Intel unveils 480GB D-series GPU for AI workloads

Intel has announced its new D-series GPU, setting a memory record with 480GB of LPDDR5X. This model, previously known as Crescent Island, is specifically designed to meet the demands of artificial intelligence inference, meaning it can handle complex models with up to 300 billion parameters.

Intel's D-series has surpassed Nvidia's Blackwell Ultra B300 GPU, which featured 288GB of HBM3e memory, thus positioning itself as the data center GPU with the highest memory capacity in the market. This development marks a significant step in the competition between tech giants in the realm of artificial intelligence and data processing.

Unlike what one might expect from a modern graphics card, Intel has chosen LPDDR5X instead of GDDR7, a choice that may seem unusual at first glance. This type of memory is more commonly found in laptops and mobile devices, suggesting that Intel is prioritizing energy efficiency and performance over sheer speed.

## Why it matters This development is crucial not only for data centers but also for any industry relying on artificial intelligence. The ability to handle larger and more complex models could lead to significant advances in AI research and practical applications.

Furthermore, Intel's decision to use LPDDR5X might signal a shift in how companies approach hardware design for AI applications, prioritizing efficiency over maximum speed.

## What we know Intel has confirmed that its D-series GPU will feature a memory capacity of 480GB and is designed for AI workloads. It is also known to support models of up to 300 billion parameters, making it a powerful tool for developers and researchers.

## What remains unclear The full performance specifications of the D-series and its commercial availability have not yet been detailed. It is also unclear how LPDDR5X will compare to GDDR7 in practical AI applications.