Intel GPU D-Series Targets AI Inference With Record 480GB Memory in Air-Cooled Design
10 Sources
[1]
Intel's 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
Crescent Island gets official performance and bandwidth specs at OCP 2026. At OCP 2026 today, Intel has given Crescent Island a proper name: the Intel GPU D-Series. For the first time, the company also shared a theoretical peak performance target for the chip: 1.3 PFLOPS of FP8. It also disclosed
[2]
'Put every prompt in its place': Intel launches 480GB GPU D-series for AI inference
* Intel's GPU D-series sets a new GPU memory record with 480GB of LPDDR5X * Intel chose LPDDR5X instead of GDDR7 for its high-capacity AI accelerator * The D-series supports models containing up to 300 billion parameters At the recent Open Compute Project Global Summit, Intel unveiled a graphics
[3]
Intel GPU D-series AI inference chip specs revealed at OCP 2026
The chip, formerly code-named Crescent Island, delivers 1.3 petaflops of FP8 compute and 480GB of memory within a 350-watt power envelope Intel $INTC shared technical details Monday on its GPU D-series AI inference chip at the Open Compute Project Global Summit, giving the product formerly
[4]
5. Intel GPU D-Series Enters AI Inference With 480GB LPDDR5X
Intel's GPU D-Series, formerly known as Crescent Island, is designed for AI inference and massively parallel computing. Intel specifies up to 1.3 PFLOPS of FP8 performance, 1.5TB/s of memory bandwidth and up to 480GB of LPDDR5X memory. The accelerator uses Intel's Xe3P architecture but is not a
[5]
Intel GPU D270 Targets 300B-Parameter Models With 480GB of LPDDR5X
Intel has officially announced the GPU D270, previously associated with the Crescent Island development name. The data-center accelerator is designed for agentic AI inference and large-memory workloads, with a maximum of 480GB of LPDDR5X memory per card. Intel's Xe3P architecture provides 32 Xe
[6]
Intel Squeezes Two Crescent Island GPUs Onto One Liquid-Cooled OAM Card, And First Xe3P Die Shots Show A Chip Smaller Than Big Battlemage
Intel's upcoming Crescent Island "D-Series" GPUs have been shown off in a dual-package on a liquid-cooled OAM card. Intel D-Series GPUs To Pack Up To 512 GB Memory, First OAM Designs Revealed While Xe3P GPUs Pictured In Single & Dual Packages Just a few days after its unveiling, the Intel
[7]
Intel Crescent Island Squeezes 1.3 PFLOPs Of FP8 Into Just 350W As D270 Targets The Competition With Value-Oriented LPDDR5X Up To 480 GB
Intel is announcing its D-Series GPUs, formerly Crescent Island, as the next AI data center lineup based on the Xe3P architecture. Intel D-Series "Crescent Island" GPUs Pack 1.3 PFLOPs of FP8 at 350W, Rivaling Existing Pro Cards From NVIDIA & AMD The Intel Crescent Island GPU is based on the
[8]
Intel unveils GPU D-series for AI inference at OCP summit By Investing.com
SANTA CLARA, Calif. - Intel Corporation (NASDAQ:INTC) announced technical details of its GPU D-series, formerly code-named Crescent Island, and introduced new Xeon-based rack designs at the Open Compute Project Global Summit on Monday. The GPU D-series is built on Intel's Xe3P architecture and
[9]
Intel Corporation Shares New Technical Details On Intel GPU D-Series And Introduces New Intel Xeon-Based Rack Designs
Intel corporation shared new technical details on the Intel GPU D-series, formerly code-named Crescent Island, introduced new Intel Xeon-based rack designs and highlighted open software and ecosystem innovations. At OCP, Intel showcased new Xeon-based rack designs developed with ecosystem partners
[10]
Intel Crescent Island explained: Specs and architecture of the new GPU D-series
Everyone who works in AI looks for the most powerful chip. Intel has now unveiled a card that wants to claim the title of being the largest card out there. At the Open Compute Project Global Summit, Intel provided some more information on their new GPU called Crescent Island. They have given it a
Share
Copy Link
Intel officially launched its GPU D-Series AI accelerator at OCP 2026, featuring 480GB of LPDDR5X memory and 1.3 PFLOPS of FP8 compute in a 350W air-cooled design. The chip targets large language models and concurrent inference services, with customer sampling already underway.
Intel GPU D-Series Debuts With Industry-Leading Memory Capacity
Intel officially unveiled its GPU D-Series AI accelerator at the Open Compute Project Global Summit, revealing performance specifications for the chip previously known by its Crescent Island codename
1
2
. The data center GPU delivers up to 1.3 PFLOPS of FP8 compute performance alongside 1.5 TB/s memory bandwidth and up to 480GB of LPDDR5X memory within a 350W TDP envelope3
. The GPU D-Series now holds the record for highest memory capacity among data center GPUs, surpassing Nvidia's Blackwell Ultra B300 which offers 288GB of HBM3e memory2
.
