SymbioticLab operates powerful in-house clusters dedicated for SymbioticLab’s full-stack research. We have two clusters specialized for memory and GPUs, respectively.
6 Memory Nodes
- 2 x Intel Xeon Processors
- 512 GB DRAM + 512 GB Intel Optane DIMM
- 1 x NVIDIA A40 GPU
6 GPU Nodes
- 2 x AMD EPYC Processors
- 512 GB DRAM
- 4 x NVIDIA A40 GPUs
In total, 30 NVIDIA A40 GPUs, 1440 CPU cores, and 9 TB memory are connected via a dedicated 200 Gbps Infiniband network. For reference, this dedicated facility is more powerful than continuously using eight AWS p3.8xlarge instances ($70,500/month), allowing us to discover many challenging problems before anyone else.