3 Reasons to Migrate from UltraScale/UltraScale+ to Versal™

The explosion of data, growth of AI and machine learning, and rising demand for faster computing have driven the development of next-generation silicon built to accelerate sensor fusion applications. AMD’s new Versal™ family, an adaptive computer acceleration platform (ACAP), combines traditional FPGA with hard IP blocks. These advancements offer powerful processing, AI, and communication capabilities that drive cutting-edge applications, particularly in the military and aerospace sector.

Despite the Versal series’ potential, migrating from UltraScale/UltraScale+ series silicon requires significant time and resources — and comes with inherent risk. Still, there are three compelling reasons to migrate: obsolescence, increased capability, and new advancements.

Obsolescence

UltraScale, like any generation of silicon, has a finite lifespan, and engineers have already been deploying it for years. The fear is if you don’t upgrade now, you could face a hardware gap where your existing FPGA becomes unavailable before the lifecycle of your program ends. This risk is critical in the aerospace and defense (A&D) market, where platform or system lifecycles often span decades.

In this scenario, engineers aren’t necessarily seeking new capabilities. Rather, they just want to port their existing code and functionality to a newer, supported silicon platform to extend the lifecycle of the system. That’s understandable, given the significant investment in developing code – often the “secret sauce” of a military system – which is cost-prohibitive to recreate simply to maintain existing functionality.

However, this scenario presents challenges, as some functionality and connections do not automatically port over to the Versal family. For example, while DSP and some I/O blocks automatically migrate, soft memory controllers, AXI Interconnects, PCIe subsystems, and processors and peripherals do not.

Partial migration from UltraScale to Versal isn’t the best option as it compromises performance. Instead, AMD advises engineers to rearchitect connections, use Versal’s integrated blocks, replace PL primitives with RTL descriptions or XPMs, regenerate IP blocks, and fully resynthesize the design.

Increased Capability

The second path to Versal migration is when engineers want to maintain current functionality while also adding capabilities made possible by Versal’s enhanced performance and bandwidth. Many modern applications require PCIe Gen 4 or 5, but UltraScale/UltraScale+ FPGAs only supports PCIe Gen 3. In addition, legacy platforms max out at 40G (Kintex) and 100G (Virtex) Ethernet, while Versal is capable of exceeding 100G.

Engineers need more PCIe connections between FPGA, GPUs, and SBCs. Previously, a PCIe gen3x8 link might have been sufficient, but now, two or more PCIe gen4x8 links might be required to move enough data through optical interfaces. Engineers can also take advantage of more transceivers, additional hard silicon resources, and better routing with the Versal family.

New Advancements

The third motivator for migrating from UltraScale/UltraScale+ to Versal is a need for genuinely new capabilities. In this scenario, programmers aren’t trying to replicate legacy designs. Rather, they’re leveraging Versal’s AI cores for real-time inference or improved DSP slices. Technology refresh cycles, the need to stay ahead of emerging threats, and incremental performance gains are all enhancements that may be achievable with Versal — but were often out of reach with previous generations of silicon.

For example, a New Wave Design client sought to add real-time sensor fusion algorithms to a mission processing subsystem, requiring massive matrix mathematics well beyond the capabilities of traditional DSP-based signal processing. New Wave Design helped develop an AI-driven solution that provided low latency and high efficiency for tasks like image recognition and radar signal fusion using a Versal AI Core series FPGA.

As performance demands continue to rise, engineers face a critical decision: should they migrate from UltraScale/UltraScale+ to Versal? The answer starts with a clear understanding of why migration is being considered in the first place, because that reasoning will shape the entire migration strategy. Ultimately, your design approach must align with your overarching goal, whether it’s to maintain functionality, enhance performance, or add new capabilities.