Source: Wccftech
Strategic Design Choices for AI Inference Workloads
Intel built the GPU D-Series on its Xe3P architecture, which represents a massively parallel computing platform rather than a traditional graphics processor. The chip lacks most graphics-specific functional units, prioritizing matrix and parallel-compute resources instead
1
4
. Intel chose 480GB LPDDR5X memory over GDDR7, a decision that prioritizes memory capacity and power efficiency for AI inference rather than conventional gaming performance2
. The large LPDDR5X capacity keeps model weights close to compute engines, reducing repeated transfers across host interconnects4
.The AI accelerator uses a standard PCIe form factor with air-cooled design, allowing deployment in existing data center environments without specialized rack or cooling infrastructure
3
. This contrasts sharply with Nvidia's Vera Rubin NVL72 rack-scale system, where each of the 72 Rubin GPUs delivers 17.5 PFLOPS of FP8 throughput but requires liquid cooling and draws power measured in kilowatts1
.
Source: Tom's Hardware
Targeting Trillion-Parameter Models and Enterprise Deployments
Intel GPU D-Series products are designed for deployment in traditional air-cooled servers scaling up to eight GPUs, similar to Nvidia's HGX systems. A single card can support large language models with up to 300 billion parameters, while an eight-GPU system can accommodate models exceeding one trillion parameters
2
3
. Intel highlights the chip's suitability for large language models, long-context inference, and concurrent inference services, creating infrastructure for multi-tenant AI services, robotics, and latency-sensitive workloads1
."Inference is becoming the defining AI workload, and enterprises need infrastructure that delivers the right performance at the right cost," said Anil Nanduri, vice president of AI Products and GTM at Intel's Data Center Group. "With Intel GPU D-Series, we're focused on making high-capacity inference practical to deploy and economical to scale"
3
. The 480GB capacity addresses AI systems that must keep substantial model weights and associated data accessible during inference workloads2
.Related Stories
Software Stack and Ecosystem Challenges
For distributed inference deployments where multiple GPUs act as one unit, Intel emphasizes its common Xe software foundation built for AI across Xe Cores in integrated GPUs and discrete GPU products
1
. The company promotes support for OpenVINO and oneAPI frameworks4
. However, the ecosystem remains a significant challenge as customers compare Intel against Nvidia's mature CUDA platform and AMD's expanding ROCm stack4
.Intel is working with a broad ecosystem of OEMs, ODMs, neoclouds, independent software vendors, appliance providers, and AI-native companies to bring Xe accelerators into data centers
1
. Customer sampling of GPU D-Series products is currently underway, though Intel has not announced a broad launch date or pricing details3
5
.Market Position and Future Implications
The Intel GPU D-Series represents another attempt to establish an AI accelerator foothold after the failure of the Data Center GPU Max family (code-named Ponte Vecchio) and its mothballed follow-ons, as well as a series of silicon acquisitions that failed to coalesce into a coherent AI strategy
1
. In a market hungry for compute cycles for inference workloads, a product with proven architectural shape targeted specifically at inference in an accessible thermal and electrical envelope might provide the traction Intel needs1
.
Source: Wccftech
The 1.3 PFLOPS FP8 figure represents theoretical peak performance and should not be compared directly with gaming GPU performance or treated as an application benchmark. Actual throughput depends on model architecture, quantization, software kernels, and memory access patterns
4
5
. Independent testing will need to evaluate real model throughput, latency, scaling across eight cards, and software compatibility to determine whether the GPU D-Series becomes a meaningful alternative to Nvidia and AMD solutions5
. As AI models advance in intelligence and agentic workflows assume greater importance, infrastructure that balances capacity, efficiency, and package complexity will become increasingly critical for enterprise deployments.References
Summarized by
Navi
Related Stories
Intel Crescent Island AI Accelerator Packs 480GB Memory, 32 Xe3P Cores for Inference Workloads
25 Aug 2026•Technology
Intel Crescent Island targets AI inference with cheaper memory, air-cooling to challenge Nvidia
01 Jun 2026•Technology
Intel Unveils Crescent Island: A New AI-Focused GPU for Data Centers
14 Oct 2025•Technology
Recent Highlights
1
Google Unveils Gemini 4 Carbon AI Model With Major Cybersecurity Gains, Beating Claude Opus 5.5
Technology
2
Binance Hack: How AI-Enabled Cyberattack Led to Record $6.3 Billion Crypto Theft
Technology
3
Pope Leo XIV Demands AI Disarmament and Ban on Autonomous Weapons at UN Geneva Meeting
Policy and Regulation
Recent Highlights
Today's Top Stories
Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Follow topics that matter to you and stay ahead.
