GoKawiilIntel renames Crescent Island accelerator as GPU D-Series, reveals specs at OCP 2026
Intel announced at OCP 2026 that its previously code-named Crescent Island chip will ship as the GPU D-Series, disclosing a theoretical peak of 1.3 PFLOPS FP8 performance and 1.5 TB/s memory bandwidth. The air-cooled PCIe card runs at 350W TDP with up to 480GB of LPDDR5X memory, built on Intel's new Xe3P architecture, and can scale to eight GPUs per server.
GoKawiil's interpretation of the reporting above, not reported fact.
The specs show Intel is positioning this product for air-cooled, lower-power inference workloads rather than competing directly with Nvidia's liquid-cooled Rubin GPUs, which offer roughly 13 times the peak FP8 throughput per chip. This suggests Intel may be targeting a distinct market segment—cost-sensitive, infrastructure-constrained data centers running large-model inference—rather than chasing Nvidia's top-end training and inference performance tier.
- Crescent Island is now officially branded the Intel GPU D-Series, unveiled at OCP 2026.
- The chip targets 1.3 PFLOPS FP8 performance and 1.5 TB/s bandwidth at 350W TDP with up to 480GB LPDDR5X memory.
- It's built on Intel's new Xe3P architecture and designed for air-cooled servers scaling up to eight GPUs, unlike Nvidia's liquid-cooled Rubin GPUs.
Source: tomshardware.com — Jeffrey Kampman, 2026-10-12
Published there as: “Intel's Crescent Island AI accelerator officially powers new GPU D-Series products”
Read the original report → The summary and analysis above are GoKawiil's own, written from reporting by the source above. Facts and quotes belong to the original publisher.