Our hardware platform consists of an internal cluster of several hundred GPUs, integration of on-site and cloud resources from multiple vendors, and research into next generation experimental FPGAs.
High Performance Computing
Our current high-performance computing platform is comprised of both internal and cloud resources. Most of our drug discovery workloads run on Neo, a 300 GPU cluster housed in four racks at MGHPCC. Neo’s first-generation nodes have Intel Xeon processors for a total of 16 cores and 64 GB of memory. The more recent nodes have Intel Skylake processors, with 16 cores and 96 GB of memory. Altogether, the system has almost 3 petaflops of peak single-precision power from almost a million Nvidia GPU cores for running CUDA and OpenCL jobs. Neo runs Ubuntu OS and utilizes Slurm for job scheduling. We are also capable of utilizing multiple cloud vendors including Microsoft, Google and AWS during special projects or for burst simulations.
For our next generation platform, we are increasingly researching state-of-the art FPGAs for some of our problems. We have a FPGA node in the office with both Intel and Xilinx FPGAs and are actively working on running molecular dynamics simulations on FPGAs in the near future.
About the Project Team
Silicon Therapeutics employs an exceptional team of outstanding life scientists and computer scientists with a unifying belief that we can make better medicines faster and cheaper using computers.