- HPE reveals rack-scale system containing 81,920 CPU cores
- AMD Venice processors energy HPE’s next-generation Cray infrastructure
- One 42U rack delivers unprecedented ranges of computing density
Throughout its current HPE Uncover 2026 occasion, the corporate revealed new Cray GX5000 {hardware} that includes next-generation AMD EPYC Venice processors, with specs that push server density properly past present deployments.
The system combines a number of compute blades, liquid cooling infrastructure, networking {hardware}, and reminiscence assets inside a single 42U rack configuration.
HPE revealed a Cray GX5000 configuration designed to ship as much as 81,920 CPU cores in a single rack.
Dense compute structure pushes rack capability larger
The HPE Cray GX5000 platform follows the AMD EPYC 9965, a 192-core processor that represented one in all AMD’s highest-core-count server CPUs earlier than Venice arrived.
Whereas the EPYC 9965 elevated processor-level density, the Venice-based system takes a broader method by combining a number of CPUs, reminiscence assets, and cooling infrastructure inside a single rack.
On the heart of the system is the HPE Cray GX250a compute blade, which homes eight AMD EPYC Venice processors.
The compute blade incorporates energy supply, liquid cooling channels, reminiscence subsystems, storage units, and networking parts inside a compact design.
HPE acknowledged {that a} totally populated rack can ship 81,920 CPU cores, though precise processor configurations weren’t disclosed.
Primarily based on the rack specs, the system reportedly makes use of 80 multi-node motherboards and may assist as a lot as 1.28PB of RAM.
Every Venice processor connects to 16 reminiscence channels, creating substantial reminiscence bandwidth for large-scale computing workloads.
The reminiscence modules themselves are liquid-cooled and seem to make use of normal DIMM kind elements.
Pictures from the occasion confirmed native Samsung E1.S EDSSF SSDs mounted above a number of processor chilly plates.
HPE representatives indicated these drives function high-speed scratch storage for non permanent information processing duties.
The put in DRAM modules, storage units, and node identifiers recommend that the displayed {hardware} was operational reasonably than a nonfunctional demonstration unit.
That distinction is critical as a result of earlier Venice demonstrations appeared nearer to prototype programs than production-ready deployments.
Venice CPUs and networking outline the platform
The rack incorporates Slingshot 400 networking {hardware}, with HPE indicating future compatibility with Slingshot 800 know-how.
Networking modules are mounted inside facet pods linked to processors by devoted interfaces designed for high-bandwidth communication.
The front-facing networking association additionally simplifies cable administration by altering how optical connections are routed all through the rack.
HPE additionally displayed a coolant distribution unit able to dealing with 1.6MW of cooling capability for big installations.
Such cooling necessities replicate the rising energy densities related to fashionable high-performance computing infrastructure and more and more complicated CPU designs.
AI instruments, scientific simulations, engineering evaluation, and enormous LLM deployments are amongst workloads requiring this degree of computational density.
The corporate didn’t disclose detailed specs for AMD’s unannounced Venice processors, though obtainable figures recommend unusually excessive core counts.
Calculations primarily based on the acknowledged 81,920-core rack capability suggest processor densities exceeding present EPYC generations by a considerable margin.
Though AMD has not launched the specs or efficiency figures for Venice, the projected core density of the HPE system has led to hypothesis that the processor may turn out to be probably the most highly effective x86 CPUs produced.
Rather a lot may change earlier than the official launch, however the Cray GX5000 platform signifies that AMD and HPE are pursuing larger compute density throughout the identical rack footprint.
Through ServeTheHome
Comply with TechRadar on Google Information and add us as a most well-liked supply to get our skilled information, evaluations, and opinion in your feeds.
Leave a Reply