HPE has confirmed support for upcoming NVIDIA Rubin PCIe graphics processors in its ProLiant DL585a Gen13 server. The revelation surfaced through presentation slides shared with The Next Platform. Specific hardware details regarding how NVIDIA plans to adapt the power hungry Rubin architecture for standard PCIe expansion slots remain unannounced.
The HPE ProLiant DL585a Gen13 uses a massive 10U rack chassis built for enterprise computing. It runs on 2 AMD EPYC 9006 processors and accommodates up to 8 double width or 16 single width expansion cards. Storage connectivity includes 16 EDSFF E3.S drive bays for high speed solid state storage drives.
Fitting the Rubin architecture into a standard PCIe slot presents serious engineering challenges. Existing data center Rubin hardware uses an SMX form factor drawing up to 2300W of power. Standard PCIe add in cards cannot handle that level of power draw or heat output. If NVIDIA uses this exact silicon, engineers must reduce the core count by at least 50% and dial back clock speeds significantly.
Another option is that this hardware belongs to the RTX PRO family. Industry reports suggest that card could use the GR202 die, which shares structural origins with next generation GeForce gaming hardware. Regardless of the silicon choice, HPE slide disclosures make clear that enterprise customers will have a PCIe path to Rubin generation acceleration.
