Intel Launches D-Series GPUs and Xeon 6+ Racks for Agentic AI Deployment
At the Open Compute Project Global Summit, Intel revealed new technical specifications for the Intel GPU D-series alongside high-density server rack configurations designed for agentic AI workloads, as reported by intc.com. The hardware rollout addresses a growing infrastructural bottleneck: modern artificial intelligence pipelines require intense general-purpose compute to orchestrate multi-step reasoning, retrieval, and tool execution before handing tasks off to accelerators. By pairing specialized graphics hardware with dense central processing units, the company aims to improve performance efficiency across large language model deployments. Held in Santa Clara, California, the event featured the corporation sharing new technical details on the graphics hardware, introducing new Intel Xeon-based rack designs, and highlighting open software and ecosystem innovations.
Intel GPU D-Series Architecture and Memory Specs
Codenamed Crescent Island, the Intel GPU D-series uses the Xe3P architecture to target memory-intensive inference tasks. The hardware pairs up to 480 GB of LPDDR5X memory with 1.5 terabytes per second of memory bandwidth and 1.3 petaflops of FP8 compute power, all operating inside a 350-watt power envelope. This configuration supports single-card execution for models containing up to 300 billion parameters. Multi-card setups linking eight GPUs scale that capacity to handle models exceeding one trillion parameters. The standard PCIe form factor and air-cooled design allow operators to integrate the hardware into existing data center environments without requiring specialized liquid cooling infrastructure. Anil Nanduri, vice president of AI Products and GTM in the Intel Data Center Group, stated, With Intel GPU D-series, we’re focused on making high-capacity inference practical to deploy and economical to scale.
Xeon 6+ Rack Configurations and Core Densities
Agentic AI workflows demand robust general-purpose compute to manage data movement, input-output operations, and workload orchestration. To meet these requirements, select Intel Xeon 6+ configurations scale beyond 46,000 cores per rack while consuming 133.45 kilowatts of total rack power. According to intc.com, these systems leverage the Open Compute Project Open Rack v3 standard to simplify deployment within modern enterprise facilities. Nicolas Dubé, corporate vice president of the Data Center Systems and Solutions Group at Intel, stated that organizations require infrastructure capable of matching varied workloads with appropriate compute resources and economics, adding, That’s what we mean by ‘put every prompt in its place.’
Anil Nanduri, vice president of AI Products and GTM in the Intel Data Center Group, noted that inference is rapidly becoming the defining enterprise AI workload.
Frequently Asked Questions
What is the codename for the Intel GPU D-series?
The hardware was developed under the codename Crescent Island.
How much memory does a single Intel GPU D-series card support?
A single card features up to 480 GB of LPDDR5X memory paired with 1.5 TB/s of memory bandwidth.
Which power envelope does the Intel GPU D-series operate within?
The architecture functions within a 350-watt power envelope while delivering 1.3 petaflops of FP8 compute.
What maximum core count do the select Xeon 6+ rack configurations reach?
Select configurations scale beyond 46,000 cores per rack at a total power draw of 133.45 kilowatts.
Keep reading