# Intel names D-Series GPU for AI inference with up to 480GB memory

Digest AI · Hardware & Compute · published 2026-10-12T22:42:31Z

Canonical: https://digestai.news/story/intel-names-d-series-gpu-for-ai-inference-with-up-to-480gb-memory

## Summary

Intel has officially named its AI-focused **Crescent Island** GPU as the **D-Series**, targeting inference workloads in server environments. The card features **32 Xe3P cores**, **256 XMX engines**, and supports **up to 480GB of LPDDR5X memory**—with a default **160GB configuration** for Intel-branded cards. It delivers **1.3 PetaFLOPS of FP8 performance** and includes a media engine for multimodal tasks like video generation.

The Xe3P architecture introduces a **16-deep systolic pipeline** in XMX engines, supports **FP4 to FP64 data formats**, and lacks traditional graphics processing. Intel claims the card can run models with **up to 300 billion parameters** on a single unit, with an **8x cluster** handling trillion-parameter models. Sampling is underway, but pricing, availability, and benchmark results remain undisclosed. The card runs on **PCIe 5.0 x16** with **350W power draw**, requiring no liquid cooling.

## Key points

- Intel’s D-Series GPU supports **480GB LPDDR5X memory** (160GB standard for Intel-branded cards)
- Designed for AI inference, with **1.3 PetaFLOPS FP8 performance** and no graphics processing
- Supports models up to **300B parameters** on one card, **trillion-parameter clusters** with 8x GPUs

## Why it matters

The D-Series targets high-memory AI workloads, competing with NVIDIA’s H100/H200 in inference-heavy environments like large-language-model deployment. Its **480GB memory** and **FP4 support** may reduce costs for long-context models, though real-world adoption depends on pricing and software maturity.

## Sources

1. [Intel "Crescent Island" GPU Gets D-Series Name, Supports Up to 480 GB of LPDDR5X](https://techpowerup.com/353647/intel-crescent-island-gpu-gets-d-series-name-supports-up-to-480-gb-of-lpddr5x) (techpowerup.com, 2026-10-12)
2. [Intel's Crescent Island AI accelerator officially powers new GPU D-Series products](https://tomshardware.com/tech-industry/artificial-intelligence/intels-crescent-island-ai-accelerator-officially-powers-new-gpu-d-series-products-chip-delivers-1-3-peak-pflops-of-fp8-compute-and-1-5-tb-s-of-memory-bandwidth) (Tom's Hardware, 2026-10-12)

Part of the developing story: [Intel Unveils Giant Memory GPUs](https://digestai.news/thread/intel-gpu-d270-offers-up-to-480-gb-lpddr5x-memory) (2 stories)

## Cite

Digest AI, "Intel names D-Series GPU for AI inference with up to 480GB memory", 12 October 2026, https://digestai.news/story/intel-names-d-series-gpu-for-ai-inference-with-up-to-480gb-memory

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/intel-names-d-series-gpu-for-ai-inference-with-up-to-480gb-memory.json