HPC Clusters
PRODUCTS
NumberSmasher™ and Navion™
Compute Clusters
Microway’s NumberSmasher and Navion clusters integrate the latest architectures and technologies to provide optimal performance for the lowest cost.
With decades of experience in HPC & AI, our design, production, and integration teams ensure that your cluster arrives fully functioning and on time. We’re good listeners and take your design criteria seriously.
Available with the following technologies:
NumberSmasher clusters are powered by leading-edge Intel Xeon Scalable Processors. With exceptional floating point, integer, and memory performance, Xeon CPUs are the building blocks of superior HPC.
Navion clusters feature AMD EPYC Server CPUs. Each Navion Compute Node provides exceptional throughput and class-leading memory bandwidth for leadership HPC performance.
lusters incorporating the latest NVIDIA H100, L40S, or RTX Professional GPUs, and NVIDIA Networking. Our NVIDIA GPU-accelerated clusters offer unparalleled throughput on pre-ported applications, GPU-enabled libraries, or custom CUDA code. Available in a variety of GPU:CPU ratios and with NVIDIA Networking.
Fully ready to run rack or facility scale AI deployments built upon the NVIDIA DGX Platform for scalable AI.
Flexible, Reliable, Cost-Effective HPC
Clusters are designed, integrated and installed following collaboration between Microway’s technicians and your designated IT personnel. For those without dedicated IT facilities, Microway is pleased to offer guidance on industry best-practices.
- Microway provides these fully-integrated clusters at very competitive prices. Users worldwide pushing the limits of technology in life sciences, universities, commercial and government research count on our expertise and attention to detail.
Key Technologies
We employ the fastest multi-core CPUs and multi-channel, error-correcting memory as a foundation for all of our high-performance systems. These guarantee the best performance for any application you or your users plan to execute.
- Additional acceleration is available through integrated NVIDIA Datacenter GPUs.
- Large-capacity, high-bandwidth and low-latency storage systems are available integrated into any of our clusters.
- Connections between systems enable the highest-bandwidth, lowest latency communication available through technologies including InfiniBand, High-bandwidth Ethernet (both from NVIDIA Networking) or 100Gbps Intel Omni-Path.
Turn-key Customization
Microway offers customized systems to fit any budget, and will gladly assemble a proposal to match your requirements. Your cluster’s design specifications and quotation will be prepared by our expert team. Depending on the complexity of your cluster, team members may include individuals with HPC programming, hardware design, technical sales, or HPC system administration expertise.
- All Microway systems are cluster-ready. Any server or workstation model which matches your requirements is available in a customized cluster. Groups of quiet WhisperStation workstations are available as small in-office compute clusters.
- The software and network configuration of your cluster may be specified by your group or prepared by our experienced integrators.
- Our proposals include all necessary specifications for your facilities personnel to verify suitable power and cooling. Those without centralized IT may speak with a technical account manager to ensure painless delivery and installation of their new compute resources.
Delivered Ready to Run
Microway clusters are integrated and thoroughly tested at our factory before shipment to your site.
- Each is shipped configured with the Linux distribution of your choice.
- Cluster management and monitoring tools are installed, configured and tested. Options include:
- Microway Cluster Management Software (MCMS™)
- OpenHPC community building blocks for HPC systems
- NVIDIA Bright Cluster Manager
- Open-source or commercially-supported workload manager, including SLURM, LSF, more
- MPI Link-Checker™ Fabric Validation verifies proper operation of all communication paths between systems.
- Cluster libraries and tools (such as BLAS, FFTW, MPI, Intel OpenAPI, NVIDIA HPC SDK) are compiled, custom-integrated and tested.
- User-level applications and libraries are available pre-installed.
- Intel, LLVM, GNU compilers are available pre-installed.
- Multiple options for delivery and installation, depending upon the budget and requirements:
- Factory pre-wiring of the rackmount cabinets
- Factory pre-installation for the server rack rails
- White-glove delivery of fully-integrated cluster racks
- Onsite installation & training services