen 2: Advances in Performance and AI
The artificial-intelligence engines (AIEs) in Gen 2 deliver twice the performance capabilities and move control logic from the FPGA fabric into hard logic (Fig. 2). Many of the AIE control processors have been moved to more efficient hard logic. The programmable logic still incorporates AI support, plus it allows for access to the AIEs to handle AI/ML-related datatypes like INT8.