Apex Compute Developing Open-Source Mesa Vulkan Driver For Their Hardware

Apex Compute is the company started in California in 2024 that aims to produce high-efficiency AI accelerators for real-time edge AI inferencing. So far they have an FPGA prototype for licensing and deployment and carries the bold claim of being "20x faster than the NVIDIA Jetson" and at under 10W and one fifth the cost. Making Apex Compute much more interesting to us now is that they are developing an open-source, Mesa-based Vulkan driver for their hardware. Unlike most of the start-ups running their own half-baked software stacks and a proprietary mess, Apex Compute is developing an open-source Vulkan driver leveraging Mesa. This immediately makes the start-up much more interesting than many others out there. Plus using Vulkan API for AI inferencing opens it up immediately for Llama.cpp and other software with existing Vulkan back-ends rather than having yet another vendor specific interface. Raising my curiosity yesterday was this Vulkan-Docs merge request adding the Apex Compute author ID, vendor ID, and driver ID. In that Vulkan merge request they are explicitly clear on being open-source and leveraging Mesa: "We're building Apex GPU with an open-source Vulkan driver based on Mesa: https://github.com/apex-compute/mesa" Their Mesa branch can be found on GitHub. With the Apex Vulkan driver being in active development with many patches as recently as today. It's great as well seeing them develop in the open as well rather than just staged or irregular code dumps to their public repository. It will be interesting to watch as this Apex Vulkan driver develops and hopefully that they'll submit it to upstream Mesa once fully developed. Those curious about Apex Compute now knowing their open-source Mesa Vulkan driver activity can learn more about the start-up at ApexCompute.com.