HPE Uncover This 12 months HPE Uncover 2024 is all about AI {hardware} developed by means of an expanded partnership with chip heavyweight Nvidia.
“Generative AI holds immense potential for enterprise transformation, however the complexities of fragmented AI know-how comprise too many dangers and limitations that hamper large-scale enterprise adoption,” mentioned HPE chief Antonio Neri. “HPE and Nvidia co-developed a turnkey personal cloud for AI that may allow enterprises to focus their assets.”
At its most elementary, Nvidia AI Compute by HPE consists of a trio of recent servers designed for a variety of wants, from LLM shoppers up by means of full service suppliers. Probably the most primary mannequin, the HPE ProLiant Compute DL384 Gen12, ships with an Nvidia GH200 NVL2, whereas the top-end HPE Cray XD670 ships with eight Nvidia H200 Tensor Core GPUs. The mid-line mannequin can maintain as much as eight Nvidia H200 NVL GPUs, and choose variations can be found with direct liquid cooling as nicely.
“By no means earlier than have Nvidia and HPE built-in our applied sciences so deeply,” Nvidia boss Jensen Huang gushed.
HPE additionally introduced that its GreenLake cloud platform had achieved Nvidia DGX BasePOD certification and OVX storage validation, and the corporate additionally shared information of recent GreenLake OpsRamp capabilities for AI infrastructure observability.
The large information: A personal Nvidia AI cloud from HPE
Slicing-edge AI {hardware} and a few new GreenLake validations are all nicely and good, however the large announcement Neri shared on the keynote was undoubtedly the brand new HPE Personal Cloud AI choices, which the corporate is advertising to prospects as having the ability to rise up and working “in three clicks.”
{Hardware} choices are included right here too. An HPE Personal Cloud AI setup consists of a number of {hardware} racks in 4 tiers: A small one for doing AI inferencing, a medium tier that provides retrieval-augmented era (RAG) capabilities, and a big and XL setup that provides numerous levels of machine studying fine-tuning choices.
Whatever the tiers, which differ largely in relation to the sort and variety of Nvidia chips packed into them, HPE Personal Cloud AI {hardware} is predicated fully on GreenLake Cloud with a personal management plan layered on high, HPE mentioned. Nvidia’s total AI enterprise suite can be included, as is its NIM inference microservices.
Relying on how prospects wish to set them up, the racks might be fully self-managed or dealt with by HPE. Every tier is ready to function as a standalone on-prem AI resolution or as a hybrid mannequin.
For the curious, HPE Personal Cloud AI racks include a minimal of 4 Nvidia L405 GPUs on the small finish, and as many as 24 Nvidia GH200 NVL2 GPUs in totally specced-out, extra-large configuration.
As for what HPE Personal Cloud provides over rivals, the enterprise biz mentioned it does not suppose it has any competitors in the case of successful prospects seeking to spend money on AI who wish to keep away from spending their total IT price range on streaming information to a cloud supplier for evaluation.
“We’ve got designed this to be a turnkey single click on,” HPE chief know-how officer Fidelma Russo mentioned throughout a press briefing final week, “so we don’t consider there are comparable choices within the market.”
In different phrases, HPE appears satisfied it is provide you with a product that is going to be good for corporates with loads of information that need to spend money on AI with out having to maneuver mentioned information out of their very own techniques.
“Enterprises around the globe are shifting from a public-cloud first strategy to a hybrid one,” Russo mentioned eventually 12 months’s HPE safety analyst assembly, suggesting HPE’s technique has been within the works for some time.
“They’ve come to the conclusion that as their information grows and turns into extra distributed in nature, it’s neither scalable nor economical to rely solely on the general public cloud.”
However trade consultants aren’t fully certain some companies will wish to tackle the opposite expense of managing their very own AI workloads: The chance that comes with shifting all that infrastructure on web site.
“HPE should use Uncover to construct a convincing story that enterprises ought to reduce threat by exerting full management over their AI initiatives,” Informa Fellow Steve Brazier instructed The Register.
“[HPE] additionally must persuade Nvidia to supply it with a enough provide of Blackwell chips,” Brazier added. In different phrases, as optimistic as HPE makes its future look, there’s nonetheless loads of work for the corporate to do to show it is a worthwhile different to all of the cloud-based AI options.
HPE plans to launch Personal Cloud AI this fall, with the identical timing for all of the {hardware} introduced at the moment as nicely, aside from the brand new Nvidia H200 NVL-powered Cray XD670, which is anticipated to be out there this summer season. ®
