FPGA architectures include increasingly complex arithmetic operators and optimized hard IPs, such as memory subsystems and Networks-on-Chip (NoC). This evolution leads to higher compute density also linked with high memory bandwidth. It represents an opportunity to tailor an architecture to niche application needs while being competitive with a costly ASIC implementation. More specifically, scientific computing requires high precision (> 32 bits) floating point computation. However, GPU vendors are progressively favoring low precision performance for AI needs, and are even phasing out support for 64-bit floating point compute. We present an analytical study motivating the need to investigate the implementation of an open source 64-bit GPGPU architecture on a state of the art FPGA, as an alternative to GPUs for scientific computing.

In the context of the Horizon Europe project, METASAT, a hardware platform was developed as a prototype of future space systems. The platform is based on a multiprocessor NOEL-V, an established space-grade processor, which is integrated with the SPARROW AI accelerator and connected to a GPU, Vortex. Both processing systems follow the RISCV specification. This is a novel hardware architecture for the space domain as the use of massive parallel processing units, such as GPUs, is starting to be considered for upcoming space missions due to the increased performance required to future space-related workloads, in particular, related to AI. However, such solutions are only currently adopted for New Space, since their limitations come not only from the hardware, but also from the software, which needs to be qualified before being deployed on an institutional mission. For this reason, the METASAT platform is one of the first endeavors towards enabling the use of high performance hardware in a qualifiable environment for safety critical systems. The software stack is based on baremetal, RTEMS and the XtratuM hypervisor, providing different options for applications of various degrees of criticality