Dell Applied sciences has introduced the PowerEdge XE8812, a brand new addition to the Dell AI Manufacturing facility with NVIDIA portfolio designed for high-performance computing (HPC) and synthetic intelligence (AI) workloads. Constructed on NVIDIA Vera Rubin NVL4 structure, the platform helps as much as 144 GPUs per rack and targets organizations working large-scale AI coaching, inference, engineering simulations, and scientific computing workloads.
The launch comes as Dell AI Manufacturing facility deployments proceed to increase globally throughout sovereign AI initiatives, engineering and design workflows, and genomic analysis.
Dell PowerEdge XE8812
The fanless, direct liquid-cooled PowerEdge XE8812 is designed for demanding HPC and AI functions, together with molecular and multi-physics simulations. The platform combines NVIDIA Vera Rubin NVL4 structure with NVIDIA CUDA-X libraries, permitting organizations to run bigger AI fashions and simulations fully in reminiscence.
In comparison with programs based mostly on NVIDIA GB200 NVL4 structure, the brand new platform delivers:
- Expanded host reminiscence
- CPU core depend elevated from 144 to 176 cores
- Greater GPU reminiscence capability
- Elevated compute efficiency
These upgrades are meant to help bigger workloads whereas decreasing dependence on information motion between storage and reminiscence.
Key Options
Excessive-density rack design
Dell says the PowerEdge XE8812 might be obtainable in an ORv3-style rack configuration designed for high-density AI and HPC deployments. Key specs embody:
- As much as 144 GPUs per rack
- Greater than 300kW energy capability
- 100% direct liquid cooling for CPUs and GPUs
- ORv3-based rack structure
The liquid-cooled design helps excessive compute density whereas enhancing cooling and energy effectivity.
Expanded reminiscence capability
The platform delivers 50% extra reminiscence per socket and better GPU reminiscence capability in comparison with the earlier era. This permits bigger AI fashions and simulations to stay totally in reminiscence, decreasing the necessity for staging information from storage or host reminiscence and minimizing swapping operations that may introduce latency and scale back obtainable bandwidth.
Open structure and system administration
Constructed on the Open Rack Model 3 (ORv3) customary, the platform is designed to help modular deployment and operational effectivity.
Administration instruments embody:
- Built-in Dell Distant Entry Controller (iDRAC)
- Dell Built-in Rack Controller
- OpenManage Enterprise
These instruments present distant administration, monitoring, telemetry, and automatic leak detection.
Rack-scale deployment
Dell PowerRack offers factory-integrated and pre-validated rack-scale programs for organizations deploying giant AI and HPC environments. Mixed with Dell ProDeploy providers, the answer is designed to cut back deployment complexity and shorten time to manufacturing.
The method replaces a lot of the guide integration historically required for large-scale deployments. In response to Dell, production-ready racks may be deployed and working workloads in simply over six hours.
Dell AI Manufacturing facility Deployments
Dell says greater than 5,000 prospects worldwide have deployed Dell AI Manufacturing facility options throughout AI, engineering, sovereign AI, and scientific workloads.
United States
Dell, NVIDIA, and NERSC are constructing the Doudna supercomputer at Lawrence Berkeley Nationwide Laboratory.
The system will characteristic:
- Dell PowerEdge XE8812 servers
- NVIDIA Vera Rubin NVL4 structure
- NVIDIA Quantum-X800 InfiniBand networking
It’s designed to help AI coaching, inference, HPC simulations, and data-intensive scientific analysis spanning fields from molecular science to astronomy.
France
Dell and NVIDIA are supporting AI firm InstaDeep because it expands its Kyber supercomputing cluster utilizing Dell AI Manufacturing facility infrastructure.
The cluster is anticipated to ship roughly 0.5 exaFLOPs of FP16 efficiency and might be used for:
- Massive-scale AI mannequin coaching
- Industrial design workloads
- Automated printed circuit board design
United Kingdom
The Wellcome Sanger Institute is utilizing Dell PowerEdge XE-Sequence servers with NVIDIA GPUs for large-scale genomic analysis.
The institute presently:
- Produces one totally assembled genome each seven hours
- Manages greater than 100 petabytes of curated genetic information on-premises
- Has contributed over 70% of genomes to the Earth BioGenome Venture
Australia
Monash College, in collaboration with Dell, NVIDIA, and CDC Information Centres, has deployed the MAVERIC supercomputer.
The system combines:
- Liquid-cooled Dell PowerRack infrastructure
- Dell PowerEdge XE9712 servers
- NVIDIA GB200 NVL72 structure
The system will help AI and data-intensive analysis in most cancers detection, local weather motion, and genomics.
Availability
Dell says the PowerEdge XE8812 might be available globally early subsequent 12 months.
Talking on the announcement, Venkat Sitaram, Senior Director and Nation Head, Infrastructure Options Group, Dell Applied sciences India, stated:
As India advances its AI ambitions and organizations more and more search to leverage AI and high-performance computing to drive innovation, the necessity for scalable, energy-efficient and future-ready infrastructure has by no means been higher. With the Dell AI Manufacturing facility with NVIDIA and the introduction of next-generation platforms such because the Dell PowerEdge XE8812, we’re enabling enterprises, analysis establishments and rising sovereign AI initiatives with the efficiency, density and open structure required to handle demanding AI and HPC workloads. Dell stays dedicated to serving to prospects in India translate AI ambition into real-world outcomes at scale.
Commenting on the announcement, Chris Marriott, Vice President, Enterprise Platforms, NVIDIA, stated:
The convergence of AI and HPC is reshaping what organizations can anticipate from their infrastructure. Dell and NVIDIA are advancing that benchmark collectively, combining NVIDIA Vera Rubin NVL4 structure and CUDA-X libraries with Dell’s engineering capabilities and large-scale deployment experience to ship the efficiency, effectivity and openness wanted for a few of the world’s most demanding AI and scientific computing workloads.