RAPID
Academic GPU Computing Infrastructure
A high-density computing platform built around NVIDIA A40 acceleration, AMD EPYC processors, high-capacity memory, and enterprise Linux — providing a powerful foundation for advanced academic computing.
64 NVIDIA A40 accelerators.
RAPID combines eight A40 GPUs in each physical compute server, creating a dense GPU platform with approximately 3 TB of aggregate GPU memory across the cluster.
NVIDIA A40
Ampere architecture data center acceleration with ECC-protected GPU memory and dedicated CUDA and Tensor processing resources.
Built for dense parallel compute.
Every RAPID node combines dual AMD EPYC processors, approximately half a terabyte of system memory, eight GPUs, and local NVMe storage.
CPU Compute
48 coresTwo AMD EPYC 7413 24-core processors per server, totaling 384 CPU cores across RAPID.
System Memory
~512 GBApproximately half a terabyte of ECC system memory per node, providing roughly 4 TB cluster-wide.
GPU Compute
8 × A40Eight NVIDIA A40 accelerators per node, with 48 GB nominal GPU memory each and 384 GB nominal GPU memory per server.
Local NVMe
~8 TBHigh-speed local NVMe storage provides node-local /scratch workspace for temporary compute data.
Eight servers. One GPU platform.
RAPID consists of eight high-density physical compute nodes, each equipped with eight NVIDIA A40 GPUs.
Mustang
8 × NVIDIA A40 · 48 CPU cores
Corvette
8 × NVIDIA A40 · 48 CPU cores
Rambler
8 × NVIDIA A40 · 48 CPU cores
Falcon
8 × NVIDIA A40 · 48 CPU cores
Nova
8 × NVIDIA A40 · 48 CPU cores
Impala
8 × NVIDIA A40 · 48 CPU cores
Edsel
8 × NVIDIA A40 · 48 CPU cores
Willys
8 × NVIDIA A40 · 48 CPU cores
Enterprise Linux at the foundation.
RAPID is being built as a standardized Rice Academic Computing Linux environment for GPU-intensive academic workloads.