Hewlett Packard Enterprise said its next generation Cray supercomputing platform will offer a choice of processors from Nvidia and AMD, even though the chips aren’t available yet and the system is likely not going to be available until 2027.
HPE announced the Cray Supercomputing GX5000 ahead of the upcoming Supercomputing 25 conference, taking place in St. Louis Nov.16-21. The GX5000 succeeds the EX4000as HPE’s top supercomputer.
The Cray Supercomputing GX5000 racks can handle up to 400 kilowatts of equipment, with the potential to go up to 1 megawatt in the future, HPE says. Each supercomputer holds up to 44 of the rack units while each rack unit can hold between 28 and 40 blades, depending on the configuration.
HPE offers three different blades for the GX5000. First is the GX440n Accelerated Blade for mixed-precision computing, with four Nvidia Vera CPUs and eight Nvidia Rubin GPUs. Up to 24 of these blades can be configured per compute rack, which translates to 192 Rubin GPUs per rack. Vera Rubin is due to ship late next year.
Second is the GX350a Accelerated Blade, designed for customers who desire a universal compute engine for mixed-precision computing with AMD CPUs and GPUs. The blade includes one next generation “Venice” CPU and four AMD Instinct MI430X GPUs. Up to 28 of these blades can be configured per compute rack, providing up to 112 AMD MI430X GPUs per rack.
Finally, there is the GX250 Compute Blade for customers who want a CPU-only partition for double-precision workloads with eight next-generation Venice CPUs per rack. Up to 40 of those blades can be configured per compute rack, delivering industry-leading flagship x86 CPU core density.
Networking for the systems will be the 400 Gbps version of HPE’s Slingshot interconnect. All three blades can be configured with either four or eight Slingshot ports per blade via a liquid-cooled PCIe network card. There are also a 64-port liquid-cooled Slingshot 400 switch blade and a blade chassis that can fit eight, 16, or 32 of these blades.
HPE also announced an updated version of Supercomputing Management Software to enable customers to deploy converged AI and HPC workloads on its Cray hardware. It adds multi-tenant, virtualized, and containerized environments and provides system-wide power management.