Science & Tech · Gadgets & EVs · 1 hr ago
Intel unveils AI GPU with up to 480 GB of memory
Intel has announced its GPU D-Series for running AI systems, with a focus on inference—the process of using trained AI models.
The chip is aimed at data centers and applications such as multi-tenant AI services and robotics.
It can offer up to 480 GB of LPDDR5X memory, 1.5 terabytes per second of memory bandwidth and 1.3 petaflops of FP8 computing within a 350-watt power limit.
This memory is a different type from the high-bandwidth memory used in Intel's Gaudi 3 chip, so the capacities are not directly comparable.
Intel is targeting inference workloads as it tries a new approach after its Gaudi GPUs struggled to compete with Nvidia and its planned Falcon Shores GPUs were cancelled.
The large memory capacity highlights AI's demand for computing resources, which could add to concerns about limited memory for consumer devices; Intel has not said when the GPU will be available.
Intel unveiled its GPU D-Series, an AI-focused processor with up to 480 GB of LPDDR5X memory.
The GPU is built on Intel’s Xe3P architecture and offers up to 1.5 TB/s of memory bandwidth and 1.3 petaflops of FP8 compute.
Intel says the chip is designed for inference workloads, including multi-tenant AI services, robotics and other latency-sensitive applications.
The GPU has a 350-watt power envelope and is designed for deployment in existing data center environments.
Unlike Intel’s previous Gaudi 3 chip, which has 128 GB of unified HBM, the new GPU uses LPDDR5X memory.
- Who
- Intel unveiled the GPU D-Series.
- What
- It is an AI-focused GPU with up to 480 GB of LPDDR5X memory.
- When
- The announcement was reported on October 13, 2026.
- Where
- Not stated.
- Why
- Intel says it is designed to make high-capacity AI inference practical to deploy and economical to scale.
This story does not have two clearly opposing sides.
Intel’s Gaudi 3 chip featured 128 GB of unified HBM.
Intel cancelled its follow-up Falcon Shores GPUs.
Intel’s GPU D-Series, previously codenamed Crescent Island, was reported as unveiled.
- Memory
- Up to 480 GB of LPDDR5X
- Architecture
- Intel Xe3P
- Memory bandwidth
- 1.5 TB/s
- FP8 compute
- 1.3 petaflops
- Power envelope
- 350 watts
- Earlier Intel chip
- Gaudi 3 had 128 GB of unified HBM
What people are saying
Inference is becoming the defining AI workload, and enterprises need infrastructure that delivers the right performance at the right cost. With Intel GPU D-series, we're focused on making high-capacity inference practical to deploy and economical to scale.
Built on Intel's latest Xe3P architecture, the GPU combines up to 480 GB of LPDDR5X memory, 1.5 TB/s of memory bandwidth and 1.3 petaflops of FP8 compute within a 350-watt power envelope.
designed to simplify deployment within existing data center environments while supporting application portability across Intel GPU platforms.
Sources
Related news
Google launches Fitbit Edge with GPS, health tracking and replaceable battery
Fiat unveils concept that switches from six-seat car to pickup
Android users find a way to restore Google Assistant for now
HP's EliteBoard G2a puts an AI PC inside a keyboard, starting at $3,028
Oppo Find X10 series to launch in India on October 22
Lectric launches $1,099 XPress Lite commuter e-bike
Gigabyte removes post hinting at new Intel LGA1700 chips in early 2027
Nothing OS 5.0 with Android 17 starts rolling out to Phone (3)
Citroën unveils electric 2CV concept, with production planned for 2028
Tesla renames its European Full Self-Driving feature Assisted Driving
Google teases Fitbit Edge tracker ahead of October 12 announcement
