Dell Technologies Inc. is marking the one-year anniversary of its launch of the Dell AI Factory with Nvidia with new PCs, infrastructure, software and services aimed at enterprise artificial intelligence development. Dell said more than 2,000 customers are now using the platform.

The company is introducing a top-of-the-line server high-performance computing and expanding its Pro Max AI PC portfolio with new developer workstations, notebooks and desktops. It’s also broadly expanding the Dell AI Data Platform with Nvidia, a co-developed package of products and services tailored for AI workloads.

The new PowerEdge XE8712 liquid-cooled server (pictured) is aimed at high-performance computing applications and uses Nvidia’s GB200 NVL4 platform, which was introduced last November. It combines two Nvidia Grace CPUs with four Blackwell B200 GPUs on a single motherboard with 1.3 terabytes of coherent memory. The server supports up to 144 Nvidia B200 GPUs per rack.

Liquid cooling allows up to 264 kilowatts of power to be supplied to each rack. The unit’s modular integration with Dell Integrated Rack 7000 infrastructure features disaggregated power shelves and a shared power bus bar capable of 480 kilowatts. It supports up to 36 notes per rack.

“This separates the PowerEdge offerings from other dense designs,” said Varun Chhabra, senior vice president of infrastructure and telecom marketing. “A core part of the value we’re delivering is not just in the server itself, but also the rack-scale view we’ve taken when designing these systems.”

Desktops for developers

The new Dell Pro Max with GB10 features the Nvidia GB10 Grace Blackwell Superchip based on the Nvidia Blackwell architecture. Dell said it delivers up to one petaflop of performance and 128GB of unified memory.

The new Dell Pro Max with GB300 is a top-of-the-line deskside system featuring the new Nvidia GB300 Grace Blackwell Ultra Desktop Superchip and sporting up to 20 petaflops of performance, 784 gigabytes of unified system memory that can be expanded to 288 gigabytes of HBME3e graphics processing unit memory and 496 gigabytes of Low Power Double Data Rate 5X CPU memory. It supports Nvidia’s ConnectX-8 SuperNIC networking interface and can be used to train up to 460-billion parameter models.

“Both of these are purpose-built for the AI developer,” said Kevin Terwilliger, general manager of Dell’s PC business. “Previously, that was exclusively for the data center, but now we’re bringing it deskside.”

New Dell Pro Max notebooks and desktops come equipped with the Nvidia RTX PRO  Blackwell Generation GPUs and Intel Corp. Core Ultra (Series 2) or Advanced Micro Devices Inc. Ryzen CPUs and AMD Threadripper processors. “These will be our most powerful PCs yet, up to 36% faster than the previous generation,” Terwilliger said.

Also on the server front, Dell said its PowerEdge servers will support the Nvidia Blackwell Ultra platform, including the upcoming Nvidia HGX B300 NVL16, Nvidia GB300 NVL72 and Nvidia RTX  6000 Blackwell Server Edition. ​The HGX B300 NVL16 is designed accelerate AI workloads with 11 times faster inference on large language models, seven times more computational power and four times larger memory capacity than previous models.

The GB300 NVL72 is a 72-GPU scalable computing unit built for trillion-parameter AI model training.  The RTX PRO 6000 Blackwell Server Edition is a high-performance GPU designed for data centers, and visual computing workloads. It features 96 gigabytes of GDDR7 memory and supports multi-instance GPU technology, which allows a single GPU to be partitioned into up to four isolated instances. ​

The forthcoming Dell servers will offer peak AI cluster performance with 800 gigabits per second with ConnectX-8 SuperNICs.

Speedier deployment

The new companies are also announcing a jointly developed deployment platform called the Dell AI Data Platform with Nvidia. It allows the deployment of AI applications securely through always-on, direct access to structured, semi-structured and unstructured data. Features are aimed at accelerating AI reasoning workloads using Nvidia’s computing and networking technologies for near real-time data processing.

They include agentic reasoning to transform diverse data types through always-on processing, faster data ingestion and reduced processing redundancy with Nvidia’s NeMo open-source framework for building, training, and deploying large-scale AI models and the company’s NIM suite of accelerated inference microservices for foundation models deployment

“We’ve optimized throughput so that the speed and throughput of the data platform is not the bottleneck for these data hungry GPUs,” said Chhabra said.

Data protection features include encryption, role-based controls, proactive threat detection and immutable backups. It support the latest retrieval-augmented generation connectors from Nvidia for the Langchain composable large language model framework and NIM microservices.

Faster data ingestion

Expanded certifications for PowerScale storage, now support deployments of all sizes. Enhancements now allow PowerScale to improve GPU utilization by delivering 220% faster data ingestion and 99% quicker data retrieval than previous systems. The AI Data Platform now supports Apache Spark and integrates with the AI Data Platform reference design.

Dell PowerScale storage is now validated for both the Nvidia Cloud Partner Program and the new Nvidia-Certified Storage designation. Agentic AI support is enhanced with Nvidia’s AI-Q Blueprint and AgentIQ Toolkit. The ​AI-Q Blueprint is a framework for developing AI agent systems capable of autonomous reasoning and action. It blueprint facilitates the integration of AI agents with enterprise data. ​The AgentIQ Toolkit is an open-source library for connecting, evaluating and optimize teams of AI agents. New Dell Accelerator Services for RAG  help optimize the use of data specific to customer environments with agents.

Dell Pro Max desktops and laptops will be available beginning this month.

Photo: Dell