Solutions · High Performance Computing
Gigabyte multi-node, many-core platforms for scientific computing, financial modeling, rendering and research clusters, with low-latency interconnects and high memory bandwidth.
InfiniBand NDRRoCE v2HBM memoryMulti-nodeParallel computing
Market landscape
In HPC, time to result depends on the slowest link: processor, memory or the network between nodes.
Simulations and numerical models hit the memory limit before using the full processor.
Parallel computing exchanges data all the time. Every microsecond of network latency is multiplied by the number of nodes.
Many cores per U concentrate heat and require a chassis and cooling sized for continuous load.
Clusters run 24 hours a day. Energy per calculation weighs on PUE and operating costs.
Engineering highlights
Processing
High core-count platforms for raw parallel computing.
Memory
Support for high-bandwidth memory for memory-bound workloads.
Interconnect
Low-latency networking for communication between cluster nodes.
Ethernet
RDMA over Ethernet for clusters running on standard network infrastructure.
Interconnect bus
High-speed buses for network cards, accelerators and storage.
Topology
Multiple nodes per chassis to scale the cluster using less rack space.
Recommended hardware
Multi-node, many-core platforms optimized for parallel computing.
For channels and resellers
Strategic partnership with Giga Computing and access to Gigabyte high-performance computing platforms.
Technical support for cluster sizing: nodes, memory, interconnect and cooling matched to the workload.
Stock availability and invoicing through the distributor to speed up the reseller's project closing.
Project quote
Describe your scenario and a technical specialist will get back to you with the recommended configuration.
