Server offers high performance for enterprise-level AI

The release of the new G4760 4U 10xGPU server, from Equus Compute Solutions has been announced. It is well suited for artificial intelligence and deep learning environments and offers higher performance, reduced rack space requirements, and lower power consumption compared with traditional CPU-centric platforms, in a server package ideal for enterprise-level workloads.

The innovative design of the G4760 features a dense 10GPU single root platform where each GPU is connected to the same CPU using PCIe switches to handle the connectivity. Equus is an NVIDIA Preferred Partner and the clever server physical layout allows up to 14 2.5” hot-swap SSD/HDD drive bays with support for 10 + 4 NVMe, up to 3TB memory, (3+1) 4,800W RPSU, and a solid mechanical chassis. This GPU platform provides the power and flexibility for high performance AI applications.

All Equus GPU servers have the advantage of flexible GPU sourcing options; install your own GPUs, consign GPUs to Equus for factory installation, or purchase turnkey GPU server systems.

“Our customers have been asking for the flexibility to source GPUs in different ways on high performance servers,” said Lee Abrahamson, CTO of Equus Compute Solutions. “Our GPU servers, such as the G4760 server, are the ideal cost-optimized solutions for a wide range of applications and workloads. At the same time, these innovative platforms provide benefits of scale and volume, component standardization, ease of service logistics, and the means to avoid vendor lock-in.”

The G4760 4U 10xGPU server uses Intel Xeon Scalable CPUs and NVIDIA Tesla GPUs, and features a wide range of SSD/HDD hot-swap storage options, OpenBMC readiness for open systems management, and Redfish support.