Intel introduced the D-Series GPU, codenamed Crescent Island, featuring up to 480 gigabytes of LPDDR5X memory to handle heavy artificial intelligence inference workloads in data centers, Media Indonesia reported on October 19, 2026. The announcement highlights a massive leap in memory capacity designed to help data centers run large artificial intelligence models without splitting them across multiple cards.
Open Compute Project Global Summit Reveal
Intel unveiled the hardware details at the Open Compute Project Global Summit. The company positioned the graphics processing unit as an inference accelerator rather than a training device. Telset.id reported that the chip is designed to run previously trained AI models efficiently. Anil Nanduri, vice president of AI Products and GTM in the Intel Data Center Group, stated that inference is becoming the defining AI workload. Enterprises need infrastructure that delivers the right performance at the right cost, Nanduri said. With Intel GPU D-series, we’re focused on making high-capacity inference practical to deploy and economical to scale,
Nanduri added.
Technical Specifications of the D270 Accelerator
The primary variant introduced is the Intel D270, built on the Xe3P architecture with 32 Xe3P cores and 256 XMX engines. Journalarta.com noted that the processor comes standard with 160 gigabytes of LPDDR5X memory. Original device manufacturers can expand that capacity to 480 gigabytes. The hardware delivers a memory bandwidth of 1.5 terabytes per second and provides up to 1.3 petaflops of FP8 performance with a thermal design power of 350 watts over a standard PCIe 5.0 interface. Intel dropped 3D and ray tracing hardware to dedicate more silicon area to artificial intelligence computation while retaining FP64 support for high-precision scientific work.
| Feature | Specification |
|---|---|
| Architecture | Xe3P (32 Cores, 256 XMX Engines) |
| Maximum Memory | 480 GB LPDDR5X |
| Memory Bandwidth | 1.5 TB/s |
| FP8 Performance | 1.3 Petaflops |
| Thermal Design Power | 350 W |
| Interface | PCIe 5.0 |
Model Capacity and Comparison With Nvidia Hardware
A single D270 card supports language models containing up to 300 billion parameters. Systems equipped with eight of these GPUs can handle models exceeding one trillion parameters. The 480-gigabyte memory footprint surpasses the 288 gigabytes found on the Nvidia Blackwell Ultra B300, which relies on HBM3e memory. Intel chose LPDDR5X over more expensive high-bandwidth memory variants to prioritize storage capacity and power efficiency during long prompt processing and multi-client enterprise services.
Data Center Integration Plans
Intel designed the D-Series to fit into existing data center racks using standard air cooling instead of specialized liquid cooling setups. Nicolas Dubé, corporate vice president of Data Center Systems and Solutions Group at Intel, noted that open models are expanding deployment options for organizations. To realize their full potential, organizations need infr,
Dubé said. The hardware connects directly with Xeon processors to handle retrieval, reasoning, and task orchestration. Pricing and a general availability date have not yet been announced.
