COTT AX series
From inference services
to module-scale training.
Build the system around your workload: accelerator topology, host memory, storage, networking and the power environment.
4U AMD GPU server
An eight-GPU PCIe platform for accelerated workloads in a 4U rack footprint.
6U Intel GPU server
A direct CPU-to-GPU architecture for up to eight high-power PCIe accelerators.
AMD high-power GPU server
An AMD EPYC platform with direct PCIe GPU connectivity and wider-card configuration options.
AMD modular AI server
An eight-accelerator module platform for large AI training and high-bandwidth computing projects.
Specifications describe supported platform options, not installed components. Processors, accelerators, memory, drives, networking, software, supply timing and commercial terms are confirmed in the quotation. Photographs show illustrative configurations.
A complete configuration
Model size, precision, training or inference, accelerator type and the communication pattern between GPUs.
Working-set size, dataset capacity, storage throughput and the network needed between systems.
Rack dimensions, service clearance, facility power, cooling and the software stack.
Talk to us about technology projects, distribution and research collaboration.
May we use Google Analytics to understand visits and improve this website? No advertising tracking. Your choice does not affect access. Privacy details
