Apex Compute

Stand: 1455
Apex Compute is building high-performance, energy-efficient compute architectures for edge AI and graphics. Its Unified Engine is designed to accelerate modern transformer workloads, including LLMs, vision-language models, and vision-language-action models, with a focus on low latency, deterministic execution, and high performance per watt. Apex has validated the architecture across multiple FPGA platforms running models such as Qwen, Llama, Gemma, and VLA workloads end to end, and is bringing the technology toward ASIC deployment while also offering it as licensable semiconductor IP.

In parallel, Apex has developed its own fully programmable GPU architecture and software stack, also running on FPGA. The GPU supports Vulkan 1.4, can boot and operate with both Windows and Linux, and has demonstrated a wide range of games and graphics workloads running on Apex hardware. Like the AI accelerator, the GPU architecture is available for IP licensing and can be integrated into custom SoCs and semiconductor platforms. 

Visit website (opens in a new tab)
Loading

Contact Exhibitor