U.S. Department of Energy

Pacific Northwest National Laboratory



Testbeds of future HPC technologies that enable and accelerate computational science at various levels of technology maturation and integration.

, Task Lead

CENATE’s testbeds cover a time horizon from early concepts to pre-release systems. Testbeds consider the subsystem space to enable the exploration of novel developments for application developers in processing, memory, networking, and input/output and storage technologies. These aspect are important in:

  • Determining mainstream and novel technologies at the state of the art.
  • Targeting subsystems that are key building blocks of future HPC system architectures, such as processors, accelerators, memory, storage, and/or software stacks.
  • Combining subsystems to ascertain complementary effects.
  • Providing technology access to the DOE community via visitor spaces and remote-access mechanisms.  
Available CENATE Testbeds
Penguin Computing SeaPearl cluster with 1-KHz power and thermal instrumentation that collects measurements at runtime.
Micron Technology Hybrid Memory Cube with ultra-high bandwidth dynamic random-access memory (DRAM) that uses less energy per bit than other DRAM technologies.
NVIDIA DGX-1 featuring eight-socket Pascal with stacked memory and software stack that includes major deep learning frameworks.

Power8 with Contutto, a non-volatile random-access memory (NVRAM) featuring spin torque transfer magnetic memory with memory controller logic, implemented in a field-programmable gate array.

Data Vortex Jolt, a high-probability, congestion-free network for fine-grained communications.
| Pacific Northwest National Laboratory