Intel to Expand AI Accelerator Portfolio with New GPU

[Brady]

Intel to Expand AI Accelerator Portfolio with New GPU

newsroom.intel.com

About the GPU: The new data center GPU code-named Crescent Island is being designed to be power and cost-optimized for air-cooled enterprise servers and to incorporate large amounts of memory capacity and bandwidth, optimized for inference workflows.

Key features include:

• Xe3P microarchitecture with optimized performance-per-watt
• 160GB of LPDDR5X memory
• Support for a broad range of data types, ideal for “tokens-as-a-service” providers and inference use cases

Intel’s open and unified software stack for heterogeneous AI systems is currently being developed and tested on Arc Pro B-Series GPUs to enable early optimizations and iterations. Customer sampling of the new data center GPU code-named Crescent Island is expected in the second half of 2026.

Bets on this Xe3P-driven GPU being built by TSMC or Intel Foundry?

And what does this say about an Xe3P (confirmed to be “Celestial” branded) discrete consumer GPU?

 

[Daniel Nenni]

Intel to Expand AI Accelerator Portfolio with New GPU

newsroom.intel.com

Bets on this Xe3P-driven GPU being built by TSMC or Intel Foundry?
And what does this say about an Xe3P (confirmed to be “Celestial” branded) discrete consumer GPU?

I was told it is on Intel 18A.

 

[KevinK]

I

And what does this say about an Xe3P (confirmed to be “Celestial” branded) discrete consumer GPU?

Getting much harder to design silicon that can do double duty for efficient data center inference as well as consumer GPU. Most of the new AI 4 bit and 8 bit floating point datatypes and associated tensor processing, as well as the new transformer acceleration engines and memory management hardware, are useless for graphics. So lots of wasted hardware for discrete GPUS.

 

[siliconbruh999]

I was told it is on Intel 18A.

18AP iirc that's what NVL iGPU is getting same uArch