A CHIP THAT ISN'T
SET IN STONE
Maverick’s unique runtime software algorithms learn your application and accelerate just the hot spots—out of the box.
Our technology is designed for seamless plug-and-play integration into HPC systems, with zero code modification and support for many application languages and frameworks, and no vendor lock-in.
The NextSilicon software suite intelligently adapts the Maverick processor into a workload-specific ASIC, at runtime. A novel architecture with distributed HBM memory access, accelerates workloads the others can't.
NextSilicon’s smart algorithms profile an entire application, identify the hotspots, and optimize the hardware autonomously, on the fly, and without manual intervention.
Iterative rounds of telemetry and optimization hone in on the main workload and ensure best-in-class acceleration, and optimal utilization of the available hardware.
We call the optimized grid projected onto our chip, a "mill core." A mill core that is optimized for a particular HPC application can run hundreds or thousands of threads or processes in parallel.
All this sits on a standard form factor PCIe card, with software algorithms working behind the scenes, ensuring you have a seamless plug and play experience.