TSMC has secured a substantial GPU production order from Intel

As Intel’s brand-new high-performance gaming graphics card brand, Intel Arc heralds the arrival of the first-generation Alchemist graphics card (DG2). According to Intel’s previously announced development roadmap, the Alchemist will be followed by Battlemage, Celestial, and Druid, totaling four generations of products. Diverging from the first three generations, the fourth-generation Druid will adopt a new Xe architecture to replace the existing Xe-HPG architecture.

As reported by ctee, TSMC has secured a substantial GPU production order from Intel, encompassing the Battlemage and Celestial successors to the Alchemist. Intel remains optimistic about the GPU market outlook, foreseeing continued demand driven by esports and artificial intelligence (AI) in the consumer sector, and growing requirements for AI and other compute-intensive tasks in the enterprise market.

Intel is currently advancing its GPU research and development projects as planned, with rumors suggesting a second-generation Battlemage graphics card utilizing a 4nm process will debut in the latter half of 2024, and a third-generation Celestial graphics card employing a 3nm process set for release in the latter half of 2026. Last year, reports claimed that the majority of Intel’s GPU development team had transitioned to working on the second-generation product, Battlemage, while early driver and software stack efforts were underway, and some team members had even begun preliminary work on the third-generation product, Celestial.

In a previous interview with the media, Intel researcher Tom Peterson explained that Intel had divided the Xe architecture into four categories for the Alchemist, based on the needs of niche markets, product scalability, and specific applications. Having learned valuable lessons, Battlemage will be simplified into two architectures—Xe2-LPG and Xe2-HPG—streamlining driver development, reducing costs, and enhancing compatibility.

To meet cost-cutting and efficiency improvement objectives, Intel had previously split its Accelerated Computing and Graphics (AXG) division into two parts. The consumer graphics team joined the Client Computing Group (CCG), while the accelerated computing team became part of the Data Center and AI Business Group (DCAI). Moving forward, the Xe2-LPG and Xe2-HPG architectures will be developed under the leadership of the Client Computing Group, while the Xe-HPC architecture will be overseen by the Data Center and AI Business Group.